Interpretable Machine Learning for Medical Applications based on Images

• Grant number: 22/05788-4
• Support Opportunities: Scholarships in Brazil - Doctorate
• Start date: May 01, 2023
• End date: September 30, 2025
• Field of knowledge: Physical Sciences and Mathematics - Computer Science
• Principal Investigator: André Carlos Ponce de Leon Ferreira de Carvalho
• Grantee: Iury Batista de Andrade Santos
• Host Institution: Instituto de Ciências Matemáticas e de Computação (ICMC). Universidade de São Paulo (USP). São Carlos , SP, Brazil
• Company: Universidade de São Paulo (USP). Instituto de Ciências Matemáticas e de Computação (ICMC)
• Associated research grant: 20/09835-1 - IARA - Artificial Intelligence in the Remaking of Urban Environments, AP.PCPE

Abstract

Machine learning has been applied to many problems in the last decades, reaching expressive results, especially on computer vision, highly attributed to approaches like deep learning algorithms. In the context of medical applications for computer-aided diagnostic from images, the use of deep learning algorithms is highly valuable, being a field traditionally linked to advances in computer vision. However, using solutions adopting deep learning finds resistance, like the problematic or impossible capability of understanding and explaining the outputs. The interpretability of the models is paramount to increasing confidence and acceptance for medical specialists and, consequently, the full exploitation of these advances in the medical field, effectively impacting better diagnosis day by day. Thus, the improvement and integration of interpretability methods focused on the context of medical applications is a field of a clear upward trend. This thesis project will investigate the exploration and improvement of image-based diagnostic methods with interpretable capability, hypothesizing that interpretable models offer more confidence in their use and allow better analysis, auditing, and inspection by medical experts and other important stakeholders. Still, integrated development is proposed together with field experts along the process, aiming to finetune and precise alight of interpretability methods with standard requirements of the professionals.