Modern surgical systems rely on Convolution Neural Networks (CNNs) to segment anatomy in real time. However, convolution is computationally expensive when ran digitally. We show a system that can replace digital convolution with optical physics while ensuring feature fidelity for segmentation, low power compared to GPUs, and reduced cost and complexity.
Team Members: Saransh Bedi, Jaedin Garces, Chris George, Jonathan Jiang, Nabil Johny
Special thanks to Dr. Ian Bruce, Dr. Spencer Smith, Dr. Rafael Kleiman, Perla Yaghi, and Ramon Maldonado at Nokia for their support and resources throughout this project.