< Back to previous page

Project

Learning-based strategies for analysis of shape patterns in medical images

Medical image computing relies on mathematical models to enable automated analysis with sufficient accuracy and robustness. The use of supervised machine learning techniques, especially deep learning, has shown great promise to derive such models automatically from training data by formulating the analysis problem as a classification task based on non-specified local features that are assumed to be homogeneous within and between all images separately. We wish to investigate how such approaches can be adapted for characterisation of anatomical shape and shape variation over time and between subjects in different applications. This typically includes dispersed non-local patterns derived from dense spatial correspondences between heterogeneous images analysed jointly for which training data is limited.
Date:1 Oct 2018 →  Today
Keywords:medical image computing, shape analysis, machine learning, image registration and segmentation, clinical and forensic applications
Disciplines:Multimedia processing, Biological system engineering, Signal processing