Distributed Deep Learning for Cancer Cell Typing and Tumor Purity

Explore a groundbreaking 18-minute conference talk on distributed deep learning for cancer cell typing and tumor purity analysis. Discover how Providence St. Joseph Health is revolutionizing digital pathology workflows using AI/ML vision models for precise tumor analysis from H&E stained slides. Learn about their innovative approach leveraging Azure Databricks to distribute complex image processing tasks across a Spark cluster, resulting in a tenfold speed increase per Whole Slide Image (WSI). Delve into strategies for overcoming OpenSlide file management challenges through caching across executors, implementing parallel processing with a pre-trained StarDist model for thousands of WSI tiles, and applying GIS-style spatial joins for accurate cell labeling. Gain insights into how this breakthrough technology significantly enhances large-scale genomics research and advances digital pathology. Presented by Robert Kramer, Principal Data Scientist at Providence Health & Services, this talk offers valuable knowledge for professionals in healthcare, data science, and AI/ML fields.

Distributed Deep Learning for Cancer Cell Typing and Tumor Purity