Computer Vision News - August 2022

53 Constance Fourcade (Medical Image Registration Regularized By Architecture) , which combines the strengths of both conventional and DL- based approaches within a Deep Image Prior (DIP) setup (Fig. 1). We validated the three types of approaches (conventional, DL and MIRRBA) on a private longitudinal PET dataset obtained in the context of the EPICUREseinmeta project. Our proposed method performed better than all conventional approaches [Fourcade et al., 2020, Deformable image registration with deep network priors: a study on longitudinal PET images ].  – Finally, the third contribution is the evaluation of the biomarkers extracted from lesion segmentations obtained from the lesion registration step. We proposed a protocol to evaluate tumor response in a case with multiple lesions. Our method provides a new visual tool for the monitoring of metastatic breast cancer (Fig. 2) [Fourcade et al., 2022, PERCIST-like response assessment with FDG PET based on automatic segmentation of all lesions in metastatic breast cancer ]. The objective of this PhD thesis was to assist physicians monitor metastatic breast cancer patients with longitudinal PET images and improve tumor evaluation by providing them tools to consider all regions showing a high uptake (aka hot regions). This thesis describes three contributions in this direction: – Our first contribution is a method for the automatic segmentation of active organs (brain, bladder, etc.) based on a combination of superpixels and deep learning [Fourcade et al., 2020, Combining Superpixels and Deep Learning Approaches to Segment Active Organs in Metastatic Breast Cancer PET Images ]. – Our second and main contribution formulates the segmentation of lesions in the follow-up examination as an image registration problem . The longitudinal full-body PET image registration problem is addressed first with conventional optimization-based methods and second, with the more recent deep-learning (DL) approaches. In particular, in this thesis, we developed a novel method called MIRRBA