Scholarlink

High-quality annotation plays a crucial role in improving deep learning-based detection of brain metastases (BMs). In this work, originally published in European Radiology Experimental [1], we investigate how annotations derived from SPACE sequences, which offer superior lesion conspicuity compared to MPRAGE, affect detection performance. We compare models trained with normal annotation quality on MPRAGE against those trained with high-quality annotation (HAQ) derived from co-registered SPACE images. Our results show that HAQ significantly enhances detection and delineation performance across multiple datasets, even when applied to MPRAGE inputs. These findings demonstrate that improved annotation quality alone can substantially boost deep learning detection of small BMs, enabling faster and more accurate fully automated diagnosis.

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

Abstract: Enhancing Deep Learning Methods for Brain Metastasis Detection Through Cross-technique Annotations on SPACE MRI

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Tassilo Wald,
  • Benjamin Hamm,
  • Julius Holzschuh,
  • Rami El Shafie,
  • Andreas Kudak,
  • Balint Kovacs,
  • Irada Pflüger,
  • Bastian von Nettelbladt,
  • Constantin Ulrich,
  • Michael A. Baumgartner,
  • Philipp Vollmuth,
  • Jürgen Debus,
  • Klaus H. Maier-Hein,
  • Thomas Welzel

摘要

High-quality annotation plays a crucial role in improving deep learning-based detection of brain metastases (BMs). In this work, originally published in European Radiology Experimental [1], we investigate how annotations derived from SPACE sequences, which offer superior lesion conspicuity compared to MPRAGE, affect detection performance. We compare models trained with normal annotation quality on MPRAGE against those trained with high-quality annotation (HAQ) derived from co-registered SPACE images. Our results show that HAQ significantly enhances detection and delineation performance across multiple datasets, even when applied to MPRAGE inputs. These findings demonstrate that improved annotation quality alone can substantially boost deep learning detection of small BMs, enabling faster and more accurate fully automated diagnosis.