NU Radiology-Founded AI-Enhanced Medical Imaging Company Clearvoya Receives NHLBI STTR Grant
AI Update
In the 1980s, the medical imaging technique digital subtraction angiography (DSA) was developed to visualize blood vessels. As the need for minimally-invasive endovascular interventions increased, DSA emerged as one of the more common technologies applied in the treatment of stroke, myocardial infarction, and peripheral vascular disease, pathology that more adversely affects underserved and minority patient populations. The MPIs of the STTR grant focused on DSA technology and artificial intelligence, Dr. Sameer Ansari, Professor of Radiology, Neurology, and Neurological Surgery, and Dr. Donald Cantrell, Assistant Professor of Radiology and Neurology, met at Northwestern while Cantrell completed his residency in the Department of Radiology and both Diagnostic/Interventional Neuroradiology fellowships. “Just prior to COVID and toward the end of [Cantrell’s] fellowship, we had been discussing neurointerventional applications that could leverage rapidly developing machine learning technology and his expertise in this field,” Ansari says. These discussions were the beginning of a medical start-up called Clearvoya. Ansari recalls, “As we filed our first patent, we realized that this work and a pipeline of neurointerventional applications could be commercialized over time and we began to consider forming a company.” Leon Cho - co-founder of Clearvoya, MPI of the grant and Ansari’s longtime friend since high school in Baton Rouge, LA - had moved to Chicago where the two reconnected. “Since I knew he had a computer science background working in Silicon Valley for some time and also had completed an MBA from the University of Chicago, I asked him to assist us in developing our vision and, hence, Clearvoya came into existence!”
Research from Cantrell and Ansari has led to improvements in DSA imaging. “Catheter angiography is performed by inserting a small catheter into an artery and recording a series of X-Ray images as the contrast traverses the patient’s blood vessels,” Cantrell says of the technique. “However, superimposed X-Ray densities from bones and soft tissues obscure the imaging details of the blood vessels.” Images can also be degraded by voluntary, respiratory, or cardiac motion during the exam, which are common to routine clinical practice. “In situations where patients are unable to remain still, which may be due to difficulty breathing or the distress of an acute stroke, the poor quality of motion-degraded DSA imaging increases the risk of complex procedures such as stroke clot removal (thrombectomy) or cardiac (coronary) stenting.” In response to this need for better imaging and procedures, Cantrell and Ansari developed a deep learning algorithm to use DSA even with motion. Using Vision-Transformer-based network architecture, images identify blood vessels and separate them from other X-ray densities such as bone and soft tissue. A novel data augmentation mechanism trains the neural network to outperform the imaging techniques during patient motion. The STTR project (R41HL164298) titled “Motion-Resistant Background Subtraction Angiography with Deep Learning: Real- Time, Edge Hardware Implementation and Product Development” will utilize this algorithm for real-time application in minimally- invasive procedures with the overall aim to integrate this technology into angiography machines in the future.
Leon Cho • Co-founder of Clearvoya
Donald Cantrell, MD • Assistant Professor of
Sameer Ansari, MD • Professor of Radiology, Neurology, and Neurological Surgery
Radiology and Neurology
The Read | Volume 6 | Page 7
Department of Radiology
Made with FlippingBook - Online magazine maker