MICCAI Industrial Talk: Deep implicit statistical shape models for 3d medical image delineation

MICCAI Industrial Talk Series @ June 30, 2022, by Dr. Adam P. Harrison from Q Bio.

Abstract: 3D delineation of anatomical structures is a cardinal goal in medical imaging analysis. Prior to deep learning, statistical shape models (SSMs) that imposed anatomical constraints and produced high quality surfaces were a core technology. Today’s fully-convolutional networks (FCNs), while dominant, do not offer these capabilities. In this presentation, we give a brief survey of this history and discuss efforts to bring SSMs properly into the deep learning era. As part of this, we present deep implicit statistical shape models (DISSMs), a new approach that marries the representation power of deep networks with the benefits of SSMs. We discuss how DISSMs use an implicit representation to produce a compact and descriptive deep surface embedding that permits statistical models of anatomical variance. Additionally, we explain how we can estimate rigid and non-rigid poses of this model using a Markov decision process (MDP). We then summarize experiments on liver and larynx segmentation where we show that DISSM outperforms the best FCNs we could find for both structures, indicating that the deep SSM approach to segmentation is a promising direction for continued research.

Speaker: Dr. Adam P. Harrison, Senior Machine Learning Scientist at Q Bio

Bio: Dr. Adam P. Harrison is a Senior Machine Learning Scientist at Q Bio where he works on developing cutting edge image analysis techniques for Q Bio's emerging MRI technology. Prior to his current position, Adam worked as a Staff Research Scientist at PAII, where he led efforts on computer-aided diagnosis and detection solutions for the liver, using ultrasound, CT, and MRI modalities. He has also worked at NVIDIA as an Applied Research Scientist and conducted a two year post-doctoral fellowship at the National Institutes of Health. He completed his PhD at the University of Alberta in Canada, where his dissertation won the best thesis award in Electrical and Computer Engineering. His research has been recognized in several manners, including two RSNA Trainee research prizes, a shortlist for the MICCAI Young Investigator Award, and work under his supervision has been presented orally at MICCAI and AAAI.