Chest: Multidetector Row CT of the Thorax Data Management ;2D and 3D Visualization of Thoracic MDCT Data
Leo P. Lawler, MD, FRCR1, Elliot K. Fishman MD, FACR1. 1All authors;
The Russell H. Morgan Department of Radiology and Radiological Science,
Johns Hopkins Medical Institutions,
601 N. Caroline Street,
Baltimore, Maryland, MD 21287
Introduction.
Multidetector row CT (MDCT) data sets improve on the already established prowess of single detector helical CT (SDCT) for thoracic imaging. Many anatomical features of the chest do not conform to a single two dimensional (2D) axial plane and full exploitation of the isotropic and near isotropic MDCT data requires two and three-dimensional (3D) post processing techniques to harness the added advantage of improved z-axis resolution and coverage . Data acquisition and processing as it relates to constructing the appropriate substrate for post-processing as well as the various two and three-dimensional techniques will be discussed. The clinical application of two and three-dimensional visualization to bronchovascular structures, lung parenchyma, chest wall and diaphragm will be addressed. Two and Three-Dimensional Post-Processing
Both optimal 2D and 3D studies require careful selection of the protocol for data acquisition.Most current two and three-dimensional studies are obtained from an 8 detector adaptive array system using 4 channel data acquisition though the principles of 2D and 3D visualization are the same for 16 detector arrays. For the majority of patients we employ the 1mm or 2.5 mm detectors producing 1.25 or 3mm slice widths respectively. The larger detectors can be used with faster table translation for poor breathholders or where greater scan coverage is required. 2D reconstructions can be made at any slice position or reconstruction interval. Unlike single detector systems the slice width is not inextricably linked to the beam collimation and data acquired may be assimilated into larger slice widths, which are of higher quality than similarly sized single detector slices. The smallest slice width available is equal to the smallest detector applied. Cardiac images with decreased motion may be obtained using the fast temporal resolution of MDCT together with prospective or retrosepective cardiac gating. 2D slices with a 512 matrix are fashioned from the helical data by a number of techniques of interpolation, which aim to approximate a true planar slice from the helical data set. Though there is some widening of the slice sensitivity profile, increased pitch has less deleterious effects on effective slice width compared with single detector systems . 180 or 360-degree linear interpolation deduces the slice data from adjacent helices 180 and 360 degrees apart. The 2D data from 180-degree linear interpolation has a higher temporal resolution and smaller effective slice width. Filter width interpolation is an MDCT technique that derives the slice data from a series of detector information not just the helices immediately adjacent the designated slice. For high resolution imaging such (e.g. intersititial lung disease or chest wall bone imaging) a high spatial frequency reconstruction kernel is used whereas a soft tissue kernel is best for most other protocols including routine chest, mediastinal and angiographic imaging. Multiplanar and curved multiplanar (MPR and CMPR) are 2D techniques that provide alternate viewing perspectives usually with conventional window settings . These images are a re-ordering of the voxels into one voxel thick tomographic sections excluding those voxels outside the imaging plane and they require minimal computer power. A curved line may be drawn to include an entire structure that does not lie in a single plane (CMPR).
3D visualization provides both a means to manage the large data sets of MDCT and to obtain novel perspectives on diseases of the chest with near-istotropic and isotropic data sets. The main 3D techniques currently available include shaded surface display (SSD), maximum intensity projection (MIP), minimum intensity projection (MinIP) and volume rendering (VR). Shaded surface display reformats the data around a threshold that defines the interface of tissues. The 3D surface is then simulated using polygon building block reconstructions and lighting models . SSD does not reveal internal detail and is mainly used for imaging the musculoskeletal components of the chest and can be used to show the mucosa surface of the airways. MIP casts a ray through the data and then only displays the data above a certain assigned value, reducing all the data in the ray to a single plane. It is most akin to a projectional technique and has been mainly applied to vascular imaging though it has been mentioned of use for nodule detection. MIP does not have any depth cues and the 3D relationships of structures are only appreciated by moving the image . MinIP is the opposite of MIP and by displaying only data below a designated threshold is best suited for showing areas of lower density such as cystic change, emphysema or air trapping. SSD, MIP and MinIP only display 10-20% of the data actually present and so do not require large computer power .
Volume rendering is a quite unique form of 3D visualization of high fidelity to the originally acquired data with preservation of depth cues and spatial relationships . In this process a ray is cast through the data and a weighted representation of all the Houndsfield units encountered is displayed depending on their representation within the tissues including voxels only partially filled with a density of interest. A histogram representation of the data may be generated (trapezoid) and then manipulated through various tools that display the data to best effect such as internal or external perspectives . There is depth information present and various levels of opacity can be assigned to tissues of interest. Volume rendering has been applied to all areas of chest CT.
Regardless of the 3D processing technique employed the editing steps must be quick and efficient. Laborious segmentation of data using slice-by-slice regions of interest is not acceptable in clinical practice. With volume rendering a region of interest may be drawn on a large volume of data at once. Clip plane editing is used to remove slabs of data in an infinite number of planes and can be performed in a real-time manner to remove overlying chest wall and can be customized to isolate individual bronchi. Sliding thin slab (STS) MIP and MinIP can be used to provide a scrolling projection of extreme attenuation values contained within multiple sections of lung tissue of 5-10mm thickness . Infinite manipulation of viewing perspective with frame rates over 20 frames per second are also a requisite for a viable 3D tool.
The choice of image display is hard or soft copy. Hard copy imaging is satisfactory for routine axial planar imaging but PACS systems with cine scrolling are preferred with smaller slice widths and intervals where larger image numbers are generated. All 3D studies are interpreted on soft copy workstations, which allow real-time clinical consultation though select hard copy images can be generated to go with the report or into the patient file.
High Resolution Chest Imaging.
When high-resolution chest CT is requested the clinical question usually relates to interstitial lung disease. Our interest centers on the secondary pulmonary lobule, the functional unit of the lung made up of terminal bronchovascular structures and alveoli gathered within connective tissue septations that carry lymphatic drainage. 2D MDCT can visualize these units with a detail that approximates gross pathologic inspection. Unlike single detector imaging MDCT using 1mm detectors can generate both the low noise larger slice widths necessary for general chest evaluation as well as the high-resolution 1.25mm slice widths required for interstitial study. It is not necessary to perform a second acquisition at pre-selected levels and dedicated reconstruction of areas of interest is possible before the raw data is deleted. This has practical implication with regard to many of the interstitial disease processes, which have static and dynamic patterns of distribution that may favor various disparate parts of the lung. Inspiratory and expiratory scans are performed to assess for air trapping and mosaic pattern whereas alternate prone and supine imaging can differentiate true disease from basatelectasis. Some advocate printing or displaying high-resolution images with a smaller field of view dedicated to one lung with side-by-side display of alternate phase or posture images to aid direct comparison. 2D reconstructions have high sensitivity for the detection of many of the signs of interstitial and airspace disease including septal lines, fibrosis, bronchiectasis and alveolar filling. By their nature many interstitial lung diseases are multifocal with asymmetry both superior to inferior and right to left. These complex distributions of disease can be reduced to single succinct images with MPR and volume rendering which can better depict the relative severity of certain areas of disease and may provide better reproducibility for longitudinal follow-up of response to therapy or progression(Figs.1-4). MinIP reconstructions have been shown sensitive for the detection of areas of low attenuation such as cystic change or air trapping (e.g. emphysema)(Fig.5) and it has been suggested MIP and STS-MIP images better depict centrilobular nodules . For focal severe bronchiectatic or emphysematous change limited lung resection can be planned and followed up using volume rendered data to establish the anatomical distribution of disease and normal lung and tailored images can be correlated with selected nuclear medicine ventilation/perfusion images. CT density and volume measurements do correlate with pulmonary function tests and it is hoped that through volume and perfusion measurements of the lung in various phases of respiration MDCT may ultimately provide a comprehensive evaluation of anatomic and functional derangements in a single examination .
Airway Imaging
Both CT and endoscopic bronchoscopy contributed to the demise of conventional bronchography. Whereas 2D CT can provide much of the information on lumen, wall and extramural structures of the tracheobronchial tree, bronchoscopy can better discern mucosal detail, sample tissues and intervene when necessary. Airway studies require the use of the narrowest collimation available for the coverage required with overlapping reconstruction. Though most fields of view extend from above the thoracic inlet to diaphragm any clinical questions regarding the glottic or subglottic area require extension to the hypopharynx. In an effort to limit dose exposure dynamic inspiratory and expiratory imaging is reserved for those patients with a clear clinical suspicion of bronchomalacia where a management benefit can be demonstrated. 2D interpretation of caliber change, irregularity and the anatomic distribution of abnormality are best achieved on lung windows whereas mediastinal windows better depict mural thickening or calcification or the extra-mural relationship of pathology. CT of the airways can best serve the pulmonologist and bronchoscopist when interpreted together with 3D reconstructions, which improve accuracy and confidence in the readings . These images provide sophisticated maps of tracheobronchial abnormalities and their exact relation to recognized anatomic landmarks that can be seen on endoscopy. In select cases 3D CT can obviate the need for bronchoscopy, in particular for follow-up of caliber change. Though MPR can be helpful the preferred technique is either SSD or VR using a wide range of trapezoids, projections and clip planes tailored to individual oblique branching airways to increase the conspicuity of disease(Figs.6,7). Solid or luminal view trapezoids with increased opacity are best to depict airway filling defects whereas these and bronchography like images can be used for mapping caliber change such as stenoses or bronchiectasis. Surface rendering and volume rendering has been used to produce endoscopic simulations of the airway and technologies that aid finding the centerline for fly-through evaluation have been explored. Such virtual endoscopic or perspective volume rendering imagesare not widely applied as they seldom give added information in a patient who has already had bronchoscopy and they cannot reproduce the hue and texture accurately. However virtual CT bronchsocpy incorportated into 3D information of extraluminal information can provide unique additional information such as safe routes for tracheobronchial biopsy . There is an increasing use of expandable endoprostheses for benign and malignant disease and trapezoids to depict the stent and the airway can be designed.
Congenital abnormalities of the airway can be difficult to discern on 2D images alone and endoscopic evalution can be limited by small airway caliber. However accessory and atretic bronchi or bronchi with abnormal angulation (e.g, bridging bronchus) are conspicuous on 3D studies and together with 3D CT angiography bronchovascular evaluation of rings and slings can be achieved. Airway filling defects are usually detected on close inspection of 2D studies with lung windows. Volume rendered studies better define such abnormalities as filling defects within otherwise normal airway images(Figs.8,9). Secondary invasion of the airway can be evaluated using 2D axial and MPR mediastinal windows or VR reconstructions that display both airway and extrinsic tissues to best effect. Defects in the wall due to tracheoesophageal fistula can be mapped for possible resection and reconstruction (Fig.10). Coronal or sagittal MPR and 3D studies are of particular value to demonstrate thin in-plane webs of the upper airway. Tracheobronchial stenoses are optimally seen with 3D reconstructions (Figs.11,12). Though caliber change perpendicular trachea can be appreciated on serial axial 2D images more subtle tapering and bronchiectasis is best demonstrated with 3D coronal or sagittal reconstructions using bronchography or solid airway trapezoids. Dedictated clip planes and projections are superior to measure the exact length of oblique running airways and are useful to provide consistent images for follow-up inflammatory stenoses (e.g. Wegeners). With the increasing use of self-expandable endoprostheses to restore or maintain airway patency for benign and malignant conditions 2D and 3D CT is of value (Fig.13). The 2D visualization can provide at thorough assessment of the etiology and complications of stenoses and assess the parenchyma for likelihood of benefit from attempts to relieve the obstruction. The 3D reconstructions are used to determine the ability of the bronchoscopist to traverse the narrowing and to customize the stent graft. Non-invasive follow-up can be provided by CT using internal and external volume rendering and bronchoscopic exams can be reserved for stent salvage or replacement. Airway dilatation (bronchiectasis) can be mapped in relation to remaining normal lung parenchyma to assess response to therapy and aid possible surgical resection (Fig.14)
Nodules and Masses
Lung nodules and masses can be well evaluated using 2D interpretation. 3mm slice widths from 2.5mm detectors with a soft tissue reconstruction kernel will usually provide all the information on shape, size, borders, multiplicity and attenuation profile. Some have suggested that the 2D contrast dynamics of lung nodules can aid the differentiation of benign and malignant disease. MIP has been shown to improve depiction and diagnosis of small nodules compared with axial 2D and MPR image interpretation . Superior sulcus tumors (Pancoast tumors) are best seen with coronal reformations of the chest apex.(Fig.15). Likewise diaphgramatic surface lesions that can be hard to differentiate from liver lesions can be discriminated on sagittal plane interpretation(Fig.16). For both benign and malignant small nodules surgical consultation for the resection approach is facilitated by a multimensional map which can facilitate segment sparing surgery(Fig.17). MPR images are used to map PET studies to CT either by computer fusion of data from different machines or with synchronous PET and CT scanning on a single machine. The images are usually visualized with planar axial, coronal and sagittal reformations though there is great potential for 3D volume studies to be used in the future.
. For TNM staging of lung cancer off-coronal volume rendering that shows the central extent of a medial lung mass and accurately names the lymph node chain involved can better serve the surgeon and oncologist in deciding on possible sleeve resection or alternate therapy(Fig.18). All lung mass locations can be described indirectly with reference to landmarks on the patient or on the chest x-ray to aid definition of the field for 3D radiation therapy that aims to treat the tumor with minimal ill-effects to the normal tissues. Most CT guided biopsies use limited 2D with static or CT fluoroscopic acquisition. Individual bronchi can be followed on 2D CT and 3D visualization to provide maps for bronchoscopic, medistinoscopy or limited surgical approaches for lymph node and lung biopsy.
Lung nodules and masses are rarely spherical leading to inter-scan and inter-observer measurement inaccuracy. It is thought that volume measurements which are independent of shape may be more accurate and can sensitively demonstrate temporal change that may predict doubling time and biological behaviour .
Chest wall and diaphragm
CT is used to evaluate chest wall and diaphgram pathology and in general 3D visualization is superior for the bone, muscle, pleura and diaphragm. SSD and volume rendering are primarily used for bony chest wall evaluation. Primary tumors (e.g. Askin or osteosarcoma) and secondary invasion of the chest wall can be better seen in relation to the normal chest on 3D reformations. Similarly congenital abnormalities such as pectus excavatum can be better measured for surgical planning with inferior and sagittal perspectives(Fig.19). Sternal trauma can be very difficult to see on axial images and sagittal MPR or volume rendering is more sensitive(Fig.20). Similarly the step-off of sternoclavicular dislocation or clavicle fracture is shown on coronal VR images rather than 2D studies. The extent of pleural disease (e.g. mesothelioma) and any change over time is best achieved with serial coronal comparisons using MPR or volume rendering. Diaphgramatic rupture can be difficult to differentiate from eventration with 2D axial visualization but discontinuity of the muscle and any mesentery or bowel herniation can be clearly seen with coronal or sagittal MPR. Vascular Imaging
Chest CT for vascular imaging includes studies of the systemic and pulmonary venous and arterial systems. The standard evaluation of all these vessels includes CT angiography (CTA) using contrast-enhanced studies with cine and multidimensional interpretation. MPR and CMPR can provide limited additional information to the axial 2D visualization. MIP and SSD suffer many of the shortcomings of digital subtraction antiography whereas volume rendering is preferred to discriminate the endoluminal contrast from vessel wall and bony thorax and may be more accurate. Endoscopic modes can show the relation of pathology to branch vessel origins such as subclavian relationship to thoracic aneurysms but have not been widely applied. The systemic veins of the thorax form a complex inter-connection of vessels between the chest wall and mediastinal structures, which have highly variable branching patterns. CT angiography can obtain better contrast opacification from a single bolus than conventional venographic studies. Contrast injection will preferentially opacify ipsilateral veins and contralateral opacification may require delayed imaging or alternate vein injection. Most commonly patients are referred for evaluation of complete or partial lumen compromise. 3D techniques are useful to measure the length of thrombosis or occlusion, to depict the relation of an impinging mass and to map collateral formation, which may aid recanalization procedures . Coronal and right anterior oblique views are best for the superior vena cava and the azygos and accessory hemiazygos systems are better seen with sagittal perspective. Since the veins tend to be close to the thoracic vertebrae volume rendering will show them better than SSD or MIP techniques. The pulmonary veins have become increasingly important of late with the use of ablative techniques to treat arrhythmogeinc foci in their myocardial sleeves. This requires an accurate map of the number and branch pattern of the veins to guide catheter based therapy and follow-up for stenoses. Though large anomalies can be seen on 2D axial images these vessels are best seen with volume rendering using posterior perspectives.
Pulmonary thrombo-embolic disease is the primary reason for pulmonary artery CT. Most emboli are well seen with 2D as low attenuation filling defects with possibly some asymmetric vessel enlargement . The ability of MDCT to capture the bolus is such that on may need to alter window width and level settings compared with SDCT so as not to obscure a small clot. Cine 2D scrolling can increase sensitivity to detection and MPR or 3D can on occasion help discriminate false positives due to obliquely coursing vessels. If the pulmonary embolism study is done together with an examination for DVT axial images from the iliac bifurcation to the deep femoral vessels are obtained. 5mm slice widths are adequate to survey the area and the attenuation difference of clot within the veins is best appreciated on 2D review. Chronic thrombo-embolic disease represents and build-up of thrombus along the walls of the pulmonary arteries gradually narrowing their lumen and distorting their walls. This condition is often treated surgically with an attempt made to remove the entire thrombus, as a cast and 3D mapping of the extent of disease can be useful in pre-operative planning.
CT examinations of the thoracic systemic arterial system center largely on aortic aneurysm and dissection including intramural hematoma, penetratic atherosclerotic ulcer and acute traumatic aortic injury (Figs.21-24). CT has a key role in the prompt evaluation and diagnosis of these conditions. For the most part in the emergency room setting 2D with cine scrolling of soft copy images can accurately depict an aneurysm or pseudoaneurysm, show dissection flaps with entry or re-entry tears and vascular complications such as organ ischemia. The subtle kink of traumatic injury may require multiplanar, MIP or volume rendered images. For therapeutic decisions both aneurysms and dissections benefit from 3D reformations and volume rendering is preferred to display all the density values of intimal calcification, contrast, thrombus and slow flow . MIP and SSD will only contain a portion of the data and may not depict an intimal flap or mural thrombus. 3D visualizationcan better depict aortic tortuosity and provide true orthogonal measurements of true and false lumina and may provide better localization of entry and re-entry tears . 180 degree linear interpolated 2D and 3D reconstructions may be helpful to assess possible false positive dissections such as pulsation artifact, which will appear as a regular irregularity on sagittal MPR or 3D studies. One can better appreciate the optimal location for clamp placement and graft or stent design and the burden of mural thrombus is clearly shown. The size of the neck of a pseudoaneurysm and its thrombus cap can be accurately measured and its often-complex shape can be appreciated. Increasingly both aneurysms and dissection CT studies must have coverage that includes the femoral vessels for an endovascular approach and 3D techniques provide a sense of their tortuosity and any subtle caliber change. Routinely antero-posterior and left anterior oblique images of the aorta are produced and oblique views of the femoral vessels just as in conventional angiography. Then additional perspectives tailored to the individual patient are generated.
Arteriovenous malfomations can be easily diagnosed on 2D CT. However they can have complex tortuous blood supply and drainage, which cannot be fully discerned on 2D imaging alone. Failure to fully map the abnormal blood vessels can lead to recanalization and recurrence in those patients treated by intravascular embolic therapy . 3D visualization has been shown to better assess their angioarchitecture in a way that positively affects outcome with decreased contrast requirements compared with conventional angiography (Fig.25,26).
There is much current interest in the role of MDCT for cardiac imaging. Axial 2D interpretation is used for coronary artery calcium scoring and estimating pericardial effusions. Multidimensional teschniques are used for cardiac masses to better localize their site with the chambers. It is unclear which will be the most efficacious way to evaluate coronary artery and perfusion abnormalities but algorithms are likely to include multidimensional visualization including cine studies and techniques to segment the coronary arteries to measure caliber change.
Conclusion.
As data acquisition continues to improve with MDCT there is a greater need to manage the large data sets and exploit their potential through the use of post-processing tools for two and three-dimensional visualization. Though two-dimensional axial interpretation will remain fundamental in practice for some time the concept of both volume acquisition and interpretation as a means to tailor chest CT studies to individual patients and pathology plays an increasing role.