This study explores the integration of a population-level context into pathology detection models using a graph-theoretic approach. By introducing the PopuSense module, the research aims to address the challenge of distinguishing between healthy and pathological distributions in medical images. Experimental results show promising improvements in contrast-based images but highlight challenges with texture-based inputs.
The study focuses on enhancing separability in pathology detection models by incorporating a population-level context through a refined latent code. By leveraging hypergraphs and graph convolutional networks, the research demonstrates potential advancements in anomaly detection within medical imaging datasets. The proposed method offers an alternative avenue for refining representations learned by convolutional autoencoders, particularly in contrast-based scenarios.
לשפה אחרת
מתוכן המקור
arxiv.org
תובנות מפתח מזוקקות מ:
by P. Bilha Git... ב- arxiv.org 03-05-2024
https://arxiv.org/pdf/2403.02307.pdfשאלות מעמיקות