Enhanced x-ray angiography analysis and interpretation using deep learning Over 1 Million diagnostic X-ray angiograms are performed annually in the US to guide treatment of coronary artery disease (CAD) and cost over $12 billion. Despite being the clinical standard of care, visual interpretation is prone to inter- and intra-observer variability. Recently as part of the NHLBI supported Prospective Multicenter Imaging Study for Evaluation of Chest Pain (PROMISE) trial, our research team showed that cardiologists misinterpreted over 19% of angiograms obstructive CAD (greater than 50% vessel stenosis). Given the centrality of angiographic interpretation to the development of a treatment plan, reduced accuracy can lead to unnecessary poor outcomes and increased costs to our healthcare system. Quantitative coronary angiography (QCA) offers a computed measure of disease severity. However, the time and training required to perform QCA has largely relegated its use to research applications. Recent advances in deep learning for medical image analysis offer an opportunity to address this problem. The potential impact is significant given increasing interpretation accuracy by 1% can positively benefit over 10,000 patients each year. Thus, our team proposes to develop an X-ray angiographic analysis system (XAngio) driven by deep learning technology to enhance physician interpretation. We are uniquely positioned to accelerate development of XAngio by leveraging our team?s PROMISE dataset of over 1,000 angiograms with expert QCA scoring. In Phase I, we will teach XAngio how to read angiographic images and how to discriminate obstructive CAD. XAngio will employ a convolutional neural network, a computer vision technique rooted in deep learning, to self-characterize angiographic features from labeled images. In order to infer additional information from vast amounts of unlabeled images, XAngio will incorporate a deconvolutional neural network technique to increase predictive performance. A pilot study of XAngio will test its ability to identify the presence of obstructive CAD in PROMISE angiogram images. If we are successful, we envision a Phase II proposal which is focused on clinical translation of the technology and assessment of its impact on improving visual interpretation in a cohort of cardiologists. Our goal is to combine recent advances in deep learning, big data from PROMISE, and scalable parallel computing to create XAngio. In the long term, we hope the combination of a cardiologist with XAngio as an assistive tool will improve the clinical accuracy of angiographic interpretation.
While invasive x-ray angiography is the gold standard diagnostic tool for guiding immediate treatment in the cardiac cath lab, visual interpretation of angiograms is challenging and subject to large inter- and intra-observer biases. In this project, we will develop and validate XAngio, a novel computer-assistive technology to enhance cardiologist?s angiographic reading and interpretation process. It is our hope that this technology will increase the accuracy of diagnosing obstructive coronary artery disease and, ultimately, improve patient outcomes.
