Sepehr Jalali

Papers

Total Citations

154

H-Index

About

Sepehr Jalali is a researcher whose work sits at the intersection of deep learning, medical image analysis, and real-world clinical deployment. His primary research focus is on developing efficient, high-performance neural network architectures for retinal image analysis—particularly for vessel segmentation and intra-operative tracking. Jalali’s most significant contribution is the M2U-Net architecture, which not only achieved state-of-the-art segmentation accuracy on two benchmark datasets but was also the first model capable of running in real time on high-resolution retinal images. With over 114 citations, this work has been widely recognized for its practical impact, as its small memory and processing footprint make it deployable on mobile and resource-constrained devices—a critical step toward democratizing AI-assisted diagnostics. Jalali has also advanced the field of robotic retinal surgery by developing a deep convolutional network that jointly learns semantic segmentation and optical flow for intra-operative tracking of the retinal fundus, enabling sustained, precise delivery of regenerative therapies. His research stands out for its dual commitment to algorithmic innovation and real-world applicability, bridging the gap between cutting-edge computer vision and the constraints of clinical environments.