Imaging Pearls ❯ Deep Learning ❯ Deep Learning and the GU Tract
|
-- OR -- |
|
- Purpose: To develop and test a deep learning model to automatically depict adrenal nodules on abdominal CT images and to simulate triaging performance in combination with human interpretation.
Materials and Methods: This retrospective study (January 2000–December 2020) used an internal dataset enriched with adrenal nodules for model training and testing and an external dataset reflecting real-world practice for further simulated testing in combination with human interpretation. The deep learning model had a two-stage architecture, a sequential detection and segmentation model, trained separately for the right and left adrenal glands. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC) for nodule detection and intersection over union for nodule segmentation.
Conclusion: The deep learning model demonstrated high performance and has the potential to improve detection of incidental adrenal nodules.
Two-Stage Deep Learning Model for Adrenal Nodule Detection on CT Images: A Retrospective Study
Chang Ho Ahn • Taewoo Kim • Kyungmin Jo et al.
Radiology 2025; 314(3):e231650 - “Deep learning algorithms have shown promising results in detecting radiologic abnormalities across various imaging modalities. However, deep learning algorithms for the segmentation of intraabdominal organs have demonstrated the lowest accuracy for adrenal glands because of their greater complexity and anatomic variation. Initial attempts at characterizing the adrenal gland were limited in scope. A recent study showed success in using deep learning for adrenal gland segmentation and classification but did not validate it in a real-world clinical setting.”
Two-Stage Deep Learning Model for Adrenal Nodule Detection on CT Images: A Retrospective Study
Chang Ho Ahn • Taewoo Kim • Kyungmin Jo et al.
Radiology 2025; 314(3):e231650 - ■ In this retrospective study of 995 patients, the deep learning model depicted adrenal nodules with areas under the receiver operating characteristic curve of 0.99 and 0.93 for right and left adrenal glands, respectively (internal test set 1, n = 153); model sensitivity was superior to that of historical reports (99% vs 21%–45%; P < .001).
■ When combined with radiologists’ interpretation in additional test sets, the model improved nodule detection; triaging performance (the percentage that could be used to confidently classify by deep learning) ranged from 77% (938 of 1214) to 98% (11 776 of 12 080).
Two-Stage Deep Learning Model for Adrenal Nodule Detection on CT Images: A Retrospective Study
Chang Ho Ahn • Taewoo Kim • Kyungmin Jo et al.
Radiology 2025; 314(3):e231650 - “This study had several limitations. First, because of the retrospective design, the original radiology reports were produced by several radiologists with varying levels of experience, and the CT scans or protocols were also heterogeneous. The higher accuracy of the deep learning algorithm compared with the radiology report does not imply superiority over human interpretation. Second, only contrast-enhanced CT scans were included, whereas noncontrast CT is often preferred for assessing CT attenuation of adrenal nodules. Finally, the clinical usefulness of the automatic detection of adrenal nodules was estimated without consideration of the downside of increased detection rates.”
Two-Stage Deep Learning Model for Adrenal Nodule Detection on CT Images: A Retrospective Study
Chang Ho Ahn • Taewoo Kim • Kyungmin Jo et al.
Radiology 2025; 314(3):e231650 - “In summary, most adrenal nodules are detected incidentally and can be overlooked during imaging interpretation. Our deep learning model showed promising results in a real-world setting. With low and high thresholds, about 90% of patients could be classified with high confidence as having adrenal nodules using the deep learning model. Further studies should focus on developing user-friendly interfaces and integration into clinical workflows.”
Two-Stage Deep Learning Model for Adrenal Nodule Detection on CT Images: A Retrospective Study
Chang Ho Ahn • Taewoo Kim • Kyungmin Jo et al.
Radiology 2025; 314(3):e231650 - “In conclusion, the potential of AI to enhance the detection of adrenal nodules is becoming more transparent, with promising results in automated detection, segmentation, and even classification. However, the clinical use of these systems remains a subject of ongoing research and debate. The study by Ahn and Kim et al offers an exciting step forward, showing that AI can detect nodules with high sensitivity and assist radiologists in improving diagnostic accuracy. Yet, as with any emerging technology, integrating AI into clinical practice will require further validation, particularly in prospective studies and real-world scenarios. Ultimately, the future of AI in adrenal nodule detection may not only lie in better diagnosis but also in more precise, personalized characterization and prediction strategies for patients with adrenal abnormalities.”
Unveiling the Future: A Deep Learning Model for Accurate Detection of Adrenal Nodules
Ashkan A. Malayeri • Baris Turkbey
Radiology 2025; 314(3):e250387 - “Finally, while AI-powered detection algorithms demonstrate ever-increasing accuracy, they must be implemented in diagnostic workflows considering the entire imaging study rather than one single phase. For example, because adrenal CT with washout calculation is the primary technique for differentiating benign versus malignant adrenal nodules, a future vision of AI in this context may entail not only detecting and segmenting nodules, but also analyzing them across multiple phases, extracting the radiomic data, and providing an objective predictability score on the malignant versus benign nature of the nodules.”
Unveiling the Future: A Deep Learning Model for Accurate Detection of Adrenal Nodules
Ashkan A. Malayeri • Baris Turkbey
Radiology 2025; 314(3):e250387
- As artificial intelligence continues to shape the medical imaging landscape, its integration into the field of urologic oncology has led to impressive results. For prostate cancer diagnostics, machine learning has shown promise in refining clinically-significant lesion detection, with some success in deciphering ambiguous lesions on multiparametric MRI. For kidney cancer, radiomics has emerged as a valuable tool for better distinguishing between benign and malignant renal masses and predicting tumor behavior from CT or MRI scans. Meanwhile, in the arena of bladder cancer, there is a burgeoning emphasis on prediction of muscle invasive cancer and forecasting disease trajectory. However, many studies showing promise in these areas face challenges due to limited sample sizes and the need for broader external validation.
AI-powered radiomics: revolutionizing detection of urologic malignancies.
Gelikman, David G.a; Rais-Bahrami, et al.
Current Opinion in Urology 34(1):p 1-7, January 2024. - Radiomics integrated with artificial intelligence offers a pioneering approach to urologic oncology, ushering in an era of enhanced diagnostic precision and reduced invasiveness, guiding patient-tailored treatment plans. Researchers must embrace broader, multicentered endeavors to harness the full potential of this field.
AI-powered radiomics: revolutionizing detection of urologic malignancies.
Gelikman, David G.a; Rais-Bahrami, et al.
Current Opinion in Urology 34(1):p 1-7, January 2024.
- “Renal cancer is responsible for over 100,000 yearly deaths and is principally discovered in computed tomography (CT) scans of the abdomen. CT screening would likely increase the rate of early renal cancer detection, and improve general survival rates, but it is expected to have a prohibitively high financial cost. Given recent advances in artificial intelligence (AI), it may be possible to reduce the cost of CT analysis and enable CT screening by automating the radiological tasks that constitute the early renal cancer detection pipeline. This review seeks to facilitate further interdisciplinary research in early renal cancer detection by summarizing our current knowledge across AI, radiology, and oncology and suggesting useful directions for future novel work.”
Artificial intelligence for early detection of renal cancer in computed tomography: A review
William C. McGough et al.
Cambridge Prisms: Precision Medicine,1, e4, 1–9 - “Deep learning-based classifiers can achieve high accuracy in CT images with very little manual intervention. Tanaka et al. (2020) sought to quantify small (≤4 cm) renal mass detection accuracy in CT using axial CT slices and a fine-tuned InceptionV3 CNN; they differentiated malignant and benign masses with a maximum AUC of 0.846 in CECT and 0.562 in NCCT. Pedersen et al. (2020) trained a similar 2D slice-classifying CNN, but used it to classify each slice within each known mass’ 3D volumes to enable a slice-based voting system to differentiate patient-level RC from oncocytoma, returning a perfect validation accuracy of 100%. Han et al. (2019) sought to differentiate between clear cell RCC (ccRCC) and non-ccRCC from known RCC masses, using radiologist-selected axial CT slices from NCCT and two CECT phases, and achieved sub-type classification AUCs between 0.88 and 0.94 in an internal testing dataset.”
Artificial intelligence for early detection of renal cancer in computed tomography: A review
William C. McGough et al.
Cambridge Prisms: Precision Medicine,1, e4, 1–9 - “Given the potential for RC early detection in LDCT, there is a need for more research quantifying RC segmentation performance in LDCT. Investigations into general NCCT segmentation have shown that using synthetic contrast enhancement as an auxiliary training task in MTL can improve segmentation accuracy. Therefore, an investigation in renal LDCT segmentation may be improved by introducing synthetic enhancement to CECT as an auxiliary learning task in MTL. Such an investigation would likely be complicated by Standley et al. (2020) findings – that MTL task relationships can be unique to each configuration of network architecture, hyperparameters, and dataset domain.”
Artificial intelligence for early detection of renal cancer in computed tomography: A review
William C. McGough et al.
Cambridge Prisms: Precision Medicine,1, e4, 1–9 - “This manuscript highlights and summarizes existing AI method in RC diagnosis and suggests how these can be repurposed to enable RC early detection. After summarizing existing segmentation, classification, and other AI methods in RC diagnosis, a review of analogous cancer detection and diagnosis methods across broader cancer literature and computer vision was conducted. Contrasting the RC-specific workflows to their equivalents across computer vision and other cancer domains allowed the generation of novel RC-specific research proposals that may enable AI-based RC early detection.”
Artificial intelligence for early detection of renal cancer in computed tomography: A review
William C. McGough et al.
Cambridge Prisms: Precision Medicine,1, e4, 1–9
- Objectives: To develop and externally validate multiphase CT-based deep learning (DL) models for differentiating adrenal metastases from benign lesions.
Materials and methods: This retrospective two-center study included 1146 adrenal lesions from 1059 patients who underwent multiphase CT scanning between January 2008 and March 2021. The study encompassed 564 surgically confirmed adenomas, along with 135 benign lesions and 447 metastases confirmed by observation. DL models based on multiphase CT images were developed, validated and tested. The diagnostic performance of the classification models, imaging phases and radiologists with or without DL were compared using accuracy (ACC) and receiver operating characteristic (ROC) curves. Integrated discrimination improvement (IDI) analysis and the DeLong test were used to compare the area under the curve (AUC) among models. Decision curve analysis (DCA) was used to assess the clinical usefulness of the predictive models
Differentiating adrenal metastases from benign lesions with multiphase CT imaging: Deep learning could play an active role in assisting radiologists
Changyi Ma et al.
European Journal of Radiology 169 (2023) 111169 - Results: The DL signature based on LASSO (DLSL) had a higher AUC than that of the other classification models (IDI > 0, P < 0.05). Furthermore, the precontrast phase (PCP)-based DLSL performed best in the independent external validation (AUC = 0.881, ACC = 82.9 %) and clinical test cohorts (AUC = 0.790, ACC = 70.4 %), outperforming DLSL based on the other single-phase or three-phase images (IDI > 0, P < 0.05). DCA demonstrated that PCP-based DLSL provided a higher net benefit (0.01–0.95).
Conclusion: The DL model based on PCP CT images performed acceptably in differentiating adrenal metastases from benign lesions, and it may assist radiologists in accurate tumor staging for patients with a history of extra-adrenal malignancy.
Differentiating adrenal metastases from benign lesions with multiphase CT imaging: Deep learning could play an active role in assisting radiologists
Changyi Ma et al.
European Journal of Radiology 169 (2023) 111169 - In this study, we developed a DL model based on CT images that can effectively differentiate adrenal metastases from benign lesions in patients with a history of extra-adrenal malignancies. We found that the DLSL model based on PCP exhibited a comparable diagnostic performance compared to each single- or three-phase CT image and could help radiologists perform accurate imaging staging for patients with a known extra-adrenal cancer history. To our knowledge, this is the first time that a DL model has been used to differentiate adrenal metastases from benign lesions in a large population.
Differentiating adrenal metastases from benign lesions with multiphase CT imaging: Deep learning could play an active role in assisting radiologists
Changyi Ma et al.
European Journal of Radiology 169 (2023) 111169 - In the comparison with radiologists, DLSL showed a diagnostic performance comparable to that of radiologists with moderate years of experience. It had good supplementary value for improving the diagnostic performance of each radiologist, especially those with less experience. Furthermore, we found that DLSL significantly improved the specificity of the radiologists. This may be because, in the first retrospective CT evaluation, the radiologists were more influenced by the patient’s cancer history in their clinical diagnostic experience[3], which may have led to an increase in subjectivity in diagnosing cases of metastases. In the second retrospective CT evaluation, the radiologists leveraged the DLSL predictions for ambiguous cases to make the final decision, thus improving the accuracy of diagnosis.
Differentiating adrenal metastases from benign lesions with multiphase CT imaging: Deep learning could play an active role in assisting radiologists
Changyi Ma et al.
European Journal of Radiology 169 (2023) 111169
- ”In summary, AI has emerged as a promising tool with the potential to impact how radiologists, and even nonradiologists, provide timely care to patients presenting with symptoms, especially those in lower-resource environments. These AI tools are accurate when classifying malignant and benign breast masses and, therefore, can streamline both workflow and the allocation of downstream care. With continued improvements in portable scanning equipment, further education of on-site nonradiologist providers, and continued advances of existing AI algorithms, it appears that AI is primed to revolutionize how we care for patients presenting with breast complaints. Integrating AI into these lower-resource environments is yet another step that will lessen the burden of disease in these vulnerable patients who currently face major obstacles to care. The hope is that such steps will ultimately translate into lower mortality and better overall outcomes.”
The Promise of AI in Advancing Global Radiology
Priscilla J. Slanetz
Radiology 2023; 00:e230895
- “Radiomics and artificial intelligence (AI) may increase the differentiation of benign from malignant kidney lesions, differentiation of angiomyolipoma (AML) from renal cell carcinoma (RCC), differentiation of oncocytoma from RCC, differentiation of different subtypes of RCC, to predict Fuhrman grade, to predict gene mutation through molecular biomarkers and to predict treatment response in metastatic RCC undergoing immunotherapy. Neural networks analyze imaging data. Statistical, geometrical, textural features derived are giving quantitative data of contour, internal heterogeneity and gray zone features of lesions.”
Artificial intelligence and radiomics in evaluation of kidney lesions: a comprehensive literature review
Matteo Ferro et al.
Ther Adv Urol 2023, Vol. 15: 1–26 - “AI evidence so far indicates a strong association with improved sensitivity, specificity, accuracy in detecting and differentiating between renal lesions, and its algorithms that can adjust scanner settings to improve image acquisition (especially the gray zone levels) and standardization of scanner protocols between institutions will improve preoperative differentiation between benign, low-risk cancers and clinically significant renal cancers. Radiomics holds the premises to enhance the diagnostic ability of imaging tools to characterize renal lesions, but integration in clinical practice will have to be preceded by standardized radiomics models and methodology, and future prospective external validation of obtained data and their comparison with existing traditional, well-validated tools, will have to be performed prior to further integration in current practice.”
Artificial intelligence and radiomics in evaluation of kidney lesions: a comprehensive literature review
Matteo Ferro et al.
Ther Adv Urol 2023, Vol. 15: 1–26
- “AI has currently applications in almost all fields of adrenal diseases. Although most studies are preliminary studies, they suggest that AI may improve AL classification, prognosis, and possibly management of patients affected with AL. However, there is a long way to go before AI is implemented into daily practice of radiologist, endocrinologist, and surgeons. In this regard, one current limitation for immediate applicability is the observed difference in recommended management among societies. Some studies suggest that AI algorithms should not be built using imaging data alone but should integrate biological data for better efficacy. Finally, large, prospective studies with external validations are needed to build effective, predictive models that will help improve patients’ care by selecting the most appropriate option among surveillance, open vs. laparoscopic surgery or “leave it alone” option.”
Artificial intelligence in adrenal imaging: A critical review of currentapplications
Maxime Barata, Martin Gaillard, Anne-S egol ene Cottereaub, Elliot K. Fishman, Guillaume Assi, Anne Jouinotb, Christine Hoeffelg, Philippe Soyera, Anthony Dohana,
Diagnostic and Interventional Imaging 104 (2023) 37−42 - “Three studies haved investigated the capabilities of AI using CT data. Kusunoki et al. evaluated a complex algorithm using deep convolutional neuronal network to identify AA among undetermined AL on CT. Using a sample of 112 ALs, they found 87% sensitivity (95% CI: 80−93%), 96% specificity (95% CI: 88−100%) and an AUC of 0.94 (95% CI: 0.86−0.99%) for the best model for the diagnosis of AA vs. other conditions . Considering large ALs (i.e., > 4 cm), a radiomics-based ML algorithm was evaluated to differentiate ACC from AA using CT data . The algorithm yield 82% accuracy (95% CI: 69−92%) for differentiating between the two entities in a study population of 29 patients with ACC and 25 with AA). Of interest, the algorithm performed significantly better that the radiologist alone (69% accuracy; P < 0.0001) . However, because of small cohorts and lack of external validation, these two studies should be considered as preliminary/feasibility studies only.”
Artificial intelligence in adrenal imaging: A critical review of currentapplications
Maxime Barata, Martin Gaillard, Anne-S egol ene Cottereaub, Elliot K. Fishman, Guillaume Assi, Anne Jouinotb, Christine Hoeffelg, Philippe Soyera, Anthony Dohana,
Diagnostic and Interventional Imaging 104 (2023) 37−42 - “Recently, deep neuronal networks (DNN) were developed to improve accuracy of adrenal gland segmentation. Luo et al. used a two-step method with a preprocessing step to reduce variabilities between examinations and the computational burden followed by a second step of small organs segmentation network using annotated input. The CT dataset was obtained from 348 patients, 60% for the training, 20% for the validation and 20% for the testing cohort . With this model, the authors obtained a DSC of 87.2%, which is the best reported one to date for adrenal gland segmentation .Despite a better DSC than those obtained with previous segmentation algorithms, DNNs also require further external validation to demonstrate their robustness.”
Artificial intelligence in adrenal imaging: A critical review of currentapplications
Maxime Barata, Martin Gaillard, Anne-S egol ene Cottereaub, Elliot K. Fishman, Guillaume Assi, Anne Jouinotb, Christine Hoeffelg, Philippe Soyera, Anthony Dohana,
Diagnostic and Interventional Imaging 104 (2023) 37−42 - “After training and fine-tuning, the test set, which should be ideally made of external and unseen data, is used to assess the generalizability of the AI model. So an important point before an AI model can be deployed in the real world, it that its performances be validated using a large validation test with a variety of diagnoses from different databases that also include rare conditions and probably anatomical variations. Large data for AI models is the key because they help increase the confidence in predictions and allows robust internal and external validations and testing. However, large datasets raise several issues such as reliability of original data but also inclusion of rare conditions. One option to increase the prevalence of rare conditions or obtain a distribution that mirrors those of the real world to better train the model is to enrich the dataset using synthetic images obtained with data augmentation techniques, but this requires further investigation.”
Applications of Artificial Intelligence in Urological Oncology Imaging: More Data Are Needed
Philippe Soyer, Anthony Dohan, and Maxime Barat
Canadian Association of Radiologists’ Journal 2023, Vol. 0(0) 1–2 - The artificial-intelligence (AI) community faces a reproducibility crisis, says Sayash Kapoor, a PhD candidate in computer science at Princeton University in New Jersey. As part of his work on the limits of computational prediction, Kapoor discovered that reproducibility failures and pitfalls had been reported in 329 studies across 17 fields, including medicine. He and a colleague organized a one-day online workshop last July to discuss the subject, which attracted about 600 participants from 30 countries.
THE REPRODUCIBILITY ISSUES THAT HAUNT HEALTH-CARE AI
Emily Sohn
Nature | Vol 613 | 12 January 2023 | C
- “In the elective field of adrenal imaging, artificial intelligence (AI) can be used for adrenal lesion detection, characterization, hypersecreting syndrome management and patient follow-up. Although a perfect AI tool that includes all required steps from detection to analysis does not exist yet, multiple AI algorithms have been developed and tested with encouraging results. However, AI in this setting is still at an early stage. In this regard, most published studies about AI in adrenal gland imaging report preliminary results that do not have yet daily applications in clinical practice. In this review, recent developments and current results of AI in the field of adrenal imaging are presented. Limitations and future perspectives of AI are discussed.”
Artificial intelligence in adrenal imaging: A critical review of current applications
Maxime Baratab, Martin Gaillardb, Anne-Segolene Cottereaub, Elliot K. Fishman, Guillaume Assieb, Anne Jouinotb, Christine Hoeffel, Philippe Soyer, Anthony Dohana
Diagnostic and Interventional Imaging 000 (2022) 1−6 - "Recently, deep neuronal networks (DNN) were developed to improve accuracy of adrenal gland segmentation. Luo et al. used a two-step method with a preprocessing step to reduce variabilities between examinations and the computational burden followed by a second step of small organs segmentation network using annotated input . The CT dataset was obtained from 348 patients, 60% for the training, 20% for the validation and 20% for the testing cohort . With this model, the authors obtained a DSC of 87.2%, which is the best reported one to date for adrenal gland segmentation Despite a better DSC than those obtained with previous segmentation algorithms, DNNs also require further external validation to demonstrate their robustness.”
Artificial intelligence in adrenal imaging: A critical review of current applications
Maxime Baratab, Martin Gaillardb, Anne-Segolene Cottereaub, Elliot K. Fishman, Guillaume Assieb, Anne Jouinotb, Christine Hoeffel, Philippe Soyer, Anthony Dohana
Diagnostic and Interventional Imaging 000 (2022) 1−6 - "Radiomics is regarded as a promising research field for the characterization and follow-up of non-typical AL. Using first order texture analysis features, Jhaveri et al. found that a 5% cut-off of negative voxels on histogram derived from unenhanced CT data had 92.3% sensitivity and 100% specificity for the characterization of lipid-poor AA . Another found that entropy (i.e., a first order texture analysis feature) yielded an AUC of 0.65 (95% CI: 0.52−0.77) to differentiate between adrenal metastasis from lung cancer (showing greater entropy) and AA Ho et al. developed a model using 21 second order features extracted from contrast-enhanced CT data that allowed differentiating lipid-poor AA from malignant AL better than standard morphological features (AUC of 0.8 vs. 0.6, respectively). Similarly, a model using four second-order features extracted from unenhanced CT data yielded 96.6% sensitivity, 81% specificity and 85.2% accuracy for the diagnosis of pheochromocytoma vs. lipid poor AA.”
Artificial intelligence in adrenal imaging: A critical review of current applications
Maxime Baratab, Martin Gaillardb, Anne-Segolene Cottereaub, Elliot K. Fishman, Guillaume Assieb, Anne Jouinotb, Christine Hoeffel, Philippe Soyer, Anthony Dohana
Diagnostic and Interventional Imaging 000 (2022) 1−6 - “AI has currently applications in almost all fields of adrenal diseases. Although most studies are preliminary studies, they suggestthat AI may improve AL classification, prognosis, and possibly management of patients affected with AL. However, there is a long way to go before AI is implemented into daily practice of radiologist, endocrinologist, and surgeons. In this regard, one current limitation for immediate applicability is the observed difference in recommended management among societies. Some studies suggest that AI algorithms should not be built using imaging data alone but should integrate biological data for better efficacy. Finally, large, prospective studies with external validations are needed to buildeffective, predictive models that will help improve patients’ care by selecting the most appropriate option among surveillance, open vs. laparoscopic surgery or “leave it alone” option.”
Artificial intelligence in adrenal imaging: A critical review of current applications
Maxime Baratab, Martin Gaillardb, Anne-Segolene Cottereaub, Elliot K. Fishman, Guillaume Assieb, Anne Jouinotb, Christine Hoeffel, Philippe Soyer, Anthony Dohana
Diagnostic and Interventional Imaging 000 (2022) 1−6 - “A machine learning algorithm was developed that can accurately segment and classify adrenal glands as normal or mass-containing on contrastvenhanced CT images, with performance similar to that of radiologists.”
Machine Learning for Adrenal Gland Segmentation and Classification of Normal and Adrenal Masses at CT
Cory Robinson-Weiss et al.
Radiology 2022; 000:1–8 - • In this retrospective study using CT images from 251 (development data set) and 991 patients (secondary test set), a machine learning algorithm segmented adrenal glands, with a performance similar to that of radiologists; the median model Dice score 0.87 versus 0.89 for normal adrenal glands and 0.85 versus 0.89 for adrenal masses.
• The algorithm differentiated adrenal masses from normal adrenal glands, with 83% sensitivity and 89% specificity for the development data set and 69% sensitivity and 91% specificity for the secondary test set.
Machine Learning for Adrenal Gland Segmentation and Classification of Normal and Adrenal Masses at CT
Cory Robinson-Weiss et al.
Radiology 2022; 000:1–8 - “ Our results are promising given that the adrenal gland is an inherently challenging organ to segment compared with larger organs (such as the liver and kidneys), as these glands are small and change in position due to respiration, and their shape, size, and location can vary by laterality and patient. In addition, adrenal glands have soft-tissue attenuation at CT that is similar to that of adjacent structures, including the liver, pancreas, kidneys, and vasculature. In patients with a paucity of intra-abdominal fat, the adrenal glands may be even more challenging to delineate from adjacent structures without the contrasting fat around the glands to separate them from other structures. Finally, there may be external mass effect on the glands, for example from an adjacent liver cyst or mass.”
Machine Learning for Adrenal Gland Segmentation and Classification of Normal and Adrenal Masses at CT
Cory Robinson-Weiss et al.
Radiology 2022; 000:1–8 - “In conclusion, we propose a two-stage machine learning pipeline to automatically segment the adrenal glands at contrast enhanced CT and then classify the glands as normal or mass containing. This tool may be used to assist radiologists in accurate and expedient image interpretation and potentially decrease interreader variability. Future work is needed to improve the classification stage of our model, as well as expand on the scope of the classification task by reviewing prior imaging and assessing for mass stability or growth.”
Machine Learning for Adrenal Gland Segmentation and Classification of Normal and Adrenal Masses at CT
Cory Robinson-Weiss et al.
Radiology 2022; 000:1–8 - “In conclusion, only one-fourth of patients received appropriate biochemical testing after incidental diagnosis of an adrenal nodule, and most nodules with indeterminate imaging characteristics did not have follow-up imaging. Advanced data analytic techniques on electronic imaging reports may aid in the clinical identification andimproved management of patients with adrenal incidentalomas.”
Automated extraction of incidental adrenal nodules from electronic health records
Max Schumm et al.
Surgery xxx (2022) 1e7 (in press) - Background: Many adrenal incidentalomas do not undergo appropriate biochemical testing and complete imaging characterization to assess for hormone hypersecretion and malignancy. With the growing availability of clinical narratives in the electronic medical record, automated surveillance using advanced data analytic techniques may represent a promising method to improve management.
Conclusion: Only one-fourth of patients received appropriate biochemical testing after incidental diagnosis of an adrenal nodule, and most nodules with indeterminate imaging characteristics did notundergo follow-up imaging. Advanced data analytic techniques on electronic imaging reports may aid in the clinical identification and improved management of patients with adrenal incidentalomas.
Automated extraction of incidental adrenal nodules from electronic health records
Max Schumm et al.
Surgery xxx (2022) 1e7 (in press)
- “This observer study investigates the effect of computerized artificial intelligence (AI)-based decision support system (CDSS-T) on physicians’ diagnostic accuracy in assessing bladder cancer treatment response. The performance of 17 observers was evaluated when assessing bladder cancer treatment response without and with CDSS-T using pre- and post-chemotherapy CTU scans in 123 patients having 157 pre- and post-treatment cancer pairs. The impact of cancer case difficulty, observers’ clinical experience, institution affiliation, specialty, and the assessment times on the observers’ diagnostic performance with and without using CDSS-T were analyzed. It was found that the average performance of the 17 observers was significantly improved (p = 0.002) when aided by the CDSS-T. The cancer case difficulty, institution affiliation, specialty, and the assessment times influenced the observers’ performance without CDSS-T. The AI-based decision support system has the potential to improve the diagnostic accuracy in assessing bladder cancer treatment response and result in more consistent performance among all physicians.”
Computerized Decision Support for Bladder Cancer Treatment Response Assessment in CT Urography: Effect on Diagnostic Accuracy in Multi-Institution Multi-Specialty Study
Di Sun et al.
Tomography 2022, 8, 644–656. - “We have developed a computerized artificial intelligence (AI)-based decision supportsystem for muscle-invasive bladder cancer treatment response assessment (CDSS-T) to assist physicians to evaluate the response to treatment of these cancers on pre- and posttreatment CT urography (CTU) scans . It is critical to gain understanding of variousfactors that may affect the impact of CDSS-T on physician performance in identifying bladder cancers with complete response after neoadjuvant chemotherapy through observer studies that can guide the design of future clinical trials. Patients with complete response may be considered for organ preservation therapy instead of cystectomy (the removal of the bladder).”
Computerized Decision Support for Bladder Cancer Treatment Response Assessment in CT Urography: Effect on Diagnostic Accuracy in Multi-Institution Multi-Specialty Study
Di Sun et al.
Tomography 2022, 8, 644–656. - "The performance comparisons of experienced and inexperienced physicians are shown in Table 5. We can see there was no observable difference between their performances. The level of statistical significance of the inexperienced radiologists was slightly higher (p = 0.007) after using CDSS-T compared to that of experienced radiologists (p = 0.06). The use of CDSS-T resulted in more consistent performance among all subgroups of physicians (all AUC = 0.77).”
Computerized Decision Support for Bladder Cancer Treatment Response Assessment in CT Urography: Effect on Diagnostic Accuracy in Multi-Institution Multi-Specialty Study
Di Sun et al.
Tomography 2022, 8, 644–656. - "In conclusion, our study demonstrated that the computerized decision support system (CDSS-T) has the potential to improve the diagnostic accuracy in assessing the complete response of muscle-invasive bladder cancer to neoadjuvant chemotherapy prior to radical cystectomy. The use of CDSS-T aid has resulted in improved and more consistent diagnostic performance among the physicians from multiple institutions and multiple specialties.”
Computerized Decision Support for Bladder Cancer Treatment Response Assessment in CT Urography: Effect on Diagnostic Accuracy in Multi-Institution Multi-Specialty Study
Di Sun et al.
Tomography 2022, 8, 644–656. 
Pancreatic Cancer Detection using Machine and Deep Learning Techniques
Gupta A et al.
2022 2nd International Conference on Innovative Practices in Technology and Management (ICIPTM)- “Chen et al. tested a machine learning technique for detecting people having cancer in their pancreas at an early phase using medical data that had been collected from digital health records. As shown in eq. 1, they utilized eXtreme Gradient Boosting (XGBoost) to create a prediction model to detect early-stage patients based on 18,220 EHR variables, including diagnoses, procedures, clinical note information, and medicines.”
Pancreatic Cancer Detection using Machine and Deep Learning Techniques
Gupta A et al.
2022 2nd International Conference on Innovative Practices in Technology and Management (ICIPTM)
- “Machine learning and radiomic features extraction can differentiate between benign and malignant indeterminate adrenal tumors and can be used to direct further workup with high sensitivity and specificity.”
Machine learning‐based texture analysis for differentiation of radiologically indeterminate small adrenal tumors on adrenal protocol CT scans
Ahmed W. Moawad et al.
Abdominal Radiology https://doi.org/10.1007/s00261-021-03136-2 - “Adrenal “incidentalomas” are adrenal lesions that are accidentally discovered during workup not related to the adrenal glands; they have an incidence as high as 5%. Small adrenal incidentalomas (< 4 cm) with high attenuation values on pre-contrast CT(> 10 HU) need further evaluation to calculate the absolute percentage of washout (APW). If the APW is < 60%, these lesions are considered non-adenomas and commonly classified as indeterminate adrenal lesions. Further workup for indeterminate lesions includes more complicated and expensive radiological studies or invasive procedures like biopsy or surgical resection.”
Machine learning‐based texture analysis for differentiation of radiologically indeterminate small adrenal tumors on adrenal protocol CT scans
Ahmed W. Moawad et al.
Abdominal Radiology https://doi.org/10.1007/s00261-021-03136-2 
Machine learning‐based texture analysis for differentiation of radiologically indeterminate small adrenal tumors on adrenal protocol CT scans
Ahmed W. Moawad et al.
Abdominal Radiology https://doi.org/10.1007/s00261-021-03136-2- ”We did not include pheochromocytomas in the current study, as they have variable imaging features with the adrenal protocol. Pheochromocytoma usually demon- strates attenuation > 10 HU owing to its low fat content, but some cases contain sufficient fat to drop the attenuation to < 10 HU. Also, there are no typical washout features of pheochromocytoma during the adrenal protocol and a wide range of washout criteria. In addition, iodinated contrast agent is not recommended for patients with pheochromocytoma.”
Machine learning‐based texture analysis for differentiation of radiologically indeterminate small adrenal tumors on adrenal protocol CT scans
Ahmed W. Moawad et al.
Abdominal Radiology https://doi.org/10.1007/s00261-021-03136-2 - “In the current study, texture analysis of contrast-enhanced CT can differentiate between benign and malignant adrenal lesions smaller than 4 cm with pre-contrast attenuation values > 10 HU. Using features extracted from contrast- enhanced studies, we have shown that machine learning- based texture analysis can be used as a non-invasive tool for differentiation prior to invasive procedures. Our study differs from other that we specifically examined use of radiomics in radiologically indeterminant adrenal lesions, which cause additional financial, and emotional burden for the healthcare system. A more uniform prospective study is needed to accurately determine the clinical suitability of our model and to incorporate it into a decision support system.”
Machine learning‐based texture analysis for differentiation of radiologically indeterminate small adrenal tumors on adrenal protocol CT scans
Ahmed W. Moawad et al.
Abdominal Radiology https://doi.org/10.1007/s00261-021-03136-2
- “For all these reasons, preoperative diagnosis of solid renal masses continues to challenge clinicians. In fact, differentiation between RCC and oncocytomas is usually made on the basis of histologic findings of the surgically removed tumor. To avoid unnecessary surgery in patients with benign lesions, preoperative diagnosis by imaging would be of great value.”
Usefulness of MDCT to Differentiate Between Renal Cell Carcinoma and Oncocytoma: Development of a Predictive Model
Blanca Paño et al.
AJR 2016; 206:764–774 - “Oncocytomas showed significantly higher enhancement than RCCs in the nephrographic and excretory phases, that is, N2 and N3 of oncocytomas were higher than N2 and N3 of RCCs (p = 0.02 and p = 0.03 for N2 and N3, respectively), and the same was true in lesions 4 cm or smaller (p = 0.03 and p = 0.004 for N2 and N3, respectively) (Table 2).”
Usefulness of MDCT to Differentiate Between Renal Cell Carcinoma and Oncocytoma: Development of a Predictive Model
Blanca Paño et al.
AJR 2016; 206:764–774 - “Our study suggests that the combination of homogeneous enhancement and the difference in attenuation for the excretory phase leads to a good level of correct classifications between RCC and oncocytoma. In addition, given that the tumor can be measured quickly and simply and the high (almost eightfold) increase in risk, a tumor larger than 4 cm should always trigger further study when imaging data is not available. Our study provides higher accuracy than that previously reported, because most studies to date have concluded that RCC and oncocytoma cannot be differentiated by analyzing each variable independently using MDCT. This higher accuracy could be attributable to the multiparametric analysis we performed.”
Usefulness of MDCT to Differentiate Between Renal Cell Carcinoma and Oncocytoma: Development of a Predictive Model
Blanca Paño et al.
AJR 2016; 206:764–774 - “We have not evaluated the presence of segmental enhancement inversion or the presence of a central, sharply marginated, stellate scar (commonly described as characteristic findings for oncocytoma) because these features have been found in both oncocytoma and RCC, particularly in the chromophobe subtype of RCC, and are therefore poor predictors of oncocytoma .Regarding the usefulness of the different phases, our study suggests that arterial phase imaging is not helpful to differentiate RCC from oncocytoma and that only unenhanced, nephrographic, and excretory phases are necessary.”
Usefulness of MDCT to Differentiate Between Renal Cell Carcinoma and Oncocytoma: Development of a Predictive Model
Blanca Paño et al.
AJR 2016; 206:764–774 - “In conclusion, this study indicates that the combination of parameters assessed using MDCT (size, homogeneity, and enhancement changes between the excretory and unenhanced phases) can help distinguish between RCC and oncocytoma. The phases that have proved useful for differentiation are the excretory and nephrographic phases; their differences in relation to the unenhanced phase also appear to provide helpful information. Our data should be confirmed and validated in a larger and independent cohort.”
Usefulness of MDCT to Differentiate Between Renal Cell Carcinoma and Oncocytoma: Development of a Predictive Model
Blanca Paño et al.
AJR 2016; 206:764–774 - OBJECTIVE. The objective of our study was to systematically review the literature about the application of artificial intelligence (AI) to renal mass characterization with a focus on the methodologic quality items.
MATERIALS AND METHODS. A systematic literature search was conducted using PubMed to identify original research studies about the application of AI to renal mass characterization. Besides baseline study characteristics, a total of 15 methodologic quality items were extracted and evaluated on the basis of the following four main categories: modeling, performance evaluation, clinical utility, and transparency items. The qualitative synthesis was presented using descriptive statistics with an accompanying narrative.
RESULTS. Thirty studies were included in this systematic review. Overall, the methodologic quality items were mostly favorable for modeling (63%) and performance evaluation (63%). Even so, the studies (57%) more frequently constructed their work on nonrobust features. Furthermore, only a few studies (10%) had a generalizability assessment with independent or external validation. The studies were mostly unsuccessful in terms of clinical utility evaluation (89%) and transparency (97%) items. For clinical utility, the interesting findings were lack of comparisons with both radiologists' evaluation (87%) and traditional models (70%) in most of the studies. For transparency, most studies (97%) did not share their data with the public.
CONCLUSION. To bring AI-based renal mass characterization from research to practice, future studies need to improve modeling and performance evaluation strategies and pay attention to clinical utility and transparency issues.
Artificial Intelligence in Renal Mass Characterization: A Systematic Review of Methodologic Items Related to Modeling, Performance Evaluation, Clinical Utility, and Transparency
Burak Kocak et al.
American Journal of Roentgenology 2020 215:5, 1113-1122 - OBJECTIVE. The objective of our study was to systematically review the literature about the application of artificial intelligence (AI) to renal mass characterization with a focus on the methodologic quality items.
MATERIALS AND METHODS. A systematic literature search was conducted using PubMed to identify original research studies about the application of AI to renal mass characterization. Besides baseline study characteristics, a total of 15 methodologic quality items were extracted and evaluated on the basis of the following four main categories: modeling, performance evaluation, clinical utility, and transparency items. The qualitative synthesis was presented using descriptive statistics with an accompanying narrative. strategies and pay attention to clinical utility and transparency issues.
Artificial Intelligence in Renal Mass Characterization: A Systematic Review of Methodologic Items Related to Modeling, Performance Evaluation, Clinical Utility, and Transparency
Burak Kocak et al.
American Journal of Roentgenology 2020 215:5, 1113-1122 - RESULTS. Thirty studies were included in this systematic review. Overall, the methodologic quality items were mostly favorable for modeling (63%) and performance evaluation (63%). Even so, the studies (57%) more frequently constructed their work on nonrobust features. Furthermore, only a few studies (10%) had a generalizability assessment with independent or external validation. The studies were mostly unsuccessful in terms of clinical utility evaluation (89%) and transparency (97%) items. For clinical utility, the interesting findings were lack of comparisons with both radiologists' evaluation (87%) and traditional models (70%) in most of the studies. For transparency, most studies (97%) did not share their data with the public.
CONCLUSION. To bring AI-based renal mass characterization from research to practice, future studies need to improve modeling and performance evaluation strategies and pay attention to clinical utility and transparency issues.
Artificial Intelligence in Renal Mass Characterization: A Systematic Review of Methodologic Items Related to Modeling, Performance Evaluation, Clinical Utility, and Transparency
Burak Kocak et al.
American Journal of Roentgenology 2020 215:5, 1113-1122 - “In this study, we systematically reviewed 30 studies about the application of AI to renal mass characterization. Our focus was on the methodologic quality items related to modeling, performance evaluation, clinical utility, and transparency. The quality items were favorable for modeling and performance evaluation categories for most studies. On the other hand, they were poor in terms of clinical utility evaluation and transparency for most studies. To move this field of research to clinical practice, future studies need to improve modeling and performance evaluation strategies by constructing their analysis on the robust features and finding ways to perform proper generalizability assessment with independent or external validation. In addition, forthcoming studies should pay attention to clinical utility and transparency issues.”
Artificial Intelligence in Renal Mass Characterization: A Systematic Review of Methodologic Items Related to Modeling, Performance Evaluation, Clinical Utility, and Transparency
Burak Kocak et al.
American Journal of Roentgenology 2020 215:5, 1113-1122
- OBJECTIVE. The objective of our study was to systematically review the literature about the application of artificial intelligence (AI) to renal mass characterization with a focus on the methodologic quality items.
MATERIALS AND METHODS. A systematic literature search was conducted using PubMed to identify original research studies about the application of AI to renal mass characterization. Besides baseline study characteristics, a total of 15 methodologic quality items were extracted and evaluated on the basis of the following four main categories: modeling, performance evaluation, clinical utility, and transparency items. The qualitative synthesis was presented using descriptive statistics with an accompanying narrative.
CONCLUSION. To bring AI-based renal mass characterization from research to practice, future studies need to improve modeling and performance evaluation strategies and pay attention to clinical utility and transparency issues.
Artificial Intelligence in Renal Mass Characterization: A Systematic Review of Methodologic Items Related to Modeling, Performance Evaluation, Clinical Utility, and Transparency
Kocak B et al.
AJR 2020; 215:1113–1122 - OBJECTIVE. The objective of our study was to systematically review the literature about the application of artificial intelligence (AI) to renal mass characterization with a focus on the methodologic quality items.
CONCLUSION. To bring AI-based renal mass characterization from research to practice, future studies need to improve modeling and performance evaluation strategies and pay attention to clinical utility and transparency issues.
Artificial Intelligence in Renal Mass Characterization: A Systematic Review of Methodologic Items Related to Modeling, Performance Evaluation, Clinical Utility, and Transparency
Kocak B et al.
AJR 2020; 215:1113–1122 - “As a broad concept, artificial intelligence (AI) covers a wide variety of machine learning (ML) methods or algorithms that create models without strict rule-based programming beforehand. These algorithms can improve and correct themselves through experience. The goal of AI tools is to predict certain outcomes using multiple variables. In the field of medical imaging, there has been extensive interest in AI tools.”
Artificial Intelligence in Renal Mass Characterization: A Systematic Review of Methodologic Items Related to Modeling, Performance Evaluation, Clinical Utility, and Transparency
Kocak B et al.
AJR 2020; 215:1113–1122 - "In this study, we systematically reviewed 30 studies about the application of AI to re- nal mass characterization. Our focus was on the methodologic quality items related to modeling, performance evaluation, clinical utility, and transparency. The quality items were favorable for modeling and perfor- mance evaluation categories for most stud- ies. On the other hand, they were poor in terms of clinical utility evaluation and transparency for most studies.”
Artificial Intelligence in Renal Mass Characterization: A Systematic Review of Methodologic Items Related to Modeling, Performance Evaluation, Clinical Utility, and Transparency
Kocak B et al.
AJR 2020; 215:1113–1122
- IMPORTANCE For prostate cancer, Gleason grading of the biopsy specimen plays a pivotal role in determining case management. However, Gleason grading is associated with substantial interobserver variability, resulting in a need for decision support tools to improve the reproducibility of Gleason grading in routine clinical practice.
OBJECTIVE To evaluate the ability of a deep learning system (DLS) to grade diagnostic prostate biopsy specimens.
CONCLUSIONS AND RELEVANCE In this study, the DLS showed higher proficiency than general pathologists at Gleason grading prostate needle core biopsy specimens and generalized to an independent institution. Future research is necessary to evaluate the potential utility of using the DLS as a decision support tool in clinical workflows and to improve the quality of prostate cancer grading for therapy decisions.
Development and Validation of a Deep Learning Algorithm for Gleason Grading of Prostate Cancer From Biopsy Specimens
Kunal Nagpal et al.
JAMA Oncol. doi:10.1001/jamaoncol.2020.2485 Published online July 23, 2020. - RESULTS For grading tumor-containing biopsy specimens in the validation set (n = 498), the rate of agreement with subspecialists was significantly higher for the DLS (71.7%; 95% CI, 67.9%-75.3%) than for general pathologists (58.0%; 95% CI, 54.5%-61.4%) (P < .001). In subanalyses of biopsy specimens from an external validation set (n = 322), the Gleason grading performance of the DLS remained similar. For distinguishing nontumor from tumor-containing biopsy specimens (n = 752), the rate of agreement with subspecialists was 94.3% (95% CI, 92.4%-95.9%) for the DLS and similar at 94.7% (95% CI, 92.8%-96.3%) for general pathologists (P = .58).
Development and Validation of a Deep Learning Algorithm for Gleason Grading of Prostate Cancer From Biopsy Specimens
Kunal Nagpal et al.
JAMA Oncol. doi:10.1001/jamaoncol.2020.2485 Published online July 23, 2020. - Key Points
Question: How does a deep learning system for assessing prostate biopsy specimens compare with interpretations determined by specialists in urologic pathology and by general pathologists?
Findings: In a validation dataset of 752 biopsy specimens obtained from 2 independent medical laboratories and a tertiary teaching hospital, this study found that rate of agreement with subspecialists was significantly higher for the deep learning system than it was for a cohort of general pathologists.
Meaning: The deep learning system warrants evaluation as an assistive tool for improving prostate cancer diagnosis and treatment decisions, especially where subspecialist expertise is unavailable.
Development and Validation of a Deep Learning Algorithm for Gleason Grading of Prostate Cancer From Biopsy Specimens
Kunal Nagpal et al.
JAMA Oncol. doi:10.1001/jamaoncol.2020.2485 Published online July 23, 2020. - “To conclude, we have presented a DLS for Gleason grading of prostate biopsy specimens that is highly concordant with sub- specialists and that maintained its performance on an external validation set. Future work will need to assess the diag- nostic and clinical effect of the use of a DLS for increasing the accuracy and consistency of Gleason grading to improve patient care.”
Development and Validation of a Deep Learning Algorithm for Gleason Grading of Prostate Cancer From Biopsy Specimens
Kunal Nagpal et al.
JAMA Oncol. doi:10.1001/jamaoncol.2020.2485 Published online July 23, 2020. 
Development and Validation of a Deep Learning Algorithm for Gleason Grading of Prostate Cancer From Biopsy Specimens
Kunal Nagpal et al.
JAMA Oncol. doi:10.1001/jamaoncol.2020.2485 Published online July 23, 2020.
- OBJECTIVE. The purpose of this study is to evaluate the potential value of machine learning (ML)–based high-dimensional quantitative CT texture analysis in predicting the mutation status of the gene encoding the protein polybromo-1 (PBRM1) in patients with clear cell renal cell carcinoma (RCC). CONCLUSION. ML-based high-dimensional quantitative CT texture analysis might be a feasible and potential method for predicting PBRM1 mutation status in patients with clear cell RCC.
Radiogenomics in Clear Cell Renal Cell Carcinoma: Machine Learning–Based High-Dimensional Quantitative CT Texture Analysis in Predicting PBRM1 Mutation Status
Burak Kocak et al.
AJR 2019; 212:W55–W63 - “Quantitative CT (QCT) texture analysis (TA) is an image processing method for measuring repetitive pixel or voxel gray-level patterns that may not be perceptible with the human eye. Several texture parameters can be produced by this method, which makes QCT TA high-dimensional. Although the field of high-dimensional QCT TA is still under development, the literature suggests that QCT TA can be used for characterizing lesions or tumors, predicting staging, nuclear grading, assessing the response to treatment, and predicting survival.”
Radiogenomics in Clear Cell Renal Cell Carcinoma: Machine Learning–Based High-Dimensional Quantitative CT Texture Analysis in Predicting PBRM1 Mutation Status
Burak Kocak et al.
AJR 2019; 212:W55–W63 - “Radiogenomics is a field of radiology in- vestigating the potential associations be- tween the imaging features of a disease and the underlying genetic patterns or molecular phenotype of that disease. The field has aimed to noninvasively obtain predictive data for diagnostic, prognostic, and, ultimately, optimal therapeutic assessment.”
Radiogenomics in Clear Cell Renal Cell Carcinoma: Machine Learning–Based High-Dimensional Quantitative CT Texture Analysis in Predicting PBRM1 Mutation Status
Burak Kocak et al.
AJR 2019; 212:W55–W63 - In conclusion, ML-based high-dimensional QCT TA is a feasible and potential method for predicting PBRM1 mutation status in patients with clear cell RCC. Nonetheless, more studies with more labeled data are absolutely required for further validation and improve- ment of the method for clinical use. We hope that the present study will provide the basis for new research.
Radiogenomics in Clear Cell Renal Cell Carcinoma: Machine Learning–Based High-Dimensional Quantitative CT Texture Analysis in Predicting PBRM1 Mutation Status
Burak Kocak et al.
AJR 2019; 212:W55–W63
- Purpose: To compare biparametric contrast-free radiomic machine learning (RML), mean apparent dffusion coeficient (ADC), and radiologist assessment for characterization of prostate lesions detected during prospective MRI interpretation.
Conclusion: Quantitative measurement of the mean apparent diffusion coeficient (ADC) improved differentiation of benign versus malignant prostate lesions, compared with clinical assessment. Radiomic machine learning had comparable but not better performance than mean ADC assessment.
Radiomic Machine Learning for Characterization of Prostate Lesions with MRI: Comparison to ADC Values
Bonekamp D et al.
Radiology 2018 (in press)
- “The authors demonstrated that the DL-CNN can overcome the strong boundary between two regions that have large difference in gray levels and provides a seamless mask to guide level set segmentation, which has been a problem for many gradient-based segmentation methods. Compared to our previous CLASS with LCR method, which required two user inputs to initialize the segmentation, DL-CNN with level sets achieved better segmentation performance while using a single user input. Compared to the Haar-feature-based likelihood map, the DL-CNN-based likelihood map could guide the level sets to achieve better segmentation. The results demonstrate the feasibility of our new approach of using DL-CNN in combination with level sets for segmentation of the bladder.”
Urinary bladder segmentation in CT urography using deep-learning convolutional neural network and level sets. Cha KH et al. Med Phys. 2016 Apr;43(4):1882