Unlocking the Power of AI: In Silico Trials in Chest Radiology

Forum Details

Date: June 21, 2024 12-1 PM Eastern time

Speakers: Joseph Lo PhD and Fakrul Islam Tushar MSc

In this talk, we explore the transformative potential of AI for evaluating in silico clinical trials, focusing on three groundbreaking studies in chest imaging. First, we address the challenge of generalizing AI for COVID-19 classification with CT and CXR imaging, demonstrating how in silico phantoms produce realistic and trustworthy results compared to clinical data that is often flawed. Next, we showcase our emulation of the renowned NLST trial for lung cancer screening. Our VLST trial deployed a fully virtual pipeline encompassing patients, scanners, and readers, illustrating how in silico trials can answer important clinical questions efficiently and rigorously. Finally, we introduce the Duke Lung Cancer Screening dataset, providing 3000+ annotated lung nodules from 2000+ patients that were acquired on modern CT scanners, and present AI performance benchmarks. This presentation underscores the promise of AI to advance clinical research in chest imaging and beyond.