Ambispective clinical study to develop artificial intelligence models using routine blood biomarkers to support the early identification of breast cancer.

• Grant number: 23/14898-0
• Support Opportunities: Research Grants - Innovative Research in Small Business - PIPE
• Start date: March 01, 2024
• End date: February 28, 2026
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Marco Aurelio Kohara
• Grantee: Marco Aurelio Kohara
• Company: Huna Ltda
• CNAE: Pesquisa e desenvolvimento experimental em ciências físicas e naturais ; Atividades de serviços de complementação diagnóstica e terapêutica
• City: São Paulo
• Pesquisadores principais: Daniella Castro Araujo
• Associated research grant: 22/07614-3 - Routine blood markers panel based on artificial intelligence for breast cancer early detection, AP.PIPE
• Associated scholarship(s): 24/04558-0 - Ambispective clinical study for the development of artificial intelligence models using routine blood biomarkers for early detection of Breast Cancer - Oncology, Clinical Analysis & Research Ethics, BP.TT; 24/04562-8 - Ambispective clinical study for the development of artificial intelligence models using routine blood biomarkers for early detection of Breast Cancer - Proof of concepts and support for commercialization., BP.TT; 24/04700-1 - Ambispective clinical study for the development of artificial intelligence models using routine blood biomarkers to support early identification of Breast Cancer, BP.PIPE

Abstract

Machine Learning (ML) has emerged as an essential tool for identifying complex patterns in blood tests, enabling the transformation of non-specific tests into accurate tools for diagnosing multiple diseases. In this context, we propose the integration of several routine blood markers into a robust panel aided by Artificial Intelligence (AI) to assess the risk of breast cancer in the age group of 40 to 70 years. This research is a pioneering cross-sectional study using ML methods to create, evaluate, and validate a model that explores the non-linear relationships between routine blood biomarkers to improve the breast cancer screening process.The central objective of this project lies in developing a reliable AI model capable of identifying breast cancer from routine blood tests. Successful implementation of this model could result in an accessible and scalable screening tool targeting early clinical interventions for women at high risk. We believe this approach can potentially optimize breast cancer screening in resource-limited regions, contributing to early detection, effective treatment, and improved clinical outcomes.Despite initial limitations, such as the lack of external validation and incomplete information on comorbidities, research carried out during Phase 1/PIPE revealed promising results in personalizing risk stratification. Therefore, we assume that the prospect of significantly improving the effectiveness and efficiency of breast cancer screening in populations with limited resources is potentially relevant. In this sense, we propose additional investigations in Phase 2/direct PIPE, including the incorporation of new blood biomarkers to mitigate possible biases and confirm the clinical utility of the proposed model, and expand the case series, repeating the methodology established in Phase 1 to externally validate the model created to, in parallel, carry out Proofs of Concepts with commercial partners. Such investigations will be conducted in collaboration with Hospital de Amor de Barretos, Grupo Fleury, Instituto Hermes Pardini, cooperatives of the Unimed system, and Grupo Hapvida Notre-Dame Intermédica (HNDI) aiming to strengthen the scientific knowledge base to significantly contribute to the area of detection early stage of breast cancer. (AU)