Scholarlink

Meningiomas are the most prevalent type of primary intracranial tumor, with radiotherapy commonly used for treating high-grade cases. Precise identification and delineation of tumor boundaries on magnetic resonance imaging (MRI) are essential for effective radiotherapy planning. However, this task is both time-consuming and labor-intensive, highlighting the need for automated segmentation solutions in clinical practice. To support progress in this area, the BraTS-MEN-RT challenge was launched in 2024, and continued in 2025, aiming to benchmark automated meningioma segmentation on MRI. In this paper, we describe our submission to the 2025 edition of the challenge, which utilizes the widely adopted nnU-Net framework. Our approach specifically focuses on improving the pre-processing steps and the delineation of tumor borders. It ranked first during the testing stage of the challenge, achieving the average Dice score of 0.816.

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

Boundary-Aware Approach for Meningioma Segmentation in Radiotherapy Planning MRI

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Valeriia Abramova,
  • Agustin Cartaya Lathulerie,
  • Uma M. Lal-Trehan Estrada,
  • Cansu Yalçın,
  • Rachika E. Hamadache,
  • Clara Lisazo,
  • Micaela Rivas Díaz,
  • Adrià Casamitjana,
  • Arnau Oliver,
  • Xavier Lladó

摘要

Meningiomas are the most prevalent type of primary intracranial tumor, with radiotherapy commonly used for treating high-grade cases. Precise identification and delineation of tumor boundaries on magnetic resonance imaging (MRI) are essential for effective radiotherapy planning. However, this task is both time-consuming and labor-intensive, highlighting the need for automated segmentation solutions in clinical practice. To support progress in this area, the BraTS-MEN-RT challenge was launched in 2024, and continued in 2025, aiming to benchmark automated meningioma segmentation on MRI. In this paper, we describe our submission to the 2025 edition of the challenge, which utilizes the widely adopted nnU-Net framework. Our approach specifically focuses on improving the pre-processing steps and the delineation of tumor borders. It ranked first during the testing stage of the challenge, achieving the average Dice score of 0.816.