Hierarchically Optimized Multiple Instance Learning With Multi-Magnification Pathological Images for Cerebral Tumor Diagnosis.

in IEEE journal of biomedical and health informatics by Lianghui Zhu, Renao Yan, Tian Guan, Fenfen Zhang, Linlang Guo, Qiming He, Shanshan Shi, Huijuan Shi, Yonghong He, Anjia Han

TLDR

  • A novel Hierarchically Optimized Multiple Instance Learning (HOMIL) method is proposed for accurate diagnosis of cerebral tumors, achieving state-of-the-art performance in brain tumor classification, glioma grading, and origin determination.

Abstract

Accurate diagnosis of cerebral tumors is crucial for effective clinical therapeutics and prognosis. However, limitations in brain biopsy tissues and the scarcity of pathologists specializing in cerebral tumors hinder comprehensive clinical tests for precise diagnosis. To address these challenges, we first established a brain tumor dataset of 3,520 cases collected from multiple centers. We then proposed a novel Hierarchically Optimized Multiple Instance Learning (HOMIL) method for classifying six common brain tumor types, glioma grading, and predicting the origin of brain metastatic cancers. The feature encoder and aggregator in HOMIL were trained alternately based on specific datasets and tasks. Compared to other multiple instance learning (MIL) methods, HOMIL achieved state-of-the-art performance with impressive accuracies: 93.29% / 85.60% for brain tumor classification, 91.21% / 96.93% for glioma grading, and 86.36% / 79.28% for origin determination on internal/external datasets. Additionally, HOMIL effectively located multi-scale regions of interest, enabling an in-depth analysis through features and heatmaps. Extensive visualization demonstrated HOMIL's ability to cluster features within the same type while establishing distinct boundaries between tumor types. It also identified critical areas on pathological slides, regardless of tumor size.

Overview

  • The study aims to develop an accurate diagnostic tool for cerebral tumors, addressing limitations in brain biopsy tissues and the scarcity of specialized pathologists.
  • A brain tumor dataset of 3,520 cases was established, and a novel Hierarchically Optimized Multiple Instance Learning (HOMIL) method was proposed for classifying brain tumor types, glioma grading, and predicting brain metastatic cancer origin.
  • The primary objective is to achieve accurate diagnosis and precise classification of brain tumors, enabling effective clinical therapeutics and prognosis.

Comparative Analysis & Findings

  • HOMIL achieved state-of-the-art performance with impressive accuracies of 93.29% / 85.60% for brain tumor classification, 91.21% / 96.93% for glioma grading, and 86.36% / 79.28% for origin determination on internal/external datasets.
  • HOMIL effectively located multi-scale regions of interest, allowing for in-depth analysis through features and heatmaps.
  • Compared to other multiple instance learning (MIL) methods, HOMIL outperformed them in brain tumor classification, glioma grading, and origin determination tasks.

Implications and Future Directions

  • The study's findings have significant implications for the accurate diagnosis of cerebral tumors, enabling effective clinical therapeutics and prognosis.
  • Future directions could involve fine-tuning HOMIL for specific tumor types, developing more advanced feature extractors, and integrating HOMIL with other diagnostic tools.
  • The study highlights the need for further research to address the limitations and challenges in brain tumor diagnosis, including the scarcity of specialized pathologists and the complexity of brain tumor biology.
Read Full Article