Mary Dong is a sophomore at Brown University concentrating in Applied Math-Computer Science. She's interested in the applications and social impacts of machine learning.
Deep Learning for ELVO Stroke Detection
Emergent large vessel occlusions (ELVOs) — the most disabling acute ischemic strokes — are primarily diagnosed through CT angiography (CTA), but require trained radiologists for rapid interpretation. Recent deep learning advancements show great promise to automate ELVO detection, accelerate downstream care delivery, and improve patient outcomes. In this project, we tested the fidelity of 2D convolutional neural networks (CNNs) to detect ELVOs. We first took maximum intensity projections of over 1000 brain scans of suspected stroke patients. We then used these preprocessed scans as training data for pre-trained architectures — including Inception v3, DenseNet 121, ResNet-50, and NASNet — as well as custom CNNs. We achieved a peak validation accuracy of 86.3% and a peak AUC of 0.917 with ResNet-50, which we adapted to the task at hand by replacing the top layers with three fully-connected layers. The high accuracy and AUC values achieved demonstrate the feasibility of using a fully automated, deep learning–based ELVO detection system to streamline ELVO diagnosis and treatment. These results can potentially be integrated into a platform that can accelerate and improve ELVO diagnosis and treatment.