Medical image analysis using deep learning has recently been prevalent, showing great performance for various downstream tasks including medical image segmentation and its sibling, volumetric image segmentation. Particularly, a typical volumetric segmentation network strongly relies on a voxel grid representation which treats volumetric data as a stack of individual voxel ‘slices’, which allows learning to segment a voxel grid to be as straightforward as extending existing image-based segmentation networks to the 3D domain. However, using a voxel grid representation requires a large memory footprint, expensive test-time and limiting the scalability of the solutions. Furthermore, it is arguable that in traditional CNNs, its pooling layer tends to discard important information such as positions and CNNs are sensitive to rotation and affine transformation.

In this presentation, we present Point-Unet (accepted by MICCAI 2021), a novel method that incorporates the efficiency of deep learning with 3D point clouds into volumetric segmentation. We also introduce 3D-UCaps, a 3D voxel-based Capsule network for medical volumetric image segmentation (MICCAI 2021).

The DART Monthly Webinar Series provides an opportunity for project faculty and students to share short presentations about their work followed by open discussion and questions. Webinar topics will rotate between project management items like reporting and overviews of DART as a project, as well as specific research topics, important research results, and more.

All webinars will be recorded and made available after presentation. Please check https://dartproject.org/webinars/ for recordings.