Artificial intelligence in the diagnosis of low-grade and high-grade head and neck lymphomas.

• Grant number: 23/11058-1
• Support Opportunities: Scholarships abroad - Research Internship - Doctorate
• Start date: March 01, 2025
• End date: February 28, 2026
• Field of knowledge: Health Sciences - Dentistry
• Principal Investigator: Pablo Agustin Vargas
• Grantee: Lucas Lacerda de Souza
• Supervisor: Syed Ali Khurram
• Host Institution: Faculdade de Odontologia de Piracicaba (FOP). Universidade Estadual de Campinas (UNICAMP). Piracicaba , SP, Brazil
• Institution abroad: University of Sheffield, England
• Associated to the scholarship: 22/03123-5 - ARTIFICIAL INTELLIGENCE IN THE DIAGNOSIS OF SMALL, ROUND AND BLUE CELL NEOPLASMS, BP.DR

Abstract

Lymphomas are hematopoietic neoplasms that present a diagnostic challenge for pathologists due to their distinct morphological, immunophenotypic, and molecular features. With the advancement of computational technologies, the utilization of artificial intelligence techniques through digital systems and the development of algorithms for image analysis have been studied to enhance diagnostic efficiency. This study aims to evaluate various pre-trained models of neural networks using a Deep Learning approach to differentiate low-grade and high-grade lymphomas and diagnose lesions with similar microscopic characteristics. Retrospective samples diagnosed as lymphomas will becollected, and the slides scanned and manually annotated to validate the images by delineating areas of lesions and connective tissue. The images will be fragmented into patches, resized to a standardized dimension of 299x299, and subsequently processed by a neural network. A random division of the images will be conducted using a network of codes based on the desired test to be performed on the neural network. Pre-trained models such as UNet, Efficient UNet and Xception, generating a classification of the distinguishing characteristics among the analyzed groups. The ultimate goal of these analyses is to develop an algorithm that supports accurate histopathological diagnosis or assists pathologists in diagnosing low-grade and high-grade head and neck lymphomas, ultimately improving patient prognosis.