Giant Cell Arteritis Detection

Machine Learning Algorithm to Analyze Histopathologic Sections of Temporal Artery Biopsy Specimens

Computer Vision
Deep Learning
Data Science
Image Processing
Statistical Analysis
Research
Author

Nivedhitha Dhanasekaran

Published

July 15, 2022

Project Poster

Download PDF file.

Project Description

In this project, I created an automated system to detect and diagnose Giant Cell Arteritis (GCA) from digital pathology slides using a deep residual convolutional network (ResNet) model for ROI-level classification, which produced a 91.65% whole-slide inference score. I collaborated with the core research team of clinical professionals to determine the inclusion and exclusion criteria for the design studies and randomized datasets, rigorously moved dates to protect patient privacy, and employed a common naming strategy across multiple health databases. Finally, I used GradCAM representations to visualize the model’s findings for medical specialists to validate. This system is the first published automated detector for Giant Cell Arteritis, and we plan to patent this technology soon.