Aortic stenosis analysis using a novel fully automatic AI powered toolbox

Principal Investigator: Farhad Nezami

Authors: Amir Rouhollahi, James Willi, Jonathan Brown, Tsuyoshi Kaneko, Elazer Edelman, Farhad Nezami
Lay Abstract

As the population ages, valvular diseases such as calcification of aortic valves also known as aortic stenosis, are becoming more prevalent mandating updated therapy such as recently acclaimed catheter-based prosthetic valve implantation. Clinical decisions to choose between surgery and minimally-invasive intervention, and procedural setting for the latter, are still at the discretion of clinicians and not properly informed by patient-specific anatomy and disease state. We have developed a novel image processing technique powered by machine leaning algorithms to delineate the morphology and distribution of calcium in patients with valvular disease and predict major possible complications for patients treated with prosthetic valves. The extent of calcium and localization were quantified with novel geometrical metrics weighted by calcification distribution patterns to better classify patients and guide therapy. The resulted information for calcium characterization will be correlated with device performance metrics and clinical outcomes for the patients to relate geometrical features with clinical observations. Engineering tools akin to what developed herein provide indispensable means for digital medicine and individualized therapy wherein physics-informed measures of disease state would provide insightful inputs for clinical decision making and therapy planning. Such digital substrates as well provide unique platforms to design and develop emerging medical devices.

Scientific Abstract

Aortic valve stenosis (AS), the most prevalent valvular heart disease in developed countries, is a condition characterized by substantial calcification of the leaflet and often surrounding aortic tissues. Surgical aortic valve replacement (SAVR) is increasingly complemented by transcatheter aortic valve implantation (TAVI). Yet, adverse clinical events associated with the design of prosthetic valves and interventional guidelines are compromising theTAVI application. Engineering approaches, including physics-informed image processing algorithms, are being leveraged to optimize the design of emerging prosthetic devices and update interventional guidelines. Herein, we introduce a fully automatic analysis of calcification morphology and distribution to predicts the outcome for the given clinical image. Medical images were obtained through internal IRB to secure chest CT for >100 AS patients and exported in DICOM format. After adequate image preprocessing, automatic segmentation of both cusps and calcifications were conducted. This fully automatic algorithm detects the patient-specific geometry of valve as well as its landmarks and, extracting the morphology of calcification, quantifies its distribution pattern relative to non-coronary as well as left and right coronary cusps. Such unique approach would allow grouping AS patients according to calcium morphology to attribute clinical events and the response to TAVI to calcium extent and phenotype.

Clinical Implications
We developed a tool which allows insightful risk stratification strategies and quantified inputs to enhance clinical decision making and therapy planning. It will be further extended leveraging AI-powered algorithms to provide real-time calcium characterization beyond traditional approaches.

If the PDF viewer does not load initially, please try refreshing the page.

Comments are closed.