Modelo não supervisionado de segmentação de estruturas em exames de tomografia de crânio

Abstract

This work proposes the development of an unsupervised method for segmenting cranial CT images. The methodology involves extracting image features and applying similarity and continuity constraints to create segmentation maps of intracranial structures and observable tissues. This approach aims to assist specialists in diagnosis by identifying regions with specific anomalies. Applied to real-world datasets, the method uses a spatial continuity evaluation function related to the desired number of structures. Results show satisfactory performance, indicating a simplified and accessible approach that reduces computational load, training time, and financial costs. This proposal serves as a practical tool for cranial CT image segmentation, providing significant contributions to the analysis of medical images in clinical and diagnostic settings.