Deep-stratification of the cardiovascular risk by ultrasound carotid artery images
Cardiovascular risk estimation functions predict the risk of cardiovascular events with clinical data and survivalmodels. These functions accurately stratify individuals into low, moderate, and high-risk categories. However,they tend to classify a considerable number of individuals into the middle-r...
| Autores: | , , , , |
|---|---|
| Tipo de recurso: | artículo |
| Estado: | Versión publicada |
| Fecha de publicación: | 2024 |
| País: | España |
| Institución: | Universidad de Barcelona |
| Repositorio: | Dipòsit Digital de la UB |
| OAI Identifier: | oai:diposit.ub.edu:2445/224486 |
| Acceso en línea: | https://hdl.handle.net/2445/224486 |
| Access Level: | acceso abierto |
| Palabra clave: | Aterosclerosi Ecografia Doppler Atherosclerosis Doppler ultrasonography |
| Sumario: | Cardiovascular risk estimation functions predict the risk of cardiovascular events with clinical data and survivalmodels. These functions accurately stratify individuals into low, moderate, and high-risk categories. However,they tend to classify a considerable number of individuals into the middle-risk category, and often, a subsequentreclassification into high-risk groups is required. Atherosclerosis is the leading cause of cardiovascular events,and ultrasound images of the Carotid Artery (CA) can detect its burden by measuring the carotid intimamediathickness and identifying atherosclerotic plaques. Current risk estimation functions do not considerultrasound imaging. This paper proposes the use of deep ultrasound CA image features in survival models toimprove risk stratification. In particular, we define new deep CA image features, extracting information froma convolutional neural network, and add them to an existing risk function. The experiments carried out showthat using deep image features improves the AUC of the risk function to 0.842, and these features are enoughto replace the information provided by blood biomarkers. Furthermore, the use of these features resulted in a20% improvement in the reclassification of risk categories, specifically for individuals who suffered an event,as shown by the net reclassification improvement metric. |
|---|