ASAIO 2021 Annual Meeting Abstracts.

Abstract

Study: Deep venous thrombosis (DVT) is said to be the most preventable cause of death in hospitals today (Centers for Disease Control, February 2020). This silent killer most frequently starts in the lower extremities as a small thrombus that initially forms in a valve pocket located behind a venous valve leaflet. Failure to initiate preventative measures to reduce the risk of DVT can result in the development of life-threatening pulmonary emboli (PE) and chronic venous insufficiency (CVI). These risks are currently exacerbated by the current SARS-CoV-2 pandemic that has increased the frequency of DVT formation and PE in patients through mechanisms that are not fully understood. The current study is using artificial intelligence (AI) and machine learning (ML) to evaluate changes in blood rheology, venous blood flow, and venous valve configuration that can lead to increased risk of DVT, PE, and CVI. Methods: In our model, AI/ML is used to evaluate previously recorded images of normal and abnormal cadaver veins and venous valves. Using these data, we can predict that previous DVT formation produced the observed leaflet damage and will increase the risk of CVI and future DVT formation. The contribution that changes in blood rheology could make to venous blood flow abnormality will also be assessed. We leverage the latest advances in image processing to quantify the valve leaflet geometry with respect to the susceptibility to recurrent DVT. Results: Based on this data, AI/ML will predict, on a patient-by-patient basis, the probability of deep venous thrombosis formation. This will allow healthcare professionals to determine when preventative measures are most appropriate. It will also allow physicians to assess whether a patient has an increased risk of developing a future DVT and/or CVI insufficiency. With this information patient care can be improved and the debilitating and life-threatening complications of DVT can be substantially reduced.