CASE-IN-POINT: Utilizing automated AI assisted imaging systems to improve workflow and avoid critical pitfalls in vascular imaging.

Edward Steiner, MD, FACR

These case studies are right off my desk. This current case was on my morning reading list, so I try look for pitfalls and learning points to benefit all. We were doing a TAVR aortic valve reconstruction and I recently read an article on renal artery stenosis and the predictive outcomes of TAVR (Atherosclerosis on CT Angiogram Predicts Acute Kidney Injury After Transcatheter Aortic Valve Replacement. AJR Am J Roentgenol. 2018 Sep;211(3):677-683.)

CASE STUDY:

81 y.o Female presents for a CTA of the Chest, Abdomen, and Pelvis for TAVR valve replacement and reconstruction. She is hypertensive and has diffuse atherosclerotic vascular disease. Vascular evaluation utilized standard technique with conventional reconstruction. Evaluation, however focused on the renal arteries because of her hypertensive history.

SCANNING PARAMERTERS AND TECHNICAL FACTORS:

·      Philips iCT 256 Slice Scanner iDOSE (4), Slice Thickness 2mm, 1556 slices in full acquisition.

·      Post Processing Philips ISP Enterprise Advanced Vascular Suite.

·      Post Processing on a standard diagnostic desktop, 19 seconds in Advanced Vascular Imaging Mode. No technologist needed for post processing for the images presented. Screen captures performed on-the-fly with no further manipulation.

FINDINGS:

Fig 1: Shaded 3D volume recons of the Aorta and renal arteries clearly demonstrate bilateral renal artery origin calcification, severe on the left. The arrow on the left renal artery warrants more interrogation.

Fig 2: One click calcium removal tool on the ISP Advanced vascular app now shows what appears to be a critical stenosis (arrow).

Fig 3: Review of the more conventional MIP's automatically displayed on my second monitor confirms that the right renal artery is clear, but still shows a very concerning calcified lesion in the left (arrow). One can be misled in dictating a critical stenosis and advising angioplasty prior to TAVR valve replacement. Given the high incidence of renal toxicity and failure, this would result in more procedural time, risk of stent complications and bleeding, and more contrast.

Fig 4: We immediately went to the ISP automated vessel extraction module and selected “Left Renal” in my Vessel List drop-down (left upper corner). With desktop processing of approximately 12 seconds the curvilinear luminal reconstructions appeared on my three right panels, essentially unfolding and dynamically rotating the vessel around an automated central endoluminal axis defined by the program. This was the difference maker in my diagnosis. This patient has a classical pattern of eccentric plaque and needed absolutely no intervention.

CONCLUSION:

Radiology is filled with pitfalls and potential errors. This case is not unique but illustrates basic known concepts in vascular imaging. Programs such as ISP are geared towards limiting such errors by putting high level computer processing and AI at our fingertips without human intervention. This application identified the vessel, subtracted bone and overlying distractions, and identified vascular calcification. On a deeper level, it identified the left renal artery central lumen, and without intervention displayed the vessel “interior” by finding and drawing the central axis. My diagnosis went from “critical left renal artery stenosis” to “eccentric left renal artery origin plaque with no significant stenosis.” All this in under 3 minutes.

I do want to make it clear that the above diagnosis should be made by most vascular imagers using conventional post processing and MPR's. A famous professor and mentor of mine in Boston once told me, however..."Ed, you can row a boat standing, but there are better ways of doing it". We are now in a position of defining and evaluating "the better ways." I try to look for such examples every day to define the extent and depth of this technology.