June 2026 Imaging Pearls - Educational Tools | CT Scanning | CT Imaging | CT Scan Protocols - CTisus

Imaging Pearls ❯ June 2026
-- OR --

3D and Workflow

  • All screening tests can lead to harm (eg, colon perforation during colonoscopy, bleeding from a subsequent biopsy). The goal is to screen people who have high enough pretest odds of important disease that the benefit outweighs the harm. If the pretest odds are too low, more people will be hurt than helped, regardless of how well intentioned the screening may be, which is why, for example, colon cancer screening is not performed in healthy 20-year-olds.
    Elective MRI Screening of the General Public-Buyer Beware.
    Davenport MS, Reeder SB.
    JAMA. 2026 May 6. doi: 10.1001/jama.2026.5888
  • The truth is that screening is complicated. Evidence-based screening programs(eg, colon cancer, breast cancer, prostate cancer, cervical cancer, lung cancer, cardiovascular disease) are confirmed  to save lives when targeted to people with high enough pretest odds of disease.However, this fact does not mean that all screening is helpful. There are numerous examples of well-intentioned screening efforts that have caused more harm than good due to invasive efforts to diagnose and treat findings that are determined later to be safe(ie, false-positive results) and from identification of diseases that create no benefit from early detection (ie, overdiagnosis), which is true even for many cancers.
    Elective MRI Screening of the General Public-Buyer Beware.
    Davenport MS, Reeder SB.
    JAMA. 2026 May 6. doi: 10.1001/jama.2026.5888
  • It is intuitive to assume that early detection always improves health. But this assumption is incorrect. Whole-body MRI will identify cancer in approximately 1 or 2 people out of 100 in the general adult population, but there is no evidence that it reduces mortalityor improves health. Most cancers detected incidentally in the general population are slow growing, and efforts to diagnose and treat them often create more harm than good, which is true for low-risk thyroid, kidney, and prostate cancer, as well as many other types.1On the other hand, highly aggressive cancers can grow so quickly that screening is unable to create a better outcome.
    Elective MRI Screening of the General Public-Buyer Beware.
    Davenport MS, Reeder SB.
    JAMA. 2026 May 6. doi: 10.1001/jama.2026.5888
  • Unproven health care interventions with meaningful risk of indirect harm require frank and accurate informed consent. Such an informed consent statement for the general population considering whole-body MRI might read as follows:
    No medical guideline recommends that you undergo this test.There is a 3 in 10chance we find something that creates uncertainty for you,1 which could result in anxiety, sleeplessness, financial strain, life disruption, more imaging, invasive procedures, or possibly surgery. Although cancer will be identified in 1 or 2 out of 100 people,9 finding cancer with this test is unlikely to help you because most types will below risk or already yadvanced.There are no studies showing that undergoing this test will improve the quality or length of your life. Based on what we know, if you undergo this test, you are more likely to be harmed than helped due to complications from efforts to diagnose and treat what we find and the low likelihoodwe find something that can improve the quality or length of your life. On average, any apparent benefit is likely illusory, even for many cancers.1-4 This test does not replace effective but often underused screening tests, such as mammography or colonoscopy. The cost of this test is borne by you out of pocket. Future care and costs will be your responsibility.
    Elective MRI Screening of the General Public-Buyer Beware.
    Davenport MS, Reeder SB.
    JAMA. 2026 May 6. doi: 10.1001/jama.2026.5888
  • Leaders affect their teams on a range of levels, whether it be microscopic exchanges among team members on a project, or in large-scale changes that shape the overarching culture of a company. Indeed, leaders seldom have the time to deliberately manage each interaction that happens in their team, often relying on defined culture and trusted coworkers to do the right things in those moments.
    The three snake rules: Lessons in leadership from an inside perspective.
    Brody WR, Fishman EK, Chu LC, Rowe SP, Smith CW
    Clin Imaging. 2026 May 9;135:110830. 
  • Although good leaders imbue values and principles that guide their team members wisely, where do we find leaders to do so? There has been a long-standing question regarding the origin of leadership that remains a relevant topic of conversation to this day. Are leaders born, or are they made? Still, despite the great leaders and inspirational figures that we have witnessed in our lives, many cannot decide where good leadership begins. You may know a peer or loved one who has always been comfortable making decisions, motivating others, and promoting a set of guiding values, to whom you would reasonably attribute the label of a “born” leader. On the other hand, you may have grown up with a shy person who now occupies a highly vocal, public leadership position in a multinational company, and you would say they were “made” a leader.
    The three snake rules: Lessons in leadership from an inside perspective.
    Brody WR, Fishman EK, Chu LC, Rowe SP, Smith CW
    Clin Imaging. 2026 May 9;135:110830. 
  • As a contrarian to the classical question of whether leaders are born or made, my contention is that leaders are found. The correct question to ask when trying to match an appropriate candidate for leadership is not “is this person a leader?”, but rather “does this person have the particular qualities that are most needed in this moment?” In the search for leaders, we often focus on a candidate's traits themselves and fail to recognize the gravity of the context in which the candidate will lead. A leader's qualities are effective only when team members are open and responsive to the leader's skills and approach. Similarly, choosing the right leader requires a thorough understanding of the people, objectives, and values that the leader will manage.
    The three snake rules: Lessons in leadership from an inside perspective.
    Brody WR, Fishman EK, Chu LC, Rowe SP, Smith CW
    Clin Imaging. 2026 May 9;135:110830. 
  • A great question I was recently asked was “If you came back now, 40 years later, and had to run a radiology department, would you still be successful?” To this question, my chief concern was the personnel within the field and how they have changed over the years. Radiologists have, over time, become increasingly disgruntled with expanding workloads rewarded by unpredictable compensation—and understandably so. Radiologists are already frustrated in their workplace, and I believed in the right circumstances I could give them hope. My self-assessment of my ability to lead the current generation of radiologists is not focused on my own traits or skillset, but rather on the landscape of radiology and how I might align with the organization I would be entering. I would examine the people I would be leading, and whether I could be a fit for them – it would not be a purely introspective thought process.
    The three snake rules: Lessons in leadership from an inside perspective.
    Brody WR, Fishman EK, Chu LC, Rowe SP, Smith CW
    Clin Imaging. 2026 May 9;135:110830. 
  • In selecting leadership, the qualities of those being led are equally important to the qualities of the leader. Those two cannot be divorced from one another but often are. Ultimately, the success of an organization depends not on finding a leader with the “right” qualities in abstract, but on matching the qualities of a leader to the specific needs and temperament of a team at a given time. Leadership is less about discovering innate greatness and more about the careful alignment of people, context, and shared values, which requires a fundamental understanding of both those being led and the leader themselves
    The three snake rules: Lessons in leadership from an inside perspective.
    Brody WR, Fishman EK, Chu LC, Rowe SP, Smith CW
    Clin Imaging. 2026 May 9;135:110830. 
  • In the current era of radiology, there is substantial burnout, which may in part be related to suboptimal leadership. Those leading your organization may not be inherently bad leaders, but they may not be the correct leaders for your team. The principles outlined in this article offer a framework for better understanding that mismatch and breaking the pattern that has progressively plagued radiologists with discontentment and frustration. Those appointing leaders within radiology—and the current leaders thereof—must recognize that the qualities needed to lead a demoralized, overextended workforce differ from those suited to other contexts. It is crucial that emerging and existent leaders within radiology advocate for sustainability and address systemic issues within the field.
    The three snake rules: Lessons in leadership from an inside perspective.
    Brody WR, Fishman EK, Chu LC, Rowe SP, Smith CW
    Clin Imaging. 2026 May 9;135:110830. 
  • Ultimately, physicians must reclaim leadership to align the logistical framework of their departments with the missions of medicine. More than anyone else, physicians understand the criteria for success in the tripartite mission of academic medicine, balancing care delivery with research and teaching, whereas administrators typically focus on the singular mission of optimizing financial performance. For medicine to succeed, whether in the setting of private practice, privademics, or classical academics, its leadership and driving force must be physicians who truly understand the inner workings of the field. We may be living in a time when a snake has presented itself, but we have yet to take decisive action to kill it. The widespread low morale among radiologists  is a multifaceted problem, but physician leadership whose qualities and experiences are congruent with the needs of the specialty could begin to transform this “snake” into a time of revitalization within our field.
    The three snake rules: Lessons in leadership from an inside perspective.
    Brody WR, Fishman EK, Chu LC, Rowe SP, Smith CW
    Clin Imaging. 2026 May 9;135:110830. 
  • Leadership is not about choosing either structure or freedom. It is about knowing how to move between them. Structure provides stability in a team: clarity of purpose, defined roles, and shared values. Freedom, on the other hand, provides possibility: creativity, innovation, and authenticity. Too much structure hinders progress and inventive thinking. Too much freedom results in chaos and downfall. Good leadership lives in the space between the two.
    The Leadership Dance: Swaying Between Structure and Freedom.
    Lynch AH, Fishman EK, Chu LC, Rowe SP, Smith CW.
    J Am Coll Radiol. 2026 Mar 25:S1546-1440
  • Radiology is a discipline rooted in deep structure. Imaging acquisition is performed according to strict protocol, reporting templates guide communication, and imaging features are synthesized meticulously to narrow differential diagnoses. Those systems are essential, indeed, ensuring quality, safety, and consistency in our specialty. However, structure alone is not sufficient. Every radiologist will encounter cases that cannot be neatly categorized. Ambiguous imaging features, changes in clinical course, and interdisciplinary collaboration may all obfuscate the radiologist’s interpretation. If we cling too tightly to structure, we risk neglecting the more complicated cases and narrowing our diagnostic breadth. If we abandon structure, we risk inconsistency. In practice, we must not choose to adhere to either structure or freedom, but to understand how to balance them.
    The Leadership Dance: Swaying Between Structure and Freedom.
    Lynch AH, Fishman EK, Chu LC, Rowe SP, Smith CW.
    J Am Coll Radiol. 2026 Mar 25:S1546-1440
  • Furthermore, the integration of artificial intelligence (AI) makes this balance even more relevant. Indeed, AI can provide unprecedented structural support by radiomic analysis, enhancing efficiency, and automating measurements. However, AI lacks the “freedom” and human insight that synthesize patient-specific and clinical context to reason through more nuanced cases. Technology should amplify the radiologist’s expertise structurally, not replace its freedom and multifaceted thinking. The disciplined improvisation we provide to bridge radiological data and clinical thinking is invaluable, and we must not sacrifice that when  clinging to the new structural capabilities of AI.
    The Leadership Dance: Swaying Between Structure and Freedom.
    Lynch AH, Fishman EK, Chu LC, Rowe SP, Smith CW.
    J Am Coll Radiol. 2026 Mar 25:S1546-1440
  • Furthermore, leadership within radiology departments requires a similar balance of structure and freedom. Productivity metrics and efficient protocols create necessary structure. But innovation and long-term sustainability require spaces reserved for autonomy and curiosity. Leaders of their departments must carefully cultivate their environments to facilitate both efficiency and innovation. And in its current state, academic radiology is facing record levels of burnout. The methods and systems that brought us here may not be the same ones that carry us forward. We must be willing to unlearn, adapt, and choreograph new approaches when circumstances demand. Mastery in radiology and leadership lives in the blend of structure and freedom. We must be prepared to move between the two to not only provide excellent care and leadership currently, but to shape environments where others can grow into future leaders as well.
    The Leadership Dance: Swaying Between Structure and Freedom.
    Lynch AH, Fishman EK, Chu LC, Rowe SP, Smith CW.
     J Am Coll Radiol. 2026 Mar 25:S1546-1440
Cardiac


  • Detecting Asymptomatic Coronary Artery Disease Using Coronary Artery Calcium Score and Coronary Computed Tomography Angiography
    Showly Nicholson , Dhrubajyoti Bandyopadhyay, Sandeep Hedgire
    Radiol Clin N Am 64 (2026) 591–603 
  • • Coronary CT angiography (CCTA) is used as initial imaging in patients to detect high-risk noncalci fied plaques, and obstructive and nonobstructive disease, even in asymptomatic individuals.
    • CCTA provides a comprehensive anatomic and morphologic assessment of coronary atheroscle rosis, including calcified and noncalcified plaques, high-risk features, and total plaque burden but is not yet recommended for routine screening in asymptomatic populations.
    • Future approaches combine CACS, selective CCTA and clinical risk scoring, leveraging emerging  technologies (photon counting detector computed tomography, quantitative plaque analysis, and artificial intelligence-enabled plaque assessment).
    Detecting Asymptomatic Coronary Artery Disease Using Coronary Artery Calcium Score and Coronary Computed Tomography Angiography
    Showly Nicholson , Dhrubajyoti Bandyopadhyay, Sandeep Hedgire
    Radiol Clin N Am 64 (2026) 591–603 
  • On a calcium score CT, CAC is defined as any lesion with an attenuation of ≥130 Hounsfield units (HU) and an area of ≥1 square millimeter. The most widely used and validated method of scoring CAC is the Agatston Method, which generates an Agatston Score by multiplying the area of CAC by a predetermined density weighting factor (DWF) based on maximal HU within the region of interest (Area × DWF), then summating the scores across all lesions. Additional methods of calculating CACS, including the calcium volume score and relative calcium mass score, are beyond the scope of this publication. 8 The Agatston Score remains integral to risk stratification and shared decision-making, either by using absolute values across established cutoff points to classify a patient’s coronary artery disease (CAD) risk, or by adjusting values for clinical risk factors to determine a percentile ranking relative to population distributions.
    Detecting Asymptomatic Coronary Artery Disease Using Coronary Artery Calcium Score and Coronary Computed Tomography Angiography
    Showly Nicholson , Dhrubajyoti Bandyopadhyay, Sandeep Hedgire
    Radiol Clin N Am 64 (2026) 591–603 
  • A CACS greater than 300 has been associated with a 10-fold higher risk of CHD compared to patients without coronary artery calcium.  Elevaetd CAC levels are linked to increased risk of myocardial infarction (MI), coronary revascularization, atrial fibrillation, all-cause mortality, and even noncardiovascular mortality. 9 Importantly, CACS has been shown to be a stronger predictor of CHD events than traditional risk factors. Clinically, identification of CAC leads to higher initiation or continuation rates of lipid-lowering therapies, antihypertensive medications, aspirin use, and adoption of healthier lifestyle behaviors such as dietary modification and increased physical activity. 11 Current guidelines recommend CAC scoring to guide statin initiation when there is uncertainty in patients at intermediate (≥7.5% to <20%) 10-year ASCVD risk, and in select patients at borderline (5% to <7.5%) risk. This represents class IIa, level A recommendation.
    Detecting Asymptomatic Coronary Artery Disease Using Coronary Artery Calcium Score and Coronary Computed Tomography Angiography
    Showly Nicholson , Dhrubajyoti Bandyopadhyay, Sandeep Hedgire
    Radiol Clin N Am 64 (2026) 591–603 
  • In asymptomatic adults, the absence of CAC carries a high negative predictive value for cardiovascular outcomes over the subsequent 5 years. 14 Large cohort studies, including data from the CACC, have demonstrated that patients with CAC = 0 experience markedly lower rates of cardiovascular and all-cause mortality compared with those with any measurable calcification (individuals with CAC 1–10 experienced a 1.4-fold higher risk of death from any cause compared to individuals with CAC = 0). 13 For low-risk individuals with calcium score 0, the ACC/AHA recommends repeating CAC screening after 5 to 10 years, whereas SCCT recommends repeating in 5 years where increase in CACS would favor intensification of preventive therapies and consider repeating it every 3 to 5 years if CAC greater than 0.
    Detecting Asymptomatic Coronary Artery Disease Using Coronary Artery Calcium Score and Coronary Computed Tomography Angiography
    Showly Nicholson , Dhrubajyoti Bandyopadhyay, Sandeep Hedgire
     Radiol Clin N Am 64 (2026) 591–603 
  • There is limited experimental evidence elucidating the benefit of CCTA as a screening modality in asymptomatic patients. Although more research is needed to determine the usefulness of CCTA in reducing cardiac outcomes, available data suggest that CCTA can identify occult CAD in this population. Choi and colleagues reported that 22% of 1000 asymptomatic, middle-aged participants had atherosclerotic plaque, with 5% showing significant stenosis (≥50%) and 2% severe stenosis (≥75%). 34 In an asymptomatic, intermediate-risk cohort, Di Cesare and colleagues found that over 25% of patients had moderate-to-severe stenoses, with more than 24% having 3-vessel or 4-vessel disease.
    Detecting Asymptomatic Coronary Artery Disease Using Coronary Artery Calcium Score and Coronary Computed Tomography Angiography
    Showly Nicholson , Dhrubajyoti Bandyopadhyay, Sandeep Hedgire
    Radiol Clin N Am 64 (2026) 591–603 
  • CAC screening is a simple, low-dose tool that refines risk stratification in asymptomatic individuals, but it cannot detect high-risk noncalcified plaques. In contrast, CCTA offers comprehensive anatomic and morphologic assessment of coronary atherosclerosis, including detection of both calcified and noncalcified plaques, characterization of high-risk features, and quantification of total plaque burden. While not yet recommended for routine screening of asymptomatic populations, CCTA holds promise in advancing precision medicine, particularly with emerging technologies such as PCCT, quantitative plaque analysis, and AI enabled plaque assessment. Ongoing large-scale randomized trials, including SCOT-HEART 2, will clarify whether early use of CCTA in asymptomatic individuals improves cardiovascular outcomes. Until then, the integration of CACS and selective use of CCTA, combined with clinical risk scoring, offers the most balanced approach for identifying subclinical disease and guiding preventive strategies. Future updates of CAD-RADS (CAD-RADS 3) may incorporate automated plaque quantitation, further enhancing risk prediction and standardized reporting.
    Detecting Asymptomatic Coronary Artery Disease Using Coronary Artery Calcium Score and Coronary Computed Tomography Angiography
    Showly Nicholson , Dhrubajyoti Bandyopadhyay, Sandeep Hedgire
    Radiol Clin N Am 64 (2026) 591–603 
Chest

  • PAVM Lung
    Pulmonary Arteriovenous Malformation (PAVM)—also referred to as a pulmonary arteriovenous fistula (PAVF)—is an abnormal direct communication between a pulmonary artery and a pulmonary vein, bypassing the normal intervening capillary bed.
    This creates a high-flow, low-resistance right-to-left shunt, which has significant implications for systemic oxygenation and embolic risks.
  • PAVM and Syndromes
    Hereditary Hemorrhagic Telangiectasia (HHT / Osler-Weber-Rendu Disease): * This is the single most important clinical association. Approximately 80%–90% of PAVMs occur in patients with HHT. Conversely, about 30%–50% of patients with HHT will develop a PAVM.
    PAVMs in patients with HHT are much more likely to be multiple and bilateral compared to sporadic cases, which are typically solitary.
  • PAVM and Complications
    The Triad of Complications: Because the capillary filter is bypassed, patients are at high risk for:
    Paradoxical Embolism: Thrombi or bacteria cross directly into the systemic circulation, leading to stroke or brain abscesses.
    Hypoxemia: Cyanosis, clubbing, and dyspnea due to the right-to-left shunt of deoxygenated blood.
    Hemorrhage: Massive hemoptysis or hemothorax if the thin-walled malformation ruptures.
  • The Triad Component Appearance
    A Feeding Artery: A dilated, tortuous pulmonary artery branches off the main arterial tree and heads directly toward the lesion.
    An Aneurysmal Nidus / Sac: A well-circumscribed, intensely enhancing vascular nodule or sac where the vessels meet.
    A Draining Vein: A dilated, early-enhancing pulmonary vein exiting the sac and traveling back toward the left atrium.
  • Pulmonary Septic Emboli
    On chest CT, septic pulmonary embolism (SPE) presents with a highly characteristic constellation of findings. Because these lesions are caused by an embolic "shower" of infected thrombi, they typically mirror blood flow patterns and appear in various stages of evolution.
  • Septic Emboli CT Findings
    Multiple Lung Nodules (80%–90% of cases): * Distribution: Predominantly bilateral, peripheral (usually within 2 cm of the pleura), and basilar, reflecting gravity-dependent pulmonary blood flow.
    Size: Typically small (ranging from 0.5 cm to $3 cm, though they can grow larger. They often vary in size, representing recurrent embolic events over time.
    Margins: May have indistinct borders or a surrounding ground-glass halo sign, which indicates localized hemorrhage or infarction.
  • Septic Emboli CT Findings
    Rapid Cavitation (~50% of cases):
    The infected emboli cause focal ischemia and microabscess formation, leading to rapid cavitation (often within 24 to 48 hours).
    On CT, you will see a mix of solid nodules, thick-walled cavities, and thin-walled cavities representing different stages of infection. Air-fluid levels are relatively uncommon compared to primary lung abscesses.
  • Differential Diagnosis
    Granulomatosis with Polyangiitis (GPA): Also causes multiple cavitating peripheral nodules, but lesions evolve much more slowly and do not respond to antibiotics.
    Cavitary Pulmonary Metastases: Classically from squamous cell carcinomas (e.g., head and neck) or sarcomas. These will also display a feeding vessel sign but lack acute clinical symptoms of sepsis and do not rapidly change from day to day.
    Rheumatoid Nodules: Typically subpleural and can cavitate, but are indolent and usually associated with known subcutaneous nodules and systemic rheumatoid arthritis.
  • Pulmonary Hamartoma
    most common benign lung tumor, accounting for approximately 75% of all benign lung nodules. They are mesenchymal tumors composed of an abnormal mixture of tissue elements typically found in the lung—predominantly mature cartilage, but also fat, fibrous tissue, and respiratory epithelium.
  • CT of Pulmonary Hamartoma
    Classic Appearance: A well-circumscribed, smooth, or lobulated solitary pulmonary nodule.
    Size: Typically small, with over 90% measuring less than 4 cm in diameter. They generally grow very slowly (around 3 mm per year, if at all).
  • CT of Pulmonary Hamartoma
    Internal Composition
    Fat Attenuation (30%–50% of cases): The presence of fat within a well-defined nodule is highly specific for a hamartoma. On CT, this registers as localized regions of low attenuation, typically between -40 HU and -120 HU.
    Calcification (20%–30% of cases): Calcification becomes more common as the lesion grows.
    --- "Popcorn" Calcification: This is the classic, pathognomonic description—coarse, irregular calcification reminiscent of popped corn.
    --- Other patterns include central, diffuse, or punctate calcified specks.
    The Combination: Finding both fat and popcorn calcification within a smooth, solitary lung nodule is entirely diagnostic of a pulmonary hamartoma.
  • Pulmonary Hemorrhage CT Findings
    Ground-Glass Opacities (GGOs): * This is the most common and earliest finding.
    It represents partial filling of the alveoli with blood. On CT, it appears as a hazy increase in lung attenuation where the underlying bronchovascular margins are still visible.
    Consolidation: * As alveoli become completely filled with blood, the GGOs progress to dense consolidation.
    Unlike GGOs, consolidation completely obscures the underlying pulmonary vessels. Air bronchograms are frequently seen within these areas
  • Pulmonary Hemorrhage CT Findings
    "Crazy-Paving" Pattern: * This occurs when there is a combination of ground-glass opacity with thickened interlobular and intralobular septa.
    In acute hemorrhage, this thickening is typically due to fluid or blood filling the interstitium alongside the alveoli.
    Centrilobular Nodules: * Ill-defined, small centrilobular nodules may appear if blood pools primarily within the smaller bronchioles and adjacent alveoli.
  • Pulmonary Hemorrhage Distribution
    Diffuse & Bilateral
    Widespread involvement, often sparing the extreme lung bases or apices.Diffuse Alveolar Hemorrhage (DAH) due to vasculitis (e.g., Granulomatosis with Polyangiitis, Goodpasture syndrome), systemic lupus erythematosus (SLE), or drug toxicities.
  • Pulmonary Hemorrhage Distribution
    Because the CT features of pulmonary hemorrhage overlap significantly with hydrostatic pulmonary edema, atypical infections (e.g., PCP or viral pneumonias), and alveolar proteinosis, clinical correlation is critical. The rapid clearing of opacities on follow-up imaging (often within days) is highly characteristic of hemorrhage compared to infectious or inflammatory processes.
Deep Learning


  • Trust, Scrutiny, or Collaboration: A Performance- Based Framework for Human–AI Interaction in Medicine
    Laura Zwaan  et al.
    NEJM AI 2026;3(5)
  • Artificial intelligence (AI) is already transforming medical innovation and drug discovery in ways that nare grabbing headlines, for example, by predicting protein structures that can be targeted by new drugs or by identifying novel antibiotic classes from large chemical libraries. But discoveries in the laboratory are often not sufficient to make new treatments available to address crucial patient needs.1 The journey from bench to bedside is delayed by the time-consuming process of enrolling patients in trials and the time it takes to measure the ultimate outcomes. AI is poised to dramatically speed up both of these processes, lowering barriers to breakthrough innovation.
    AI Will Accelerate Drug Discovery by Accelerating Clinical Evidence
    Katherine Baicker, Ziad Obermeyer
    JAMA Health Forum. 2026;7(4):e261596.
  • AI offers a different, empirical approach to finding surrogate outcomes, one that does not depend on causal understanding but rather broad measurement of causal factors—even ones that are as yet unknown to humans. Consider a thought experiment: a comprehensive dataset of variables describing the body’s current physiology collectively predicts a future outcome with high accuracy. If that readout is sufficiently comprehensive, a treatment that changes the future outcome by changing the current physiology will also change the readout. And, critically, even if it’s not known which elements are changing or why, as recent work in statistics and econometrics has shown, a predicted outcome based on the readout will change proportional to the eventual effect of the drug. Such surrogate indexes using physiological data—images,waveforms, complex time series—can only be constructed using machine learning, as traditional methods cannot handle such high dimensional data.
  • The AI revolution will surely dramatically accelerate medical discovery. But protein folding and molecular design alone will not get new treatments to patients without fixing the longest lags: enrolling and evaluating clinical trials. Despite important caveats, the use of AI for clinical trials— predictions to speed and better target enrollment, generate finely calibrated outcomes from readily available data, and determine safety and efficacy over shorter time horizons—can reduce costs, accelerate innovation, and unlock huge benefits to patients.
    AI Will Accelerate Drug Discovery by Accelerating Clinical Evidence
    Katherine Baicker, Ziad Obermeyer
    JAMA Health Forum. 2026;7(4):e261596.
  • There are of course important cautions and limitations of this approach. An AI surrogate is still a surrogate, and clinical evaluation against true endpoints remains essential. A surrogate will fail to predict a drug’s true effect if the drug changes the relationship between the predictor and the ultimate outcome, or if the drug effects are not measured in the current data (eg, idiosyncratic late toxic effects). Furthermore, if training data reflect historical disparities, the surrogate can reproduce them, making it critical to check performance across subgroups. There is only one way to know whether a surrogate is causally valid: run the trial and measure the effect on real outcomes. Nonetheless, surrogates can be very helpful in the lead-up to this definitive trial. Researchers and sponsors must prioritize which drug development programs to invest in and how to design the next generation of clinical trials for “undrugged” conditions; and regulators must consider how to address incorporation of AI predictions into trial design to support pathbreaking research on high-burden conditions with no available treatment.
    AI Will Accelerate Drug Discovery by Accelerating Clinical Evidence
    Katherine Baicker, Ziad Obermeyer
    JAMA Health Forum. 2026;7(4):e261596.
  • No one predicted that artificial intelligence (AI) would first surpass physicians on empathy—diagnostic accuracy, perhaps; imaging interpretation, certainly. Empathy was supposed to be the last refuge of the clinician; however, chatbot responses now outscore physician responses on empathy across specialties in blinded, text-based evaluations. A louder public narrative has fastened onto that finding, declaring medical schools will soon be pointless and the physician an expensive anachronism in a system that no longer needs one.Those are not clinical conclusions; they are commercial extrapolations being amplified by interests that benefit from framing medicine as replaceable. The generation deciding whether to enter the medical profession may believe them.
    Artificial Intelligence Is Not the End of the Physician.
    Martinelli C, Carnevale V, Ercoli A, Giordano A
    JAMA. 2026 Apr 29. doi: 10.1001/jama.2026.4356. 
  • The administrative work accumulated regulation by regulation and click by click until the medical profession no longer recognized what it had lost. Over 4 decades, the practice of medicine buried physicians under this new process with an increased focus on indirect patient care. In ambulatory practice, physicians spend 49%of their office day on electronic health record and desk work and only 27%in direct clinical contact.
    Artificial Intelligence Is Not the End of the Physician.
    Martinelli C, Carnevale V, Ercoli A, Giordano A
    JAMA. 2026 Apr 29. doi: 10.1001/jama.2026.4356. 
  • The prevailing prediction is that AI will finish what administrative medicine began and substitute the physician entirely, but that gets the technology exactly backward. AI cannot replace the physician, but it can remove the burden of what replaced the physician’s clinical time, which were fer to as the excavation .Documentation, coding, triage, and pattern recognition are essential but substitutable. A physical examination, a hand on the shoulder, and the willingness to stay with patients and listen to their fears are irreducible and non delegable.
    Artificial Intelligence Is Not the End of the Physician.
    Martinelli C, Carnevale V, Ercoli A, Giordano A
    JAMA. 2026 Apr 29. doi: 10.1001/jama.2026.4356. 
  • AI does not humanize medicine, but it can remove one of the most durable excuses for failing to do so. Each administrative task the algorithm absorbs is time the profession can reclaim for the clinical work that made medicine a vocation.The bedside was always the foundation, but in building everything above it we simply forgot that we were standing on it. AI is not the end of the physician—it is the return of every reason to be one.
    Artificial Intelligence Is Not the End of the Physician.
    Martinelli C, Carnevale V, Ercoli A, Giordano A
    JAMA. 2026 Apr 29. doi: 10.1001/jama.2026.4356. 
  •  Whether a physician should defer to, scrutinize, or collaboratively engage with AI depends on two key dimensions: the relative accuracy of humans and AI on a given task, and the degree to which their errors are complementary. We propose a framework mapping these dimensions onto four interaction zones: human-dominant, AI-dominant, hybrid review, and disagreement resolution. Each requires a distinct clinical workflow strategy. This framework is intentionally dynamic: As AI performance evolves and physician expertise develops, optimal interaction strategies must be recalibrated accordingly. Trust calibration, not blanket skepticism, is the appropriate goal.
    Trust, Scrutiny, or Collaboration: A Performance- Based Framework for Human–AI Interaction in Medicine
    Laura Zwaan  et al.
    NEJM AI 2026;3(5)
  • In the human-dominant zone — where human accuracy substantially exceeds AI — the rational strategy is to act as if automation bias was at play — trust the human, treat AI output with skepticism, and preserve clinical autonomy. This may describe some open-ended clinical reasoning tasks, particularly those requiring patient context, values, and clinical intuition that current AI does not fully capture. When error complementarity is high, AI can be used to flag human errors.
    Trust, Scrutiny, or Collaboration: A Performance- Based Framework for Human–AI Interaction in Medicine
    Laura Zwaan  et al.
    NEJM AI 2026;3(5)
  • The most strategically valuable zone, however, may be the hybrid review zone — where accuracy is roughly comparable, but errors are complementary. Here, neither the human nor the AI alone is sufficient, but together each can catch what the other misses. This zone is most consistent with the popular aspiration for human–AI collaboration and offers a meaningful opportunity for performance improvement, but it requires careful workflow design.6 The human users should review AI output for missed findings the AI is prone to, and the AI should surface patterns the human is prone to overlook.
    Trust, Scrutiny, or Collaboration: A Performance- Based Framework for Human–AI Interaction in Medicine
    Laura Zwaan  et al.
    NEJM AI 2026;3(5)
  • Finally, there is the disagreement resolution zone. Here, accuracy is roughly comparable, but errors overlap substantially. The AI rarely offers the physician information they do not already have and, thus, does not reduce uncertainty in a straightforward way. No single strategy applies across all cases in this category. Instead, decisions must be made on a case-by-case basis, for example by identifying task features, case features, or confidence signals. These can help characterize the nature of uncertainty and determine when the human or AI is more likely to be correct, particularly by matching the response to the type of uncertainty encountered.
    Trust, Scrutiny, or Collaboration: A Performance- Based Framework for Human–AI Interaction in Medicine
    Laura Zwaan  et al.
    NEJM AI 2026;3(5)
  • The study by Qazi and colleagues makes an important contribution, as it quantifies the costs of AI errors in a controlled clinical reasoning setting.3 The finding that more experienced physicians were more susceptible than less experienced physicians is provocative and clinically important, potentially reflecting the very fluency-induced heuristic processing and authority bias that makes sophisticated AI output harder to resist. But the policy response should not be uniform exhortations to be more critical of AI. It should be a contextually calibrated, evidence-based determination of when trust, scrutiny, or structured human–AI collaboration is most appropriate, along with recognition that, as AI tools evolve, so too must the interaction strategies designed to help physicians use them safely.
    Trust, Scrutiny, or Collaboration: A Performance- Based Framework for Human–AI Interaction in Medicine
    Laura Zwaan  et al.
    NEJM AI 2026;3(5)
  • Although there have been extensive discussions about these clinical competencies, expression of appropriate self-doubt hasn’t been included as a core competency in medical education or embedded into AI product development. Recognizing and acknowledging when the degree of selfdoubt crosses a threshold to a state of “I don’t know” is the first step in operationalizing the commitment to “do no harm”; the crossing of this threshold indicates that critical thinking is necessary. The ability to say “I don’t know” — particularly in highstakes scenarios — may be the truest hallmark of an expert.
    Can AI Say “I Don’t Know”?
    Andrea Sikora, Leo A. Celi, M.D., Raja‑Elie E. Abdulnour,
    n engl j med 394;19 May 2016
  • We believe clinical AI systems should be trained to express calibrated uncertainty in ways that complement and indicate the need for epistemic humility in human users — and their performance in these areas should be assessed. Systems without the ability to execute uncertainty behaviors will continue to produce persuasive fiction in precisely the moments when patients need their clinicians to pause and ask for help. Contemporary LLMs have passed many Turing tests, but will they pass this modern test of not knowing? We don’t know.
    Can AI Say “I Don’t Know”?
    Andrea Sikora, Leo A. Celi, M.D., Raja‑Elie E. Abdulnour,
    n engl j med 394;19 May 2016
  • Artificial intelligence and computational modelling are increasingly being applied in PDAC research. Multiomic AI platforms have shown potential to predict PDAC survival outcomes , deep-learning approaches can link histopathological features with molecular and proteomic states , and computational drug-screening models have been used to nominate candidate synergistic drug combinations in pancreatic cancer . Although these approaches remain largely investigational, they may help prioritise future molecularly selected trials and accelerate precision oncology in PDAC .
    KRAS and Beyond: Emerging Targeted and Molecularly Stratified Strategies in Pancreatic Ductal Adenocarcinoma
    Alicia Y. Lefas , Hazel Lote * and Ian Chau
    Precis. Oncol. 2026, 1, 9
  • Pancreatic ductal adenocarcinoma (PDAC) is a highly lethal malignancy, with rising incidence and a 5-year survival rate of 13%. Late presentation, early metastasis, and intrinsic resistance constrain the efficacy of cytotoxic chemotherapy, which remains the backbone of PDAC treatment, with only modest survival gains and resistance nearly universal. Although KRAS mutations dominate tumour biology (~90% of cases), PDAC is a heterogeneous disease with distinct molecular subtypes that confer differential therapeutic vulnerabilities. Advances in comprehensive molecular profiling have catalysed a paradigm shift toward precision oncology in PDAC. In KRAS-mutant PDAC, mutation-specific inhibitors have established proof-of-concept, particularly in KRAS G12C disease, while nextgeneration approaches including KRAS G12D inhibitors, RAS-“ON” inhibitors, proteolysistargeting chimeras (PROTACs), and KRAS-targeted vaccine strategies are expanding the therapeutic landscape.
    KRAS and Beyond: Emerging Targeted and Molecularly Stratified Strategies in Pancreatic Ductal Adenocarcinoma
    Alicia Y. Lefas , Hazel Lote * and Ian Chau
    Precis. Oncol. 2026, 1, 9
  • Combination strategies targeting upstream and downstream effectors of the RAS–MAPK pathway are also being explored to enhance the depth and durability of response. In parallel, KRAS-wild-type PDAC has emerged as a molecularly distinct subgroup enriched for rare but actionable alternative oncogenic fusion drivers including NRG1, NTRK, RET, ALK, and FGFR. Additional molecularly directed strategies targeting HER2 alterations, BRAF mutations, EGFR-dependent signalling, and tumour selectively exposed surface antigens such as CLDN18.2 are under investigation across PDAC irrespective of KRAS mutation status. Synthetic lethal approaches, including targeting the PRMT5/CDKN2A/MTAP axis, represent a further emerging therapeutic strategy. Germline homologous recombination repair defects, particularly involving BRCA1/2 and PALB2, further define clinically important subsets with sensitivity to platinum chemotherapy and PARP inhibition.
    KRAS and Beyond: Emerging Targeted and Molecularly Stratified Strategies in Pancreatic Ductal Adenocarcinoma
    Alicia Y. Lefas , Hazel Lote * and Ian Chau
    Precis. Oncol. 2026, 1, 9
  • REDMOD is an automated, mechanistically grounded, longitudinally stable, externally validated AI that surpasses radiologists for PDA detection at its visually occult pre-diagnostic stage. These attributes position it for prospective validation in high-risk cohorts, a necessary step towards shifting the paradigm from late-stage symptomatic diagnosis to proactive pre-clinical interception.
    Next-generation AI for visually occult pancreatic cancer detection in a low-prevalence setting with longitudinal stability and multi-institutional generalisability.
    Mukherjee, S., et al.
    Gut. 2026 DOI: 10.1136/gutjnl-2025-337266.
  • On an independent test set (n=493), REDMOD identified occult PDA (AUC 0.82; 73.0% sensitivity) at a median 475-day lead time. This represented nearly twofold higher sensitivity than radiologists (38.9%; p<0.001), which grew to nearly threefold (68.0% vs 23.0%) at >24 months lead time. REDMOD showed strong longitudinal stability (90–92% concordance) and generalisable specificity across multi-institutional (81.3%; n=539) and public (87.5%; n=80) datasets. Mechanistic analyses confirmed predictive power derived principally from multi-scale wavelet-filtered textural features (90% of selected signature), which outperformed unfiltered features (AUC 0.82 vs 0.74; p=0.007) in capturing subvisual architectural disruptions
    Next-generation AI for visually occult pancreatic cancer detection in a low-prevalence setting with longitudinal stability and multi-institutional generalisability.
    Mukherjee, S., et al.
    Gut. 2026 DOI: 10.1136/gutjnl-2025-337266.
  • * The performance of REDMOD surpasses that of radiologists, demonstrating nearly double the sensitivity (73.0% vs 38.9%; p<0.001) for detecting visually occult PDA, with an advantage that increases to nearly threefold for cases detected more than 24 months prior to diagnosis.
    * For the first time, this work establishes the longitudinal stability of a pre-clinical radiomic signal (90–92% test–retest concordance) and confirms the robust specificity of the model across independent multi-institutional (81.3%) and public (87.5%) datasets.
    Next-generation AI for visually occult pancreatic cancer detection in a low-prevalence setting with longitudinal stability and multi-institutional generalisability.
    Mukherjee, S., et al.
    Gut. 2026 DOI: 10.1136/gutjnl-2025-337266.
  • * REDMOD offers an automated and longitudinally stable anchor for multicentre prospective studies in risk-enriched early detection programmes, with pre-specified thresholds and an observe–retest strategy for intermediate results.
    * REDMOD could operate as a triage and longitudinal monitoring tool for CT requests triggered by NICE guidelines, aligning with NHS pathways while awaiting prospective clinical evidence and cost-effectiveness.
    * The release of a fully automated and externally benchmarked pipeline offers a reproducible platform for method comparisons, bias audits and cost-effectiveness modelling to guide policy.
    Next-generation AI for visually occult pancreatic cancer detection in a low-prevalence setting with longitudinal stability and multi-institutional generalisability.
    Mukherjee, S., et al.
    Gut. 2026 DOI: 10.1136/gutjnl-2025-337266.
  • Our group achieved a breakthrough by demonstrating that machine learning (ML) models, using radiomic features derived from pre-diagnostic CT scans—incidental CTs acquired months to years prior to clinical diagnosis—exhibit high accuracy in identifying such subclinical alterations indicative of carcinogenesis. These models substantially outperformed radiologists in detecting visually occult PDA (stage 0 disease) with a considerable lead time before clinical diagnosis.
    Next-generation AI for visually occult pancreatic cancer detection in a low-prevalence setting with longitudinal stability and multi-institutional generalisability.
    Mukherjee, S., et al.
    Gut. 2026 DOI: 10.1136/gutjnl-2025-337266.
  • This study validates REDMOD as a fully automated AI framework capable of identifying the imaging signatures of stage 0 PDA in normal pancreas, achieving this with substantial lead times and performance superior to expert radiologists. The demonstrated ability of the framework to consistently detect these occult signals on a large clinically-oriented dataset, combined with its high longitudinal stability and validated specificity, establishes a robust foundation for AI-augmented early detection. This work overcomes key barriers in the field by providing a scalable objective tool that addresses a critical diagnostic gap. While prospective validation is paramount to confirm clinical utility, the REDMOD framework represents a significant advance towards shifting the paradigm for sporadic PDA from a late-stage symptomatic diagnosis to proactive pre-clinical interception, offering tangible hope for improving outcomes in this challenging disease.
    Next-generation AI for visually occult pancreatic cancer detection in a low-prevalence setting with longitudinal stability and multi-institutional generalisability.
    Mukherjee, S., et al.
    Gut. 2026 DOI: 10.1136/gutjnl-2025-337266.
  • In Gut, Mukherjee et al introduce a radiomics-base dearly detection model (REDMOD) designed precisely to enable earlier diagnosis of PDAC.1 The presented model is a fully automated AI framework that interrogated standard contrast-enhanced CT scans for subvisual radiomic signatures of prediagnostic PDAC. The REDMOD algorithm was trained on routine CT scans without visually evident pancreatic tumours from a multi-institutional cohort and was subsequently validated on an independent test set.
    Seeing the unseen: AI radiomics unmasking occult pancreatic cancer?
    Patrick Michl , Laura Roth
    GUT 2026 (in press)
  • REDMOD was able to retrospectively detect PDAC at a median of 475 days before clinical diagnosis, with robust performance across different institutions, CT vendors and slice thicknesses and stable predictions on longitudinal scans. The AI algorithm clearly outperformed blinded abdominal radiologists with an area under the curve (AUC) of 0.82 (REDMOD) versus 0.69 for radiologists, reaching superior sensitivity (73% vs 38.9%), while maintaining a broadly comparable—though slightly lower—specificity (REDMOD 81.1% vs radiologists 87–97%).
    Seeing the unseen: AI radiomics unmasking occult pancreatic cancer?
    Patrick Michl , Laura Roth
    GUT 2026 (in press)
  • Despite the significant progress achieved by this study, it has also to be kept in mind that an 81% specificity at a 1:6 case:- control ratio implies that still a quite substantial number of individuals would be incorrectly flagged as PDAC-positive. Mukherjee et al explicitly acknowledge this and position their current operating point as a development benchmark.1 As the field moves forward, we will need to define what level of false positives—and thus what number of asymptomatic (high-risk) individuals exposed to unnecessary downstream investigations and/or invasive procedures—we consider acceptable in exchange for earlier detection and improved survival.
    Seeing the unseen: AI radiomics unmasking occult pancreatic cancer?
    Patrick Michl , Laura Roth
    GUT 2026 (in press)
  • The presented algorithm clearly provides a highly significant step towards radiomics-guided early PDAC detection. Based on these data, prospective validation of REDMOD in high-risk cohorts such as new-onset diabetics or familial pancreatic cancer is warranted, ideally in a multiparametric approach alongside blood-based biomarkers and accompanied by cost-effectiveness modelling using different thresholds. In addition, extension of comparable AI approaches to MRI and EUS would be highly desirable, avoiding the need for repetitive exposure of (asymptomatic high-risk) individuals to radiation.
    Seeing the unseen: AI radiomics unmasking occult pancreatic cancer?
    Patrick Michl , Laura Roth
    GUT 2026 (in press)
  • Mukherjee et al have established a strong methodological foundation to move beyond the low yields of current surveillance programmes. With further optimisation and prospective validation, this approach has the potential to substantially increase the proportion of PDACs diagnosed at a curable stage, before a lesion becomes clinically apparent and radiologically visible—by ‘seeing the unseen’. Ultimately, this will translate into meaningful improvements in real-world survival.
    Seeing the unseen: AI radiomics unmasking occult pancreatic cancer?
    Patrick Michl , Laura Roth
    GUT 2026 (in press)
Esophagus

  • CT Features of Achalasia
    Esophageal Dilation: Moderate to severe, fluid- and debris-filled dilation of the esophageal lumen. In long-standing, severe cases, this can manifest as a massive, redundant, tortuous "sigmoid esophagus."
    Distal Narrowing: Smooth, symmetrical, concentric narrowing of the lumen at the gastroesophageal junction, creating a "beaked" transition zone.
    Wall Thickness: The esophageal wall typically demonstrates normal thickness or mild, uniform hypertrophy.
    Air-Fluid Levels: Chronic retention of ingested material and saliva frequently creates prominent air-fluid or air-debris levels within the dilated thoracic esophagus.
  • Complications of Achalasia
    Aspiration Pneumonia: Chronic stasis and regurgitation can lead to recurrent aspiration, visible on CT as patchy airspace opacities, tree-in-bud nodules, or bronchiectasis, predominantly in the dependent lung zones.
    Esophageal Squamous Cell Carcinoma: Long-standing stasis causes chronic irritation. CT can screen for focal, eccentric wall thickening or intraluminal masses that suggest malignant transformation.
    Esophageal Perforation: Though rare, it can occur spontaneously (megaesophagus rupture) or secondary to therapeutic interventions like pneumatic dilation, visible as pneumomediastinum, mediastinal fluid collections, or hydropneumothorax.
Musculoskeletal

  • Renal Osteodystrophy
    Renal osteodystrophy refers specifically to the alteration of bone morphology and remodeling found in patients with Chronic Kidney Disease (CKD). It is the skeletal component of the broader systemic disorder known as Chronic Kidney Disease-Mineral and Bone Disorder (CKD-MBD).
    Because the kidneys cannot properly excrete phosphate or activate vitamin D, a cascade of secondary hyperparathyroidism occurs, leading to classic, dramatic radiologic findings.
  • Renal Osteodystrophy Bone Changes
    On imaging, renal osteodystrophy typically manifests as a complex mixture of two processes occurring simultaneously: bone resorption (driven by high parathyroid hormone [PTH] levels) and osteosclerosis (an abnormal increase in bone density).
  • Renal Osteodystrophy Bone Changes
    Osteosclerosis & The "Rugger-Jersey" Spine:
    While PTH resorbs bone, it paradoxically stimulates osteoblastic activity in the axial skeleton.
    This creates dense, sclerotic bands at the upper and lower endplates of the vertebral bodies, while the middle remains osteopenic. On a lateral spine X-ray, this striped pattern closely resembles the horizontal stripes 
  • Fibrous dysplasia (FD) of the skull
    a benign, chronic bone disorder where normal bone and marrow are replaced by fibrous connective tissue and poorly formed, weak trabecular bone. When it affects the skull and facial bones, it is referred to as craniofacial fibrous dysplasia.
    It can occur as an isolated finding in a single bone (monostotic), involve multiple bones (polyostotic), or present as part of McCune-Albright syndrome (associated with café-au-lait skin spots and endocrine abnormalities).
  • Key Clinical Features
    Asymmetry and Swelling: The most common presentation is a gradual, painless enlargement of the skull or facial bones, often leading to facial asymmetry.
    Neurological Findings: Depending on the location, bony overgrowth can narrow the cranial foramina. This can potentially lead to:
    Vision loss or changes (due to optic canal involvement)
    Hearing loss (due to internal auditory canal involvement)
    Headaches
    Progression: Growth typically slows down or stabilizes after skeletal maturity (post-puberty), though it can occasionally continue to expand in adulthood.
OB GYN

  • Gestational Trophoblastic Disease
    Computed Tomography (CT) plays a pivotal role in staging, risk stratification, and detecting metastatic disease once a malignant progression—such as Gestational Trophoblastic Neoplasia (GTN) (e.g., invasive mole, choriocarcinoma)—is suspected or confirmed by rising beta-hCG levels.
  • CT Features of primary uterine manifestation of GTD/GTN
    Uterine Enlargement: The uterus is often disproportionately large for expected gestational age.
    Heterogeneous Myometrial Mass: A central, hypoattenuating, focal mass or diffuse cavity enlargement with mixed attenuation due to areas of hemorrhage, necrosis, and retained trophoblastic tissue.
    Marked Hypervascularity: Striking, intense, irregular enhancement of the myometrium or tumor tissue during the early arterial and venous phases, often accompanied by dilated, tortuous pelvic vessels (uterine and ovarian arteries/veins).
  • CT Features of primary uterine manifestation of GTD/GTN
    Myometrial Invasion: Loss of the normal discrete myometrial borders, indicating invasive mole or choriocarcinoma.
    Theca Lutein Cysts: Large, multilocular, bilateral ovarian cysts are frequently visualized, resulting from extreme ovarian overstimulation by markedly elevated beta-hCG levels. These characteristically regress following treatment.
  • CT Evaluation for Metastatic Disease
    Pulmonary Metastasis (Most Common Site)
    Hepatic Metastasis
    Gastrointestinal and Splenic Metastasis
  • Differential Diagnosis on CT
    Hypervascular Uterine Leiomyoma / Leiomyosarcoma: May show similar enhancement but lacks the typical association with prominent, tortuous gonadal vessels and bilateral theca lutein cysts.
    Retained Products of Conception (RPOC): Can display overlapping enhancement patterns within the endometrium, though deep, destructive myometrial invasion and distant hypervascular metastases point definitively toward GTN. Clinical correlation with serial beta-hCG tracking is definitive.
  • Hypervascular Uterine Tumors
    Uterine Hemangioma: Extremely rare, benign vascular tumors consisting of proliferated, dilated blood vessels within the myometrium or endometrium. They can be congenital or acquired (sometimes associated with pregnancy) and may cause abnormal uterine bleeding (AUB).
    Hemangioendothelioma / Angiosarcoma: Exceedingly rare, aggressive malignant vascular tumors of the uterus with a poor prognosis.
  • Hypervascular Uterine Tumors
    Hypervascular Uterine Fibroids (Leiomyomas): While fibroids are smooth muscle tumors, some variants are highly vascularized. On color Doppler ultrasound, they typically display circumferential vascularity (a "ring of fire" pattern) with discrete feeding vessels.
    Uterine Sarcomas (e.g., Leiomyosarcoma, Endometrial Stromal Sarcoma): Malignant smooth muscle or mesenchymal tumors that characteristically display rich, chaotic, and irregular central and peripheral vascularization with low-resistance, high-velocity flow on Doppler.
    Gestational Trophoblastic Neoplasia (GTN): Such as invasive mole or choriocarcinoma. These are highly vascular, aggressive tumors related to pregnancy that deeply invade the myometrium and exhibit intense hypervascularity and arteriovenous shunting.
Pancreas

  • REDMOD is an automated, mechanistically grounded, longitudinally stable, externally validated AI that surpasses radiologists for PDA detection at its visually occult pre-diagnostic stage. These attributes position it for prospective validation in high-risk cohorts, a necessary step towards shifting the paradigm from late-stage symptomatic diagnosis to proactive pre-clinical interception.
    Next-generation AI for visually occult pancreatic cancer detection in a low-prevalence setting with longitudinal stability and multi-institutional generalisability.
    Mukherjee, S., et al.
    Gut. 2026 DOI: 10.1136/gutjnl-2025-337266.
  • On an independent test set (n=493), REDMOD identified occult PDA (AUC 0.82; 73.0% sensitivity) at a median 475-day lead time. This represented nearly twofold higher sensitivity than radiologists (38.9%; p<0.001), which grew to nearly threefold (68.0% vs 23.0%) at >24 months lead time. REDMOD showed strong longitudinal stability (90–92% concordance) and generalisable specificity across multi-institutional (81.3%; n=539) and public (87.5%; n=80) datasets. Mechanistic analyses confirmed predictive power derived principally from multi-scale wavelet-filtered textural features (90% of selected signature), which outperformed unfiltered features (AUC 0.82 vs 0.74; p=0.007) in capturing subvisual architectural disruptions
    Next-generation AI for visually occult pancreatic cancer detection in a low-prevalence setting with longitudinal stability and multi-institutional generalisability.
    Mukherjee, S., et al.
    Gut. 2026 DOI: 10.1136/gutjnl-2025-337266.
  • * The performance of REDMOD surpasses that of radiologists, demonstrating nearly double the sensitivity (73.0% vs 38.9%; p<0.001) for detecting visually occult PDA, with an advantage that increases to nearly threefold for cases detected more than 24 months prior to diagnosis.
    * For the first time, this work establishes the longitudinal stability of a pre-clinical radiomic signal (90–92% test–retest concordance) and confirms the robust specificity of the model across independent multi-institutional (81.3%) and public (87.5%) datasets.
    Next-generation AI for visually occult pancreatic cancer detection in a low-prevalence setting with longitudinal stability and multi-institutional generalisability.
    Mukherjee, S., et al.
    Gut. 2026 DOI: 10.1136/gutjnl-2025-337266.
  • * REDMOD offers an automated and longitudinally stable anchor for multicentre prospective studies in risk-enriched early detection programmes, with pre-specified thresholds and an observe–retest strategy for intermediate results.
    * REDMOD could operate as a triage and longitudinal monitoring tool for CT requests triggered by NICE guidelines, aligning with NHS pathways while awaiting prospective clinical evidence and cost-effectiveness.
    * The release of a fully automated and externally benchmarked pipeline offers a reproducible platform for method comparisons, bias audits and cost-effectiveness modelling to guide policy.
    Next-generation AI for visually occult pancreatic cancer detection in a low-prevalence setting with longitudinal stability and multi-institutional generalisability.
    Mukherjee, S., et al.
    Gut. 2026 DOI: 10.1136/gutjnl-2025-337266.
  • Our group achieved a breakthrough by demonstrating that machine learning (ML) models, using radiomic features derived from pre-diagnostic CT scans—incidental CTs acquired months to years prior to clinical diagnosis—exhibit high accuracy in identifying such subclinical alterations indicative of carcinogenesis. These models substantially outperformed radiologists in detecting visually occult PDA (stage 0 disease) with a considerable lead time before clinical diagnosis.
    Next-generation AI for visually occult pancreatic cancer detection in a low-prevalence setting with longitudinal stability and multi-institutional generalisability.
    Mukherjee, S., et al.
    Gut. 2026 DOI: 10.1136/gutjnl-2025-337266.
  • This study validates REDMOD as a fully automated AI framework capable of identifying the imaging signatures of stage 0 PDA in normal pancreas, achieving this with substantial lead times and performance superior to expert radiologists. The demonstrated ability of the framework to consistently detect these occult signals on a large clinically-oriented dataset, combined with its high longitudinal stability and validated specificity, establishes a robust foundation for AI-augmented early detection. This work overcomes key barriers in the field by providing a scalable objective tool that addresses a critical diagnostic gap. While prospective validation is paramount to confirm clinical utility, the REDMOD framework represents a significant advance towards shifting the paradigm for sporadic PDA from a late-stage symptomatic diagnosis to proactive pre-clinical interception, offering tangible hope for improving outcomes in this challenging disease.
    Next-generation AI for visually occult pancreatic cancer detection in a low-prevalence setting with longitudinal stability and multi-institutional generalisability.
    Mukherjee, S., et al.
    Gut. 2026 DOI: 10.1136/gutjnl-2025-337266.
  • In Gut, Mukherjee et al introduce a radiomics-base dearly detection model (REDMOD) designed precisely to enable earlier diagnosis of PDAC.1 The presented model is a fully automated AI framework that interrogated standard contrast-enhanced CT scans for subvisual radiomic signatures of prediagnostic PDAC. The REDMOD algorithm was trained on routine CT scans without visually evident pancreatic tumours from a multi-institutional cohort and was subsequently validated on an independent test set.
    Seeing the unseen: AI radiomics unmasking occult pancreatic cancer?
    Patrick Michl , Laura Roth
    GUT 2026 (in press)
  • REDMOD was able to retrospectively detect PDAC at a median of 475 days before clinical diagnosis, with robust performance across different institutions, CT vendors and slice thicknesses and stable predictions on longitudinal scans. The AI algorithm clearly outperformed blinded abdominal radiologists with an area under the curve (AUC) of 0.82 (REDMOD) versus 0.69 for radiologists, reaching superior sensitivity (73% vs 38.9%), while maintaining a broadly comparable—though slightly lower—specificity (REDMOD 81.1% vs radiologists 87–97%).
    Seeing the unseen: AI radiomics unmasking occult pancreatic cancer?
    Patrick Michl , Laura Roth
    GUT 2026 (in press)
  • Despite the significant progress achieved by this study, it has also to be kept in mind that an 81% specificity at a 1:6 case:- control ratio implies that still a quite substantial number of individuals would be incorrectly flagged as PDAC-positive. Mukherjee et al explicitly acknowledge this and position their current operating point as a development benchmark.1 As the field moves forward, we will need to define what level of false positives—and thus what number of asymptomatic (high-risk) individuals exposed to unnecessary downstream investigations and/or invasive procedures—we consider acceptable in exchange for earlier detection and improved survival.
    Seeing the unseen: AI radiomics unmasking occult pancreatic cancer?
    Patrick Michl , Laura Roth
    GUT 2026 (in press)
  • The presented algorithm clearly provides a highly significant step towards radiomics-guided early PDAC detection. Based on these data, prospective validation of REDMOD in high-risk cohorts such as new-onset diabetics or familial pancreatic cancer is warranted, ideally in a multiparametric approach alongside blood-based biomarkers and accompanied by cost-effectiveness modelling using different thresholds. In addition, extension of comparable AI approaches to MRI and EUS would be highly desirable, avoiding the need for repetitive exposure of (asymptomatic high-risk) individuals to radiation.
    Seeing the unseen: AI radiomics unmasking occult pancreatic cancer?
    Patrick Michl , Laura Roth
    GUT 2026 (in press)
  • Mukherjee et al have established a strong methodological foundation to move beyond the low yields of current surveillance programmes. With further optimisation and prospective validation, this approach has the potential to substantially increase the proportion of PDACs diagnosed at a curable stage, before a lesion becomes clinically apparent and radiologically visible—by ‘seeing the unseen’. Ultimately, this will translate into meaningful improvements in real-world survival.
    Seeing the unseen: AI radiomics unmasking occult pancreatic cancer?
    Patrick Michl , Laura Roth
    GUT 2026 (in press)
  • Background Solid pancreatic serous cystadenoma (SCA) may present as a hypervascular mass and mimic nonfunctioning pancreatic neuroendocrine tumour (NF-pNET), potentially leading to pancreatic resection for benign disease. We synthesized the available evidence on this diagnostic pitfall.
    Conclusions NF-pNET–mimicking SCA is an uncommon but clinically consequential benign mimic. In the available retrospective comparative cohorts of resected, pathology-proven lesions, lower unenhanced CT attenuation (and attenuation ratios) and higher DWI–ADC values consistently favored SCA over NF-pNET, whereas arterial hyperenhancement and positive somatostatin receptor imaging uptake were not specific. These findings may support cautious multiparametric assessment in selected hypervascular solid-appearing pancreatic lesions, but should not be applied as definitive diagnostic thresholds without external validation.
    Hypervascular solid-appearing serous cystadenoma as a mimic of nonfunctioning pancreatic neuroendocrine tumour: a systematic review.
    Annecchiarico M, Loiaco G, Argenio G, et al.
    Abdom Radiol (NY). 2026 Apr 25. Epub ahead of print.
  • Solid SCA is a recognised radiological pitfall; it may present as a hypervascular solid lesion, thus mimicking potentially malignant pancreatic lesions, particularly nonfunctioning pancreatic neuroendocrine tumour (NF-pNET) . The solid variant is rare, and diagnostic confusion typically arises when individual microcysts fall below the spatial resolution of conventional cross-sectional imaging (computed tomography [CT] or magnetic resonance imaging [MRI]), resulting in a pseudo-solid hypervascular appearance.
    Hypervascular solid-appearing serous cystadenoma as a mimic of nonfunctioning pancreatic neuroendocrine tumour: a systematic review.
    Annecchiarico M, Loiaco G, Argenio G, et al.
    Abdom Radiol (NY). 2026 Apr 25. Epub ahead of print.
  • This systematic review highlights that SCA may occasionally mimic NF-pNET on preoperative assessment and that enhancement behavior alone is insufficient to resolve the differential diagnosis. When uncertainty persists in a resectable hypervascular solid pancreatic lesion, management may escalate to pancreatic resection for what ultimately proves to be a benign disease. This is particularly relevant because guideline frameworks support conservative management when a serous cystic neoplasm is confidently diagnosed, whereas surgical resection remains an accepted option for selected NF-pNET within guideline-based pathways [3, 8]. In our synthesis, most NF-pNET–mimicking SCA were reported as the solid variant (67/89, 75.3%), although microcystic lesions were also represented (12/89, 13.5%). This reflects the recognized spectrum of serous cystic neoplasms, in which small cystic spaces and enhancing septa may produce an apparently solid hypervascular mass.
    Hypervascular solid-appearing serous cystadenoma as a mimic of nonfunctioning pancreatic neuroendocrine tumour: a systematic review.
    Annecchiarico M, Loiaco G, Argenio G, et al.
    Abdom Radiol (NY). 2026 Apr 25. Epub ahead of print.
  • NF-pNET–mimicking SCA is an uncommon but clinically consequential benign mimic that may appear solid or pseudo-solid and may also show somatostatin receptor imaging avidity. In the available retrospective comparative cohorts of resected, pathology-proven lesions, lower unenhanced CT attenuation (and attenuation ratios) and higher DWI–ADC values consistently favored SCA over NF-pNET, whereas arterial hyperenhancement and positive 68Ga-DOTA–peptide uptake were not specific. Accordingly, in hypervascular solid-appearing pancreatic lesions with discordant features, these quantitative CT/MRI markers may serve as useful adjuncts within multidisciplinary assessment, while EUS-guided tissue acquisition may be considered when tissue confirmation would alter management. Because current evidence derives from small, selected retrospective series without external validation, these findings should be applied cautiously and not as definitive diagnostic thresholds.
    Hypervascular solid-appearing serous cystadenoma as a mimic of nonfunctioning pancreatic neuroendocrine tumour: a systematic review.
    Annecchiarico M, Loiaco G, Argenio G, et al.
    Abdom Radiol (NY). 2026 Apr 25. Epub ahead of print.
  • Somatic mutations in the Kirsten rat sarcoma viral oncogene homologue (KRAS) occur in more than 90% of pancreatic ductal adenocarcinomas. In this issue of the Journal, Wolpin et al.1 evaluated the side-effect profile of and response to a first-in-class small-molecule drug called daraxonrasib, which targets the active state of RAS (see Key Concepts) in patients with metastatic pancreatic ductal adenocarcinoma who have received at least one previous line of chemotherapy. In the context of second-line treatment, they reported an objective response according to the Response Evaluation Criteria in Solid Tumors (RECIST), version 1.1, in up to 35% of patients and a median response duration of 8.2 months.
    Advances in RAS Therapeutics for Pancreatic Cancer.
    Der CJ, Yeh JJ.
    N Engl J Med. 2026 May 7;394(18):1857-1861. 
  • Discovered in 1982, KRAS is a critical driver of tumor initiation, found in the earliest precursor of pancreatic ductal adenocarcinoma, pancreatic intraepithelial neoplasms.The KRAS protein drives myriad signaling pathways that promote cell growth and survival and alters cellular metabolism and the tumor microenvironment. Preclinical studies in mouse models of pancreatic ductal adenocarcinoma established the critical role of KRAS in tumor initiation and maintenance, supporting KRAS as a key dependency in pancreatic ductal adenocarcinoma.
    Advances in RAS Therapeutics for Pancreatic Cancer.
    Der CJ, Yeh JJ.
    N Engl J Med. 2026 May 7;394(18):1857-1861. 
  • In pancreatic ductal adenocarcinoma, KRAS missense mutations predominantly affect glycine 12 (G12): 42% result in G12D, 31% in G12V, and 15% in G12R. Mutations in other codons such as glycine 13 (G13) and glutamine 61 (Q61) are infrequent, and G12C mutations are found in fewer than 2% of pancreatic ductal adenocarcinomas. These single amino- acid substitutions impair the intrinsic and GAP-stimulated KRAS GTPase activity, favoring the formation of active RAS-GTP.
    Advances in RAS Therapeutics for Pancreatic Cancer.
    Der CJ, Yeh JJ.
    N Engl J Med. 2026 May 7;394(18):1857-1861. 
  • A second, independent approach for the development of RAS inhibitors was based on rapamycin, a natural product isolated from the bacterium Streptomyces hygroscopicus. Long used as an immunosuppressive agent for transplantation medicine,rapamycin binds to the FKBP12 protein, forming a binary complex that then binds to the mammalian target of rapamycin (mTOR) protein kinase and forms a tri-complex that inhibits mTOR complex 1. This mechanism, involving a molecular glue (rapamycin), provided the conceptual foundation for the development of RAS(ON) inhibitors.
    Advances in RAS Therapeutics for Pancreatic Cancer.
    Der CJ, Yeh JJ.
    N Engl J Med. 2026 May 7;394(18):1857-1861. 
  • Reactivation of RAS pathway signaling through genomic alterations has been reported in preclinical models and tumor-derived cell-free DNA of patients treated with daraxonrasib ; combination strategies are being evaluated (Table 1). Identifying predictive biomarkers of response and rationally designing combination strategies to overcome resistance mechanisms to daraxonrasib will be essential to improving on the unprecedented responses to this single-agentsmall-molecule inhibitor in pancreatic ductal adenocarcinoma.
    Advances in RAS Therapeutics for Pancreatic Cancer.
    Der CJ, Yeh JJ.
    N Engl J Med. 2026 May 7;394(18):1857-1861. 
  • BACKGROUND Current therapies for patients with pancreatic ductal adenocarcinoma (PDAC) provide modest benefit. Activating RAS mutations occur in more than 90% of PDAC tumors. Daraxonrasib (RMC-6236) is an oral RAS(ON) multiselective inhibitor that targets guanosine triphosphate–bound mutant and wild-type RAS.
    CONCLUSIONS Daraxonrasib was associated with treatment-related adverse events of grade 3 or higher in one third of patients with previously treated RAS-mutated PDAC; antitumor activity was also reported.
    Daraxonrasib in Previously Treated Advanced RAS-Mutated Pancreatic Cancer
    Brian M. Wolpin et al.
    n engl j med 394;18
  • RESULTS Among the 168 patients with PDAC who received daraxonrasib at a dose of 300 mg or less, treatment-related adverse events of any grade were reported in 96%; such events of grade 3 or higher were reported in 30%. Treatment-related adverse events hat occurred in at least 10% of the patients included rash, diarrhea, nausea, stomatitis or mucositis, vomiting, and fatigue. In a subgroup of 26 patients with RAS G12 mutations who were treated with second-line daraxonrasib at a dose of 300 mg, an objective response to therapy was reported in 35% (95% confidence interval[CI], 17 to 56). The median duration of response was 8.2 months (95% CI, 3.8 to not evaluable), with median values of 8.5 months for progression-free survival and 13.1 months for overall survival. Among the 38 patients with RAS G12, G13, orQ61 mutations, 29% (95% CI, 15 to 46) had an objective response. The median duration of response was 8.2 months (95% CI, 3.8 to 8.8), with median values of8.1 months for progression-free survival and 15.6 months for overall survival.
    Daraxonrasib in Previously Treated Advanced RAS-Mutated Pancreatic Cancer
    Brian M. Wolpin et al.
    n engl j med 394;18
  • Because this is a phase 1–2 study, the results should be interpreted with caution. Small subgroup sample sizes restrict the precision of estimates of antitumor activity and their interpretation on the basis of allele or co-mutation status. The single-group design precludes definitive comparisons with standard treatment regimens. Nonetheless, the consistent evidence of clinical activity,  rapid and durable responses, and a profile of mainly low-grade adverse events support the further evaluation of daraxonrasib in randomized trials involving patients with advanced PDAC. These results support the ongoing RASolute 302 study, a global randomized phase 3 trial of daraxonrasib as compared with chemotherapy as second- line treatment in patients with metastatic PDAC (ClinicalTrials.gov number, NCT06625320). Among patients with previously treated RASmutated PDAC, daraxonrasibshowed antitumor activity and was associated with treatment-related adverse events of grade 3 or higher in one third of patients.
    Daraxonrasib in Previously Treated Advanced RAS-Mutated Pancreatic Cancer
    Brian M. Wolpin et al.
    n engl j med 394;18
  • Artificial intelligence and computational modelling are increasingly being applied in PDAC research. Multiomic AI platforms have shown potential to predict PDAC survival outcomes , deep-learning approaches can link histopathological features with molecular and proteomic states , and computational drug-screening models have been used to nominate candidate synergistic drug combinations in pancreatic cancer . Although these approaches remain largely investigational, they may help prioritise future molecularly selected trials and accelerate precision oncology in PDAC .
    KRAS and Beyond: Emerging Targeted and Molecularly Stratified Strategies in Pancreatic Ductal Adenocarcinoma
    Alicia Y. Lefas , Hazel Lote * and Ian Chau
    Precis. Oncol. 2026, 1, 9
  • Pancreatic ductal adenocarcinoma (PDAC) is a highly lethal malignancy, with rising incidence and a 5-year survival rate of 13%. Late presentation, early metastasis, and intrinsic resistance constrain the efficacy of cytotoxic chemotherapy, which remains the backbone of PDAC treatment, with only modest survival gains and resistance nearly universal. Although KRAS mutations dominate tumour biology (~90% of cases), PDAC is a heterogeneous disease with distinct molecular subtypes that confer differential therapeutic vulnerabilities. Advances in comprehensive molecular profiling have catalysed a paradigm shift toward precision oncology in PDAC. In KRAS-mutant PDAC, mutation-specific inhibitors have established proof-of-concept, particularly in KRAS G12C disease, while nextgeneration approaches including KRAS G12D inhibitors, RAS-“ON” inhibitors, proteolysistargeting chimeras (PROTACs), and KRAS-targeted vaccine strategies are expanding the therapeutic landscape.
    KRAS and Beyond: Emerging Targeted and Molecularly Stratified Strategies in Pancreatic Ductal Adenocarcinoma
    Alicia Y. Lefas , Hazel Lote * and Ian Chau
    Precis. Oncol. 2026, 1, 9
  • Combination strategies targeting upstream and downstream effectors of the RAS–MAPK pathway are also being explored to enhance the depth and durability of response. In parallel, KRAS-wild-type PDAC has emerged as a molecularly distinct subgroup enriched for rare but actionable alternative oncogenic fusion drivers including NRG1, NTRK, RET, ALK, and FGFR. Additional molecularly directed strategies targeting HER2 alterations, BRAF mutations, EGFR-dependent signalling, and tumour selectively exposed surface antigens such as CLDN18.2 are under investigation across PDAC irrespective of KRAS mutation status. Synthetic lethal approaches, including targeting the PRMT5/CDKN2A/MTAP axis, represent a further emerging therapeutic strategy. Germline homologous recombination repair defects, particularly involving BRCA1/2 and PALB2, further define clinically important subsets with sensitivity to platinum chemotherapy and PARP inhibition.
    KRAS and Beyond: Emerging Targeted and Molecularly Stratified Strategies in Pancreatic Ductal Adenocarcinoma
    Alicia Y. Lefas , Hazel Lote * and Ian Chau
    Precis. Oncol. 2026, 1, 9
  • Pancreas Schwannoma
    Pancreatic schwannoma is an extremely rare neurogenic tumor arising from the autonomic nerve fibers (vagus or sympathetic) within the pancreas. On CT, its appearance varies significantly depending on the internal histological composition—specifically the ratio of Antoni A (hypercellular) to Antoni B (hypocellular/myxoid) areas.
  • Pancreas Schwannoma: CT Findings
    Morphology: Usually a well-defined, encapsulated, round or oval mass.
    Location: Most frequently found in the head of the pancreas (~40%), followed by the body (~20%).
    Attenuation:
    Solid Components: Typically hypodense to the adjacent pancreatic parenchyma on non-contrast CT.
    Cystic Changes: Larger tumors (often >3 cm) frequently undergo cystic degeneration, hemorrhage, or necrosis, appearing as complex cystic or multiloculated masses.
  • Pancreas Schwannoma: Enhancement Patterns
    Antoni A Areas: Usually show inhomogeneous or "progressive" enhancement (becoming more prominent in the delayed phase).
    Antoni B Areas: Often show little to no enhancement due to their hypocellular, myxoid nature and tendency for cystic change.
    Cinematic Rendering: Advanced 3D post-processing like cinematic rendering can sometimes better delineate the sharp margins and internal nodules compared to standard axial imaging.
  • Pancreas Schwannoma: Differential Dx
    Because pancreatic schwannomas lack specific "pathognomonic" imaging signs, they are often mistaken for more common lesions:
    Solid Lesions: Pancreatic neuroendocrine tumors (though these are typically more hypervascular in the early arterial phase) or adenocarcinoma (which usually shows ductal dilation and vascular invasion, both rare in schwannoma).
    Cystic Lesions: Mucinous cystic neoplasms, serous cystadenomas, or solid pseudopapillary neoplasms (SPN).
Stomach

  • Gastric cancer is the fifth most common cancer worldwide, and is more common in East Asia, Eastern Europe, and South America than in North America. In the US, approximately 30 300 patients are diagnosed with gastric cancer annually.  Gastric cancer is more common among males than females in the US and in older patients, although the diagnosis is increasing among patients younger than 50 years.
    Gastric Cancer
    Walter K
    JAMA Published online May 21, 2026
  • The most common risk factor worldwide is infection with the bacterium Helicobacter pylori, which is associated with nearly90% of gastric cancer cases involving the central and lower part of the stomach. Other risk factors include tobacco smoking, heavy alcohol use (4 drinks daily), high intake of salty foods, obesity, acid reflux, history of gastric ulcer, family history of gastric cancer, and certain inherited genetic variants. In the US, more than 90% of patients diagnosed with gastric cancer have symptoms such as weight loss, abdominal pain, nausea, loss of appetite, and difficulty swallowing. Iron deficiency anemia affects about 40% of patients with gastric cancer.
    Gastric Cancer
    Walter K
    JAMA Published online May 21, 2026
  • The 5-year survival rate for early-stage gastric cancer is about 75% with treatment. Patients with locally advanced gastric cancer treated with chemotherapy have a median survival of approximately 4 years. For unresectable or metastatic gastric cancer, median survival is approximately 1 to 2 years. Less than 10% of patients with metastatic gastric cancer survive longer than 5 years.
    Gastric Cancer
    Walter K
    JAMA Published online May 21, 2026
  • Treating Helicobacter pylori infection decreases the risk of gastric cancer. Other prevention strategies include avoiding tobacco and heavy alcohol use, decreasing salt intake, maintaining a healthy weight,and exercising regularly. In the US, guidelines recommend consideration of screening endoscopy for high-risk populations (such as people born in moderate- to high-incidence regions, including East Asia, Eastern Europe,and the Andean region of South America),andf or thosewho have first-degree family members with gastric cancer.
    Gastric Cancer
    Walter K
    JAMA Published online May 21, 2026

Privacy Policy

Copyright © 2026 The Johns Hopkins University, The Johns Hopkins Hospital, and The Johns Hopkins Health System Corporation. All rights reserved.