Classification of COVID-19 via Homology of CT-Scans

Explore a 46-minute conference talk on applying topological data analysis to classify COVID-19 using CT scan images. Delve into the innovative approach of using persistent homology to quantify topological properties of SARS-CoV-2 features in medical imaging. Learn about the model's impressive performance metrics, including a 99.42% F1 score and 99.41% accuracy, when tested on a dataset of 2,481 CT scans. Discover how this TDA-based method mimics professional medical analysis and offers an intuitive way to detect anomalies in biomedical images. Follow the presentation through various topics, including perceptronomology, intensity plots, persistent diagrams, ground glass opacities, and the robustness of the model against noise. Gain insights into the visualization techniques and topological variations used in this cutting-edge application of algebraic topology to COVID-19 diagnosis.

Introduction
Perceptronomology
Death
Problem
Intensity plots
Persistent diagrams
Ground glass opacities
Low star filtration
Results
Questions
Persistence images
Learning about COVID
Robustness against noise
Pipeline time
Visualization
Topological variation
Feature vector
Image