Advanced Cinematic Rendering
Advanced Cinematic Rendering |
Welcome and Overview Cinematic Rendering (CR) is at its core volume rendering with better lighting and texture mapping because of the availability of better hardware particularly GPU’s from Nvidia. The original paper by Kroes et al. builds on the work by Drebin et al. nearly 25 years earlier. Although the core technology in cinematic rendering has not changed over the last few years, how we use or potentially can use Cinematic Rendering is evolving. |
RSNA 1985  |
LucasFilms 1985  |
Pixar Image Computer 1988  |
Drebin et al 1988  |
“ Volumetric rendering differs from surface rendering in that all the information from the CT scans is preserved, not just surface boundaries. Object thickness and internal contours can be seen in the 3D projection.” Volumetric Rendering Technique: Applications for Three-dimensional Imaging of the Hip Fishman EK, Drebin RA, Ney DR et al. Radiology 1987 Jun;163(3):737-738 |
Kroes et al 2012  |
 |
 |
“In addition to the fact that photo-realistic volume renderings tend to be aesthetically more pleasing, it has been shown that realistic lighting contributes to 3D understanding and can improve depth-related task performance . With this work and the implementation that we have made available, we hope to contribute to the uptake of realistic illumination in interactive direct volume rendering applications.” Exposure Render: An Interactive Photo-Realistic Volume Rendering Framework Thomas Kroes et al. PLOS ONE 7(7): e38586. doi: 10.1371/journal.pone.0038586 |
Introduction Cinematic Rendering still is not mainstream but in fact 3D imaging which has been around for nearly 40 years (beginning with shaded surface technique) still is done in under 10% of practices worldwide. The literature has seen an uptick in articles about cinematic rendering and as of October 30, 2020 has had 86 articles published that are either specifically on cinematic rendering or use cinematic rendering as part of their work. |
86 Articles using Cinematic Rendering as of 10-30-2020  |
Introduction (cont.) However many of the articles are case reports of interesting cases while other are pictorial essays that focus on potential clinical applications. There are few articles which look at and measure the true impact of CR in clinical practice. In part as with any post processing technique it is hard to measure specific impact on clinical care with a rigorous study as it is impossible to create a blinded study with post processed images. Yet, work is beginning to try to measure the impact on our referring physicians and there use of CR. |
Introduction (cont.) In this refresher course on Advances in Cinematic Rendering which is now in PowerPoint format and not “live” in person we will try to approach the topic by using the PowerPoint format as a way of asking questions about where CR is going in our opinion but also provide you with some ideas that might be of value in your practice. We wish we could share and discuss the topic with you in person but perhaps RSNA 2021 or 2022 will provide that opportunity. |
Cinematic Rendering in Clinical Practice 2020 Based on the literature and our own experience some of the areas where CR has been applied to most successfully have been;
|
Cinematic Rendering in Clinical Practice 2020 The common ground for applications tends to be complex anatomy that is best defined in a 3D view and CR meets these needs by providing the most photorealistic display possible. Our colleagues doing the most complex surgeries realize the value of this technology. The question I often get is “why is this not done on all cases and when will this be the standard of care?”. Perhaps this will then be the focus of my presentation. That is, where is CR going and what are the advanced techniques and applications that will need to evolve to make this a reality. |
Challenges for Cinematic Rendering in Practice Some of the global challenges for Cinematic Rendering to become mainstream in Radiology are not unique to CR. These include;
|
Challenges for Cinematic Rendering in Practice We believe that if CR is valuable then the actual rendering will in most cases need to be done by the Radiologist. This is not to say that 3D labs are not valuable but with CR the current variability of image creation makes it easy to obscure critical findings and unless you are able to understand the pathology you are addressing it is easy to create errors in the images. The use of presets helps avoid some of these issues but we currently have over 150 presets so significant experience is needed. The creation of images that the referring doc wants and needs is usually best understood by the radiologist. This is especially true in multi-disciplinary conferences. |
Challenges for Cinematic Rendering in Practice
|
Challenges for Cinematic Rendering in Practice The optimal workflow for cinematic rendering would be if the “optimal preset” or a series of optimal presets (3-5 various presets)could be created by the computer and be presented to the radiologist or the referring physician. This would mandate that the presets are optimized based on the clinical question, the phase of acquisition and the quality of the dataset. Variations based on contrast delivery and timing of injection as well as patient size and body habitus would be needed. |
The range of visualizations in any case must change to visualize structures ranging from bone, to vasculature to muscle. Here is a case with IVDA and groin infection  |
Could Automated Presets be done by the Computer? The introduction of Artificial Intelligence and Deep Learning has the promise of reinforced learning where the computer can learn from information previously generated by expert annotation. We have previously begun work to look at whether the computer had the parameters of the look up table for each case and the image data it could design 3-5 best fit renderings for each case. This could then be shown to the radiologists who would select the best rendering with the ability to adjust the parameters in some cases. Th presets can transform into the data that the computer could generate on its own. |
Trapezoid Creation Presets are the key to the workflow and maintaining high quality  |
Trapezoid Creation  |
Current Preset Values  |
AI Direct Cinematic Rendering We believe that if the computer could help choose optimal CR parameters that the output to the radiologist could look like the following case studies. As AI is being trained to segment organs s well as to detect pathology (i.e. Felix Project for the early detection of pancreatic cancer) the process could be integrated into a new level of accuracy and patient care. Here might be the output we could expect in the future. |
Pancreatic Tumors and Cinematic Rendering
|
PNET Tail of Pancreas 1cm  |
Neuroendocrine Tumor TOP With Calcifications  |
Subtle 1 cm Neuroendocrine Tumor Tail of Pancreas  |
SPEN in 62 Year Old Female  |
Liver Tumors and Cinematic Rendering
|
Hepatoma  |
Hepatoma in Cirrhotic Liver  |
Hepatoma Invades the Portal Vein Using Various CR Presets  |
Small Bowel Tumors and Cinematic Rendering
|
1cm Duodenal GIST Tumor Looks Like PNET  |
GIST Duodenum Presenting as an Exophytic Mass with Cinematic Rendering  |
GIST Tumors in Patient with Neurofibromatosis Type 1  |