Unlocking Hearts by Locking Models: Solving Echocardiogram Problems Using a Self-Supervised Approach

Explore a groundbreaking self-supervised approach to solving multiple echocardiogram problems in this 20-minute conference talk from the 2023 GAIA Conference. Delve into the Adaptive Locked Agnostic Network (ALAN) concept, which enables label-free, self-supervised learning of visual features for medical imaging applications. Discover how this methodology overcomes the challenge of acquiring labeled data by training on large unlabeled datasets, producing rich informational features applicable to various downstream tasks. Examine the implementation of this approach in two typical echocardiogram tasks: anatomical region segmentation and view classification. Gain insights into the evaluation process using three publicly available datasets: EchoNet-Dynamic, CAMUS, and TMED-2. Learn from Anders Hildeman, a PhD holder in Spatial Statistics with extensive experience in predictive modeling of digital images and geospatial data, as he shares his research in medical image analysis and explores diverse methods and data types in the pharmaceutical industry.