Object Recognition for Large-Scale and Weakly-Labelled Medical Image Data

The main objective of this research project is to investigate, develop and evaluate state-of-the-art image processing and machine learning algorithms, which are suitable for accurate modeling and recognition from large-scale medical image datasets that are weakly labeled. In particular, we will focus on the learning of recognition models in medical image computing applications that are of high interest/priority to Corstem, for instance, finding automatically the left and right ventricles in short- and long-axis cardiac magnetic resonance (MRI) sequences, which yields diagnosis measures that are of high interest to clinicians. Learning recognition models in medical image computing typically leads to difficult computational problems, where imaging data sets are weakly annotated.
This set of projects will leverage some limited and targeted interactions with medical experts, as needed, to set anatomical constraints and to drive advanced learning methods. TO BE CONT'D

Jose Dolz
Hoel Kervadec
Faculty Supervisor: 
Ismail Ben Ayed
Partner University: