By Beth W. Orenstein
Vol. 21 No. 1 P. 22
Black-blood cinematic rendering offers photorealistic views of the inside of the heart.
Imagine filming the inside of the heart with a high-definition movie camera. The images of the walls and vessels would be breathtakingly lifelike. In a sense, that’s what researchers at Johns Hopkins University (JHU) have been able to do with a process they call black-blood cinematic rendering (BBCR). Steven P. Rowe, MD, PhD, and Elliot K. Fishman, MD, FACR, and colleagues at JHU recently published a paper in the Journal of Cardiovascular Computed Tomography in which they describe their method of generating photorealistic 3D images from volumetric CT data. The method provides extremely high levels of intraluminal anatomic detail for cardiac CT imaging, according to Rowe, the lead author.
“It’s startling how good the images look,” Fishman agrees.
Fishman describes the origins of using BBCR to shed light on intraluminal heart anatomy: “We’ve been using 3D forever and using cinematic rendering for roughly the past three years. One of our jobs is trying to figure out how to best use 3D in clinical practice.” In doing so, Fishman has created a set of windows/parameters that he uses each time he reads a study so he’s not starting from scratch. “What you try and then do,” he says, “is to develop different settings to show different things.” Fishman started with bones, “as they’re the easiest to do and to be able to show in detail.” He then moved from bones to soft tissue.
“Then I began working in the abdomen,” he says. “We put a lot of effort in looking at the pancreas. From there, we did work on the heart and trying to look at the heart and its vessels. We had done some intraluminal imaging with volume rendering way back when, which was OK, but we always thought that with cinematic rendering we could do it better. Cinematic rendering is so realistic, it’s like looking inside the vessel. So, I played around with formula calculations and came up with answers. It’s really just a lot of playing around, trying to adjust parameters, and trying to figure out where to place the different trapezoids [functions that overlay diagrams called voxel histograms, which dictate the color and transparency of voxels within the imaged volume] so you can make different features you’re looking at/for stand out.”
The timing for this method is ideal, according to Fishman and Rowe, because computers have become increasingly powerful in recent years. “It takes computing power to be able to pull this off,” Rowe says, and that power is now available on individual workstations. “Computing power is so readily available now, this is a pretty fast process.”
Rowe and Fishman have done their research with cinematic rendering using syngo.via VB30 software by Siemens Healthineers. The software was already integrated into CT workflow at JHU. “We have been creating cinematic rendering images daily for a variety of clinical scenarios,” Fishman says. Fishman also notes that, while he and his colleagues use Siemens—he has gotten some research support from the company—he believes his presets would work on other manufacturers’ systems and with other software.
Lighting Adds Depth
The idea for cinematic rendering, Rowe notes, has its origins in animation for movies and video games. It employs a global lighting model that adds depth to CT data sets. Cinematic rendering views thousands of photons and determines how they traverse an imaging volume—the way they would interact with material in the volume, he explains.
The researchers call their method “black blood” because the blood appears to go black or disappear while the walls of the heart appear lighter colored. “The blood, by definition, doesn’t disappear,” Fishman says. “But it gives you a setting where it’s transparent. So you have an interface between the blood, which is transparent, and the wall of the heart’s chambers or other structures. As a result, it looks as though you’re looking inside—as if you had the heart in your hand and you took the blood out. You would clearly see the wall of the chamber, and that’s what we’re trying to create with this perspective.”
Rowe says that the difference between 3D volume rendering and cinematic rendering is dramatic. “With older volume rendering, generally, there are no shadows, so the images, even though they are 3D, look artificial. With cinematic rendering, there are realistic shadows, so it is more lifelike than anything we’ve ever seen before.”
Fishman and Rowe developed presets to visualize intraluminal structures in the heart from previously acquired patient CTs. “You use the presets as a starting point and are able to make subtle adjustments to window width and level settings,” Fishman says.
He estimates that it takes him less than five minutes per study to create the cinematic rendering images. Of that five minutes, less than one minute is spent on manual adjustments, he says. “Everyone is pretty busy these days,” Fishman says, “and if they had to figure out the best way of doing this to the images, it would be too time consuming. But, thanks to the presets, it wouldn’t add too much time to their caseload.” Fishman also figures that fewer than 25% of cases would need manual adjusting.
Rowe says there is certainly a learning curve in creating the cinematic rendering images. “It does take a while to learn how to get the images we are seeking,” he says. “But we now have a lot of shortcuts built in. Radiologists aren’t starting from square one every time.”
Three Case Studies
In their journal article, the researchers discussed the advantages of BBCR over conventional CT and volume rendering in the following three cases:
• A 71-year-old woman being evaluated for a known ascending aortic aneurysm. The patient’s 2D CT scans appeared to be overexposed and showed little surface detail of her underlying cardiac chamber. To create the BBCR images, two presets, primarily of two nonoverlapping trapezoids—one subsuming low attenuation values and the other centered at higher attenuation—were used. When the blood appeared to be black, it showed the interior of the heart wall in much greater detail and revealed thinning of the left ventricle (myocardial trabeculation) and anterolateral papillary muscle.
• A 51-year-old woman being evaluated for renal infarcts secondary to left ventricular mural thrombus. Conventional CT, taken after the patient complained of abdominal pain, showed renal infarcts. The mural thrombus and renal infarcts were obvious to readers on both the CT and the BBCR images. However, with BBCR, radiologists could more clearly see the associated myocardial thinning as well as other structures, such as the pulmonary vasculature.
• An 80-year-old male undergoing follow-up imaging for prior transcatheter aortic valve replacement. The volume rendering image was optimized to show the patient’s left ventricular myocardial trabeculae. BBCR not only showed the myocardial trabeculation in the left ventricle, but it also provided a much more detailed view of the metallic stent in his aorta. It showed that the stent was well seated within the aortic annulus and a closed valve during diastole.
These cases demonstrate that BBCR provides high levels of intraluminal anatomic detail for cardiac CT imaging and can help physicians detect important pathologic entities, such as intramural thrombus, Fishman says. It also suggests a promising role in improving evaluation of cardiac devices such as stents.
One potential drawback, Rowe says, is that BBCR could, in some cases, obscure some pathology as a result of the shadowing effects. That could limit its application in some studies, he says, and he emphasizes the need for radiologists to be diligent in assessing how to use cinematic rendering. Fishman says the solution to this challenge is to do both conventional 3D volume rendering and BBCR.
“Radiologists follow search patterns and make sure they’re looking at all the key structures and things they need to evaluate with every case they read,” he says. “With any modality, you can obscure an important finding if you’re not careful. A lot of this is spending time getting used to it so you know what you’re looking for and making sure you’re comfortable using cinematic rendering.”
Would payers consequently reimburse radiologists who take the time to create two data sets? As with any read that involves a recent technological advance, it’s hard to know, Fishman says.
What other potential uses could this software technique offer? “We’re still working to figure that out, exactly,” Rowe says. “We have some ideas of what it’s good for.” Rowe definitely sees a role for cinematic rendering in presurgical planning for heart surgery and in talking to patients about their condition and treatment. “How wonderful it is to be able to pull up these cinematic rendering images and show our colleagues exactly what we are talking about,” Rowe says. “It gets everyone on the same page very quickly. I think that’s a great application and especially when it’s something that’s hard to quantify and communicate well with colleagues. This is a significant step forward.”
It could be very useful for the cardiac surgeon who is going in to correct the problem, Rowe says. “Their understanding of those organs and their relative positions to major vascular structures is important for the surgeon to properly approach a surgery. With BBCR, the structures and the relationships to each other show up very well.”
The same is true with patients. “Instead of showing them two-dimensional black-and-white images that take years of training to understand, you are able to show them something that they can easily see and explain things very well,” Rowe says. “I think many patients are going to be very appreciative of having these incredibly realistic images available to them.”
In addition to the heart, the researchers see a role for cinematic rendering for other organs including the pancreas, liver, and kidney. “Wherever the anatomy is complex, this could be important,” Rowe says. “We like it for things like the base of the skull and neck, where there are lots of tiny structures overlapping and crammed into a tiny area.” Some research has been done with cinematic rendering in imaging facial structures, fetal and placental anatomy, the esophagus, and the lungs.
Rowe says it will be difficult to design studies that put BBCR head to head with other 3D and volume rendering techniques. “With any new way of looking at imaging data, where what you’re looking at is so different than what you’re used to, there are always going to be readers’ inherent biases built in,” he says. It can be hard for readers to compare what they see that’s different on a BBCR image vs a 3D volumetric rendering or even a 2D reconstruction, he continues.
Although early indications are that the use of BBCR is warranted in many cases, the researchers say, much more research is needed before the technique becomes mainstream and widely accepted. “Radiologists are always going to have their preferences about what they find useful,” Rowe says. “It may be that they could learn many of the same things from scrolling through 2D images, but they can do it so much more quickly with this, and, in many cases, it makes it more intuitive to understand what you’re seeing.”
— Beth W. Orenstein of Northampton, Pennsylvania, is a freelance medical writer and regular contributor to Radiology Today.