Interests: Physics Based Modeling | Animation | Medical Imaging

Histopathology Direct 3D Reconstruction with ML

Jul 2021–Cur 2024 | Ultrasound Lab @UW-Madison

[work in progress] This is one of my current projects at UW-Madison. I wrote a 3D CNN without a dense layer to directly output 3D models instead of transformation matrices for each slice or image. This tab is in progress but feel free to look around and view some images!

Carotid Plaque Histopathology & In Vivo Ultrasound Registration in 3D

Scroll to

see plaque

Mar 2023–Cur 2024 | Ultrasound Lab @UW-Madison

[work in progress] This is one of my current projects at UW-Madison. I reconstructed carotid plaque ultrasound and histology in 3D and correlated longitudinal in vivo ultrasound with longitudinal histology. This tab is in progress but feel free to look around and view some images!

Deep Learning-Based Carotid Plaque Histopathology Image Segmentation

Jul 2021–Sep 2024 | Ultrasound Lab @UW-Madison

[work in progress] This is one of my current projects at UW-Madison. I proposed that varying pixel resolution of the carotid plaque histology can help Mask RCNN learn histological features and artifacts with more "context" and successfully trained the model to segment carotid plaque histology into five classes (calcified, lipid core, fibrous, lumen, and background) by doing so. The current model has 90% pixel accuracy. This work has been published at Scientific Reports [view here]. This tab is in progress but feel free to look around and view some images!

Low Cost 3D Ultrasound

Jun 2019–Aug 2019 | REU Under Ultrasound Lab and IMMERS Lab @Stanford University

The reconstruction software previously developed in the ultrasound lab relied on openGL to reconstruct the ultrasound images in 3D, and it was only capable of displaying the reconstructed model on the openGL window. I rewrote the reconstruction process that actually builds models without relying on openGL in python with a runtime of 0.76 sec/image, and encoded the reconstructed models in a way that they can be exported into Unity for AR visualization, and 3D Slicer for segmentation, volume calculation, and other medical uses.

Modeling Water Surface

Jun 2018–Aug 2018 | Research Under Professor Gibou @UCSB

I worked on rendering water surfaces modeled by Professor Gibou. The surfaces were modeled using a novel approach that uses both octrees, and parallel implementation for free surface. These are images of the same water surface at different angles. They were rendered using radiosity, and photon mapping on PovRay.