Towards Efficient AI for Biomedical Imaging: Reducing Computation and Data Demands

Explore a comprehensive lecture on efficient AI solutions for biomedical imaging presented by Md Mostafijur Rahman, a PhD candidate from the University of Texas at Austin. Learn about innovative decoder architectures designed to overcome computational limitations and data scarcity in medical imaging applications. Discover the Cascaded Attention Decoder (CASCADE), Graph Convolution-Based Cascaded Decoder (G-CASCADE), Efficient Multi-Scale Convolutional Attention Decoder (EMCAD), and Efficient 3D Decoder (EffiDec3D) - all engineered to enhance segmentation accuracy while reducing computational demands. The talk also covers methods for improving data efficiency through Perturbed Prompts for Segment Anything Model (PP-SAM) and a training-free dataset pruning framework (PRIME) that significantly reduces annotation costs. This Stanford University MedAI Group Exchange Session provides valuable insights for researchers and practitioners working at the intersection of artificial intelligence and healthcare imaging technologies.