OncoData: artificial intelligence and machine learning for cancer pathology

• Grant number: 23/11600-0
• Support Opportunities: Research Grants - Innovative Research in Small Business - PIPE
• Start date: February 01, 2024
• End date: January 31, 2026
• Field of knowledge: Health Sciences - Medicine
• Principal Investigator: Viviane Teixeira Loiola de Alencar
• Grantee: Viviane Teixeira Loiola de Alencar
• Company: Oncodata Soluções Tecnológicas Ltda
• CNAE: Atividades de serviços de complementação diagnóstica e terapêutica
• City: São José dos Campos
• Pesquisadores principais: Felipe Navarro Balbino Alves
• Associated researchers: Fabio Rocha Fernandes Tavora ; Vladmir Cláudio Cordeiro de Lima
• Associated research grant: 21/14596-9 - OncoData: artificial intelligence and machine learning for cancer pathology, AP.PIPE
• Associated scholarship(s): 24/02537-6 - Machine learning for prostate cancer diagnosis in whole-slide pathology images, BP.TT; 24/00132-9 - Machine learning for histopathological samples cancer signaling and development of digital pathology and medical education web interfaces., BP.TT; 24/02348-9 - Artificial intelligence for prostate cancer Gleason score classification, BP.TT; 24/00201-0 - OncoData: Artificial intelligence and machine learning for Cancer pathology, BP.PIPE

Abstract

Pathology is essential for the diagnosis and treatment of cancer. It consists of an assessment made by a pathologist, based on previously described standards. Its accuracy is directly related, among other factors, to the evaluator's experience. Therefore, discrepancies between reports are more common than we would expect and can reach up to 25% when the material is reviewed by an experienced pathologist. These changes can make the difference between a patient receiving or not receiving the correct cancer diagnosis and treatment, with significant social and financial impacts. Additionally, the number of pathologists has been decreasing over the years, resulting in a shortage of this crucial professional. Consequently, delays in diagnosis, which can take up to 3 or more weeks, especially in hard-to-reach areas, are common. Given the number of biopsies performed each year, this becomes an urgent issue. To address this problem, in phase 1 of the PIPE project, we developed a proof-of-concept for an artificial intelligence (AI) algorithm capable of accurately classifying the most common histological subtypes of lung cancer. We also developed a telepathology platform that allows for viewing images through a digital microscope and the combined use of AI tools. For PIPE-2, we aim to improve the AI tool to cover other histological subtypes of lung cancer and validate it in real-life cases of Brazilian patients, especially biopsy samples, which pose a diagnostic challenge due to the scarcity of tumor material. We also aim to develop a new AI algorithm capable of classifying prostate tumors and the Gleason score (GS). This type of cancer is the most common in Latin America, and the GS is essential for therapeutic and prognostic definition, with studies showing considerable inter and intra-pathologist variability. Additionally, we will evolve the platform already developed in the first stage to make it more functional for pathology laboratories, with the ability to integrate with LIS systems. We will also create a digital pathology education platform, focusing on colleges and medical residencies, using what has already been developed for telepathology. All tools proposed for the second phase of the project are complementary to solving the proposed problem and have a high potential impact on the public and private health of the country. (AU)