July 1, 2019 – vRad (Virtual Radiologic), a Mednax company recently made a scientific presentation, “Screening for Aortic Dissection on CT Angiography Using a Convolutional Neural Network,” at the Society for Imaging Informatics in Medicine (SIIM) Annual Meeting, June 26-28 in Aurora, Colo.
The presentation highlighted how vRad, using machine learning on retrospective aortic dissection data, created an artificial intelligence (AI) model for classifying and prioritizing aortic dissection and rupture for all post-contrast chest computed tomography (CT) imaging data passing through vRad’s patented reading platform. The model is integrated into vRad radiologists’ workflow to accelerate time-to-care. The presentation also outlined how vRad, in conjunction with the Mednax Radiology Solutions Artificial Intelligence (MDR-AI) Incubator, developed a convolutional neural network model trained on these aortic datasets to assess how likely these pathologies would be present in a chest CT.
Aortic dissection is a condition in which the aorta wall tears, causing blood to flow through the torn region as the middle and inner layers of the wall separate. This condition can be fatal if not recognized and treated immediately. The ideal imaging technique for identification of aortic dissection is computed tomography angiography (CTA) of the aorta. A fast method of screening chest CT studies for this pathology can assist with efficient patient triage and diagnosis.
The work developing the model also includes efforts of the MDR-AI Incubator, which launched in November 2018. The Incubator brings together radiologists, rich clinical datasets and technology partners to stimulate innovation and product development in radiology, improving clinical accuracy, efficiency, and the overall quality of patient care. The AI Incubator’s approach involves developing models that have immediate and measurable impact, so that medical and technical leaders can build tools that improve the practice of radiology for physicians and their patients.
For more information: www.vrad.com
