Stage Data Science - Cancérologie

Réf. 996760 - publié le 16 octobre 2024

Informations générales

Missions

Unicancer vous propose une offre de stage dans les domaines Informatique, télécom, Bases de données, Développement, Génie logiciel, Informatique de gestion,... à Lyon.

Title : multi-agent for clinical decision support in molecular tumor board

Description of the project :
Many improvements in AI for oncology have recently led to the early testing of multi-agent for clinical decision support systems.
For a review see https://authors.elsevier.com/sd/article/S266677622400231X.
A recent paper (https://www.nature.com/articles/s41591-024-03097-1) have shown limitations in the case of abdominal emergencies, while providing a very nice open-source code available at https://github.com/paulhager/MIMIC-Clinical-Decision-Making-Framework/tree/main

Our applications at CLB-CRCL are to orient patients affected by cancer to a treatment accessible from their molecular analysis.
Compared to this paper, the specificities of our task requires to :
1) use RAG and/or web-based data related to the patient case derived from PubMED and other educational ressources
2) use other agents such as image analysis for ct scans (for example MedSam, or https://arxiv.org/abs/2405.12971), or digital pathology (https://www.nature.com/articles/s41586-024-07618-3) molecular data (CGI-clinics), etc…
3) adapt our internal clinical cases dataset to a format dedicated to the code
4) adapt the prompting to our domain

For that, we have a team with an experienced Data Scientist, engineer and medical trainees, and a coordinator. Our data consist in a small number of generated synthetic (not identifiable) patient cases that can be accessible remotely.

The objective of the internship is to master and adapt the code to the tasks and data format, and make a preliminary evaluation of the performance of the tools.

The deliverable is the code, comments, and proof of concept it works on few patients’ cases. A perspective is to evaluate this code on reall patients, internally at CLB.

Environment
At CRCL, the Saintigny team focuses on optimizing personalized treatments in oncology. The Data Science group conducts research in machine learning applied to textual and molecular data for personalized medicine in oncology. PôleData is a team of senior Data Scientists, Data Managers, Bioinformaticians and Doctors at CLB and CRCL.
The team, coordinator and CLB-CRCL are located in Lyon, but remote supervision is possible.

Profil

Mastering Python, and the libraries used in the Github above, ability to work collaboratively and independently, curiosity and scientific research spirit.

Postuler

Nom du recruteur : Sherwin Rebecca, Najar Stéphane