Last week, September 10-11, a researcher from the medical imaging research group of VICOROB attended the 3rd annual AI in Health Conference in Houston, home of the largest medical center in the world – the Texas Medical Center, hosted by the Ken Kennedy Institute at Rice University.
A collaborative work between our team from the medical imaging research group and McWilliams School of Biomedical Informatics at UTHealth Houston was presented at the conference and accepted as both an oral presentation and a poster. The recording of the talk has been posted to the Ken Kennedy Institute YouTube channel
The presented work proposes learnable 3D pooling (L3P), a convolutional neural network (CNN)-based module that encodes 3D data into multiple 2D feature maps in a learnable way. The L3P module is followed by a 2D lightweight CNN, to perform 3D brain image classification in an end-to-end manner, achieving results equivalent to full 3D models, but with fewer parameters and lower computational cost. L3P simplifies interpretability and model development and generalizes well across different datasets, which makes it a promising tool for a wide range of clinical applications.
Attending this conference was a great opportunity to connect with and learn from researchers, engineers, clinicians, and entrepreneurs at the forefront of artificial intelligence in healthcare and public health.
