How I Do It: Maximizing Efficiency in CTA Interpretation
They’re coming, and in many hospitals, they have already arrived: multidetector CT (MDCT), CT angiography (CTA), and advanced 3D imaging. This wonderful new modality offers the promise of evaluating disease processes from atherosclerosis to cancer to trauma more quickly, safely, and accurately than older techniques can. CTA is already replacing diagnostic catheter angiography in many institutions. The emergency evaluation of chest pain is changing rapidly, with MDCT used to evaluate the coronary arteries, pulmonary arteries, and aorta (often in a single breath hold), while at the same time allowing adjacent soft tissues to be seen for assessment for lung disease, chest-wall abnormalities, and upper-abdominal disease. Complex fractures are being evaluated in 3D for presurgical planning, with the anticipation of shorter, smarter surgical procedures and improved outcomes. Complex organs like the liver and pancreas can now be evaluated in detail prior to surgery, allowing surgeons to define optimum dissection planes for resection of disease. CT perfusion is being used more and more in the acute setting for stroke, in order to guide patient management in the crucial early hours after the onset of symptoms. New applications for this technology are being developed every day. It is clear that advanced visualization has moved from being a luxury to being a necessity in the practice of medicine. This article will focus on ways to maximize radiologist efficiency in processing and interpreting CTA studies. In the world of 3D, CTA is far and away the highest-volume procedure. It is critical, therefore, to develop both the skill sets and the workflow processes to handle ever-expanding volumes of cases as efficiently as possible. For the time and effort involved, professional reimbursement for CTA is low, generally being about $15 to $30 more than for a conventional CT. If it takes four times as long to read a CTA of the carotids than to read a CT of the neck, you are losing money. How, then, can we, as radiologists, offer high-quality CTA to our patients and referring physicians without getting bogged down in workflow inefficiencies? There are four key steps to maximizing efficiency in CTA: knowledge, support, access, and routine.
Figure 1. A 3D volumetric image of the carotid arteries with bone shadowing illustrating the relationship of the bifuration to the mandible.Knowledge Radiologists interpreting CTA must become familiar with the reconstruction algorithms commonly used in advanced visualization. The four most common reconstruction processes are 3D volumes, multiplanar reconstruction (MPR), maximum intensity projection (MIP), and curved planar reformats (CPRs). 3D volumetric images are created by taking the original CT axial slices and finding common boundaries based on Hounsfield units. These boundaries are converted into shaded surfaces to create 3D structures. Different structures can be brought out by varying the Hounsfield unit thresholds, and artificial-intelligence algorithms can be applied in order to segment out certain anatomic structures of similar density. As an example, programs have been written to recognize and remove the chest wall in coronary CTA automatically, or to remove the bones in a lower-extremity runoff CTA. 3D volumetric imaging provides an excellent overview of the spatial relationship between structures and offers a guide for defining other optimal reconstructions. The disadvantages of 3D volumetric images are that the resolution is low and heavy calcium can overwhelm the image and obscure underlying blood vessels. Typical uses of 3D volumes are bypass-graft orientation in coronary CTA, defining the relationship of the carotid bifurcation to the angle of the mandible in carotid CTA (Figure 1), and evaluating and comparing the degree of angulation across the neck of an abdominal aortic aneurysm. MPRs are the backbone of CTA interpretation for many anatomic regions. With the advent of MDCT, slices have become so thin that voxels are now isotropic. In other words, the voxel has become cubic in shape. This means that images can be reconstructed in any plane with no loss of resolution, and none of the annoying stair-step artifacts that we used to deal with in the early days of CT. For CTA, MPRs are extremely useful for evaluating stenoses, as well as for evaluating the superior and inferior ends of curved structures like blood vessels and organs. The main disadvantage of MPR is that because of the thinness of the slice, there is limited anatomic coverage per image. As a blood vessel moves in and out of the plane, only short segments are seen, and the reader must integrate the images mentally from slice to slice. MIPs are used as a way to overcome some of the limitations of MPRs. Other limitations creep in, however. With MIP, the 3D density data are projected into a 2D plane, incorporating the maximum densities of each individual slice and summing them. The thickness of the MIP can be varied by the user. MIP is ideal for highlighting uniformly dense structures that curve through a much less dense background. The main advantage of MIP is that large segments of dense curved structures, such as contrast-enhanced arteries, can be visualized in one image, making the images very intuitive to view. Unfortunately, when structures near a blood vessel are denser than contrast, the benefit of MIP becomes a liability, as these structures actually obscure the lumen. The two most common situations where this problem arises are the presence of heavily calcified arteries or of stents; Figure 2 illustrates this. Note how calcified plaque obscures the lumen of the artery as the thickness of the MIP is increased. MIP techniques have become a real workhorse for advanced visualization. This technique is used commonly for evaluating the carotid bifurcation, the pulmonary arteries (for pulmonary embolism), the coronary arteries, the circle of Willis, and the lower extremity. Virtually every CTA study will employ the MIP as part of its routine protocol. CPRs combine the best of MIP and MPR into one technique that provides the anatomic detail of thin sections with the coverage of MIP. CPRs provide a complex, curved reconstruction that follows the centerline of the structure of interest, usually a curving blood vessel. These are the sometimes anatomically confusing images that when rotated around their axis give a dancing-worm appearance. Any anatomy not associated with the curved structure of interest is distorted and appears misplaced. The images are created by depositing points along the centerline of the blood vessel, enabling the computer to connect these points into a continuous, thin-section image of the entire length of the blood vessel. This image is typically rotated around its axis, bringing different structures into profile. CPRs overcome the problem of calcified plaque and stents that obscure the lumen. Common applications for CPRs include assessment of the carotid bifurcations, carotid siphons, coronary arteries (Figure 3), or any moderately to heavily calcified artery. This technique is extremely useful as a problem-solving tool for interrogating diseased segments of arteries, and is one with which radiologists should be familiar.
Figure 2. Calcified plaque obscures the lumen as the thickness of the maximum-intensity projection grows from 1.2 mm(left) to 12 mm (right).Support Radiologists have no business being the sole users of advanced visualization software. Our job has always been to interpret studies, not produce them. I have spoken with many radiologists who were the early champions of 3D in their departments, and who took on the task of doing all of the 3D reconstructions at the workstation. At one or two cases per day, it was manageable and even somewhat fun. Unfortunately, there was no underlying infrastructure to manage increased volumes, and as the cases grew from two per day to 20, and the radiologist was three days behind and staying late at the hospital to get the cases processed and read, it wasn’t so much fun anymore. The bulk of 3D postprocessing is being done, and should be done, by well-trained technologists in a protocol-based fashion in a 3D laboratory setting. This leads to quality and consistency, with datasets returned to PACS that can be viewed in a logical and repeatable fashion. I can’t emphasize the phrase well-trained enough. Three days of on-site applications training or a week at a 3D camp are not nearly enough training for a technologist. Advanced visualization requires an understanding of far more than buttonology, and it will take months for even the best technologists to get up to speed on how to process a case properly. Whether laboratory services are outsourced or developed in-house, the effort put into the 3D lab will reap huge benefits as case volume grows.
Figure 3. Example of a curved planar reformat of the soft plaque in the left anterior desending coronary artery.Access The best advanced visualization software in the world is worthless if it can’t be accessed by the radiologist. Given the relatively poor reimbursement for CTA, workflow efficiency is a key to reading larger volumes of CTA studies. If the radiologist needs to get up from the PACS reading station and go to a workstation in order to view or perform postprocessing, frustration will quickly set in, and this wonderful new technology will be underused. Thin-client architecture is the key to handling and distributing CTA studies within and outside the radiology department. CTA studies should be available for viewing, interpreting, and processing from any PACS reading station and, in fact, from any PC, whether in the hospital, at the office, or at home. The most efficient architecture for reading CTAs allows for the following workflow: CT images are sent to a thin-client server, where a 3D technologist can access them and produce a protocol-based set of reconstructed images that are pushed back to the PACS; the radiologist reads the study just like any other CT in the stack, but has the ability to access the thin-client server and use the 3D software right from the reading station for those few times when additional postprocessing by the radiologist is needed. Ideally, workflow can be resumed, so you don’t have to start from scratch, but can take advantage of the processing performed by the technologist. Those images can also be saved to PACS; when the interpretation has been completed, you move on to the next case. Routine Ben Felson, MD, taught us all the importance of a routine approach to image interpretation. This is critical to accuracy and efficiency in CTA, given the large number of images typically produced and the variety of disease processes that radiologists look for in them. In the simple world of angiography, only the blood vessels are seen and, therefore, only the blood vessels are interpreted. CTA requires a primary focus on vascular disease, but also a search for other findings of importance. This, indeed, is one of the strengths of CTA (and, of course, one of the strengths that radiologists have the unique ability to bring to CTA). I like to read a study in the same fashion in which an angiogram is read: by breaking it down into a series of vascular territories. Since the reports tend to be longer than typical CT reports, I break up my report into sections, each labeled for ease of reading. I start with the extravascular findings and essentially read the case as a straightforward CT scan, with the exception that in my reporting, I don’t explicitly mention negative findings. If there are no significant extravascular findings, I simply state that. I then move to the vascular portion of the report. A typical template for a carotid CTA report includes:
- extravascular findings,
- aortic arch,
- posterior circulation,
- cervical carotids,
- intracranial circulation, and