Deep predictive modeling of cancer imaging based on weak supervision and atypical losses
Recrutement: 
Recrutement en cours/passé: 
Recrutement en cours
Periode: 
2020-2021
Contact: 
carole.lartizien@creatis.insa-lyon.fr

See a detailled description in the attached pdf

---------------------------------------------------------

Host laboratory : Laboratoire CREATIS, 69 Villeurbanne- MYRIAD Team

Supervisors : Carole Lartizien - carole.lartizien@creatis.insa-lyon.fr

Keywords : Medical Image analysis and Modeling, Deep Learning, Diagnosis model, weakly supervised learning

Duration : 6 months.

Starting date : Autumn or Winter 2020-21

Gratuity  : ~560 euros/month

---------------------------------------------------------

Our team has developed strong skills in the design of computer aided diagnosis and detection tools (CAD) for cancer and brain imaging based on the most advanced machine learning techniques.Such systems are designed to assist clinicians in their diagnosis by highlighting abnormal regions in an image.

One current active project concerns the prototyping of a CAD tool for prostate cancer screening based on multiparametric magnetic resonance imaging (MRI). One of the recent model reported (see below) is a novel end-to-end multi-class deep attention network that jointly performs prostate segmentation and cancer lesions detection with Gleason score (cancer aggressiveness) grading [Duran et al, MIDL 2020]. Our model is among the first to perform multiclass segmentation and achieves good performance without requiring any prior manual region delineation in clinical practice.

The purpose of this master project is to improve the  performance achieved with the current model architecture by exploring two methodological research axes on weakly supervised learning and atypical loss functions (see attached pdf file for details). The research program will be defined from the review of the state-of-the art bibliography at the beginning of the master project.The successful master candidate will have access to a multiparametric MR imaging database shared by our clinician partner (including more than 250 annotated cases and 250 non annotated clinical exams) as well as   datasets from challenges such as PROSTATEx-2 (https://www.aapm.org/GrandChallenge/PROSTATEx-2/).

Candidate should have background either in machine learning and/or deep learning or image processing and some experience in both fields as well as good programming skills.

We are looking for an enthusiastic and autonomous student with strong motivation and interest in multidisciplinary research (image processing and machine learning in a medical context). The candidate will also have the opportunity to interact with a PhD student working on this project.