Normalization of H&E stain by autoencoders with analysis of ensemble learning for histological images

• Grant number: 24/00936-0
• Support Opportunities: Scholarships in Brazil - Scientific Journalism
• Start date until: June 01, 2024
• End date until: November 30, 2024
• Field of knowledge: Engineering - Biomedical Engineering
• Principal Investigator: Thaína Aparecida Azevedo Tosta
• Grantee: Leon Nilson Gomes de Lima
• Host Institution: Instituto de Ciência e Tecnologia (ICT). Universidade Federal de São Paulo (UNIFESP). Campus São José dos Campos. São José dos Campos , SP, Brazil
• Associated research grant: 22/03020-1 - Normalization of H&E stain by autoencoders with analysis of ensemble learning for histological images, AP.PNGP.PI

Abstract

In areas related to histological research, the staining technique of tissues with suspicion of cancer in biopsies is important for confirmation or not of the respective investigated pathology. Staining methods, such as using H&E (Hematoxylin-Eosin) are commonly applied in such laboratory practice. However, there are possibilities of equivocated diagnosis due to differences in sample staining. The project associated with this proposal aims to develop robust computational techniques for normalizing images of interest, besides better understanding the biological interface of the used pigment with the biological tissues in analysis. This proposal presents the diffusion of this work recognizing its relevance by (i) unawareness of the general public about the characteristics of cancers, (ii) how this neoplasia affects the population, (iii) its diagnoses process, (iv) how these diagnoses are obtained and interpreted by pathologists, (v) how computational systems can aid in this process, making it more objective and faster, and (vi) aspects until now known but little comprehended related to machine learning, artificial intelligence and image processing, as it is highlighted by the correlated scientific literature. Furthermore, the wide dissemination of the topics previously summarized can highlight the interdisciplinarity of scientific research developed in Brazilian public universities, with explicit correlation between biomedicine and computer science/engineering.