Scholarlink

The workload of pathologists is increasing due to the continuous rise in cancer cases. To classify tumors and assess their level of aggressiveness, pathologists must analyze a large number of pathological images, sometimes hundreds or thousands, which is both costly and time-consuming and does not necessarily guarantee perfectly accurate results. To address these challenges, automating the analysis of tissue slices mounted on glass slides using microscopes is essential. Computer-assisted techniques, particularly artificial intelligence, offer significant potential to improve tumor classification. We propose to develop a new multi-input convolutional neural network architecture, leveraging both MRI and histological data to refine glioma classification. This approach will be validated using data from the Radiology-Pathology Challenge (CPM: RAD-PATH 2020). Our results show a precision of 0.75 for MRI image classification, 0.80 for pathological image classification, and 0.83 when combining pathological and MRI images.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Enhanced Glioma Classification Through Multi-modal Deep Learning: Integrating Histopathological and MRI Data

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Linda Ait Mohammed,
  • Fatiha Alim-Ferhat,
  • Farid Talbi

摘要

The workload of pathologists is increasing due to the continuous rise in cancer cases. To classify tumors and assess their level of aggressiveness, pathologists must analyze a large number of pathological images, sometimes hundreds or thousands, which is both costly and time-consuming and does not necessarily guarantee perfectly accurate results. To address these challenges, automating the analysis of tissue slices mounted on glass slides using microscopes is essential. Computer-assisted techniques, particularly artificial intelligence, offer significant potential to improve tumor classification. We propose to develop a new multi-input convolutional neural network architecture, leveraging both MRI and histological data to refine glioma classification. This approach will be validated using data from the Radiology-Pathology Challenge (CPM: RAD-PATH 2020). Our results show a precision of 0.75 for MRI image classification, 0.80 for pathological image classification, and 0.83 when combining pathological and MRI images.