Mission
Build a clinically interpretable imaging workflow where segmentation priors and deep learning outputs can be reviewed together instead of as isolated results.
About PRISM-3D
Periskeletal Region–aware Imaging for Survival Modeling
PRISM-3D is a research pipeline exploring whether targeted anatomical context can improve prognostic modeling for NSCLC. The current demo focuses on transparent workflow design, visual explainability, and clean clinical-facing interaction.
Demonstration Focus
End-to-end UI for uploading CT, running segmentation-guided inference, and inspecting Grad-CAM overlays across three anatomical planes.
Team
A team of three seniors from Green Level High School, we are the developers of PRISM-3D.
Tianxi Liang
Tianxi is an incoming Emory freshman majoring in Chemistry who is chasing the sweet spot where medicine meets machine learning. He wants to become a physician, yet can't resist building medical AI. On our team, he serves as the ML engineer, designing the end-to-end pipeline, training and tuning 3D model systems of PRISM.
Nishanth Sathisha
Nishanth is an incoming Carnegie Mellon mechanical engineering freshman who weirdly lives at the intersection of hardware brain and bio curiosity. When he is not working on PRISM, he is probably geeking out over golf as a side quest. On our team, he leads segmentation, and is the key behind the anatomical reasoning behind our model.
Sai Maruvada
Sai is an incoming UT Austin McCombs freshman headed toward investment banking, but he still gets oddly excited about statistics. He works closely with Tianxi, and complements him by zooming in on the results side of the pipeline, making sure our numbers are real and our validation is rock solid. He also leads external outreach with partner hospitals.