Renal Cell Carcinoma Classification Project

The last project I worked on during my undergrad

Above Reference Image Source: Toquero, Lawrence & Aboumarzouk, Omar & Abbasi, Zahir. (2009). Renal cell carcinoma metastasis to the ovary: A case report. Cases journal. 2. 7472. 10.4076/1757-1626-2-7472.

View Code [Github]

This project, completed through the Informatics Skunkworks group in collaboration with the University of Wisconsin-Madison School of Medicine, focused on automating the grading of renal cell carcinoma (RCC) using deep learning and radiomics.

  • Developed a machine learning workflow to classify CT scans of renal cell carcinoma into low-grade or high-grade categories, assisting in non-invasive diagnostic efforts.

  • Data Engineering & Pipeline Development: Optimized a segmentation script to isolate kidney regions from three-dimensional CT scans. Engineered a data pipeline to convert 3D .mat files into 2D image slices based on specific volumetric criteria (e.g., largest cross-sectional area or median slice).

  • Computer Vision & Transfer Learning: Implemented a transfer learning approach using a ResNet architecture. This required adapting pre-trained weights from general image datasets to the specific grayscale and textural nuances of medical CT imagery.

  • Technical Impact: Successfully bridged the gap between raw medical volumetric data and standard deep learning architectures, establishing a foundation for automated radiomic classification.