| KEYWORD |
Self/weakly-supervised learning to improve Renal Cancer subtyping
Thesis abroad
keywords IMAGE ANALYSIS, MACHINE LEARNING, ARTIFICIAL NEURAL NETWORKS, MEDICAL IMAGING, SELF-SUPERVISED LEARNING
Reference persons SANTA DI CATALDO
External reference persons francesco.ponzio@polito.it
Research Groups DAUIN - GR-06 - ELECTRONIC DESIGN AUTOMATION - EDA
Description Renal cell carcinoma (RCC) is currently categorized into several main histological subtypes. Importantly, the outcome of RCC malignant neoplasms deeply relies on the accurate determination of the histological subtype. Although histological diagnosis of renal neoplasm is usually straightforward by routine light microscopy, the categorization of histological RCC subtypes typically shows a poor to fair agreement between pathologists, with a mean inter-observer value in the range 0.32– 0.55.
The aim of the thesis activity is to design and to validate a self/weakly-supervised learning-based system able to automatically classify tissue samples (in form of histological slides) into four RCC subtypes: clear cell RCC (ccRCC), papillary RCC (pRCC), chromophobe RCC (chRCC), and renal oncocytoma (ONCO).
The thesis period is abroad at the research center Inria Sophia Antipolis - Méditerranée.
Required skills Good programming skills (Python). Prior experience with machine learning/deep learning design frameworks is a plus.
Deadline 28/07/2024
PROPONI LA TUA CANDIDATURA