Sistema de recuperación de imágenes histopatológicas basado en contenido visual

Abstract

This work present the develop a histopathological image retrieval system based on its visual content, which serves as an essential tool to improve search accessibility, interpretation of histopathological images, which are guided by the principle of evidence-based medicine. The construction and operation process of this system consists of four fundamental stages. Initially, a repository containing whole slide images (WSI - Whole slide Images) is built. Once the images are digitized, a methodology is established that enhances the visual content of the images by means of appropriate processing techniques for the subsequent extraction of low-level visual features. Next, a machine learning model designed to learn and extract the relevant visual features is implemented. This model plays a crucial role in the search phase of the CBIR system based on low-level feature comparison, ensuring more accurate and efficient results. Finally, the performance of the system in operation is evaluated. This analysis aims not only to improve the practice of histopathology, but also to contribute to the advancement of knowledge in evidence-based medicine. The research aims to make a significant contribution to the field of virtual microscopy