Supervised Machine Learning and Federated Learning Methods for Solving Classification Problems in Healthcare

• Grant number: 24/20660-0
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: May 01, 2025
• End date: April 30, 2026
• Field of knowledge: Physical Sciences and Mathematics - Probability and Statistics - Applied Probability and Statistics
• Principal Investigator: Benilton de Sá Carvalho
• Grantee: Erick Jun Miyagi
• Host Institution: Instituto de Matemática, Estatística e Computação Científica (IMECC). Universidade Estadual de Campinas (UNICAMP). Campinas , SP, Brazil

Abstract

Among women, breast cancer is the most prevalent type of cancer and the leading cause of death worldwide (according to 2020 data). In Brazil, excluding non-melanoma skin tumors, it is the most incident (as per estimates for 2023). To improve the treatment of this type of cancer, the study of gene expression profiles in tumor tissues has been intensively conducted, and oncologists have been applying the obtained results. These data are used to classify tumors into 4 subtypes. Machine Learning (ML) methods have been employed to predict cancer types based on RNA sequencing (RNA-seq) data, a method for analyzing cell transcriptome (RNA) profiles. To ensure efficiency and generalizability of the constructed models, it is necessary to use large quantities of data from various sources, which is conventionally solved by data centralization. However, this mechanism may infringe upon governmental regulations concerning data privacy. To circumvent this problem, Federated Learning (FL) techniques have been implemented, allowing the development of collaborative Machine Learning models without sharing individual data. As a result, this research project aims to develop technical-scientific knowledge about constructing an ML and FL model for classifying breast cancer subtypes using publicly available data from TCGA, neural network algorithms, and ML and FL frameworks. (AU)