Scholarlink

Purpose <p>Metachromatic leukodystrophy (MLD) is a&#xa0;rare lysosomal storage disorder characterized by progressive white matter demyelination. Quantification of demyelinated white matter on MRI—typically expressed as the <i>demyelination load</i>—serves as a&#xa0;key imaging biomarker of disease burden, enabling objective monitoring beyond visual rating scales. However, current semi-automated pipelines are limited by manual interaction, pediatric brain variability, and differences in MRI acquisition. This study aimed to develop and validate a&#xa0;self-configuring convolutional neural network (CNN) for automated segmentation of demyelinated white matter in MLD and to compare its performance with a&#xa0;conventional semi-automated method across heterogeneous MRI datasets.</p> Methods <p>An nnU-Net was trained on 189 3D T1- and axial T2-weighted scans from 35&#xa0;MLD patients using visually controlled conventional masks as ground truth. Independent testing was performed on 130 scans (73&#xa0;high-resolution 3D, 57&#xa0;lower-resolution 2D T1-weighted) from 49&#xa0;patients. Performance was assessed by Dice coefficient, Bland-Altman bias, correlation with Gross Motor Function Classification (GMFC-MLD), MLD MRI severity score, longitudinal consistency, and qualitative review of outliers.</p> Results <p>CNN-based segmentation showed strong spatial agreement with the reference method, with a&#xa0;median Dice coefficient of 0.82 for 3D T1-weighted scans and 0.75 for 2D scans. Volumetric bias was minimal on Bland-Altman analysis. CNN-derived demyelination load correlated significantly with motor impairment (r<sub>S</sub> = 0.38 for 3D and r = 0.56 for 2D; both <i>p</i> &lt; 0.001) and showed a&#xa0;stronger association with the MLD MRI severity score than conventional segmentation (3D: r<sub>S</sub> = 0.48 vs. 0.28; 2D: r<sub>S</sub> = 0.83 vs. 0.29). Correlations with clinical status were slightly lower (CNN: r<sub>S</sub> = 0.38, <i>p</i> &lt; 0.001; conventional: (r<sub>S</sub> = 0.26, <i>p</i> &lt; 0.025)) Longitudinal analyses demonstrated stable, monotonic changes over time, and qualitative review revealed fewer boundary misclassifications.</p> Conclusion <p>The nnU-Net enables fast, reproducible, and clinically meaningful segmentation of demyelinated white matter in MLD. It generalizes across MRI protocols, correlates with motor function, and offers a&#xa0;scalable tool for standardized biomarker extraction in clinical trials and other leukodystrophies.</p>

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

Automated Deep Learning-Based Demyelination Load Segmentation in Metachromatic Leukodystrophy

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年4月21日

查看全文
  • Pascal Martin,
  • Joël Schaerer,
  • Thomas Cajgfinger,
  • Alessandro Delmonte,
  • Benjamin Bender,
  • David Whiteman,
  • C.J. Malanga,
  • Alexandre Fusellier,
  • David Scott,
  • Joyce Suhy,
  • Hendrik Rosewich,
  • Samuel Groeschel,
  • Luc Bracoud

摘要

Purpose

Metachromatic leukodystrophy (MLD) is a rare lysosomal storage disorder characterized by progressive white matter demyelination. Quantification of demyelinated white matter on MRI—typically expressed as the demyelination load—serves as a key imaging biomarker of disease burden, enabling objective monitoring beyond visual rating scales. However, current semi-automated pipelines are limited by manual interaction, pediatric brain variability, and differences in MRI acquisition. This study aimed to develop and validate a self-configuring convolutional neural network (CNN) for automated segmentation of demyelinated white matter in MLD and to compare its performance with a conventional semi-automated method across heterogeneous MRI datasets.

Methods

An nnU-Net was trained on 189 3D T1- and axial T2-weighted scans from 35 MLD patients using visually controlled conventional masks as ground truth. Independent testing was performed on 130 scans (73 high-resolution 3D, 57 lower-resolution 2D T1-weighted) from 49 patients. Performance was assessed by Dice coefficient, Bland-Altman bias, correlation with Gross Motor Function Classification (GMFC-MLD), MLD MRI severity score, longitudinal consistency, and qualitative review of outliers.

Results

CNN-based segmentation showed strong spatial agreement with the reference method, with a median Dice coefficient of 0.82 for 3D T1-weighted scans and 0.75 for 2D scans. Volumetric bias was minimal on Bland-Altman analysis. CNN-derived demyelination load correlated significantly with motor impairment (rS = 0.38 for 3D and r = 0.56 for 2D; both p < 0.001) and showed a stronger association with the MLD MRI severity score than conventional segmentation (3D: rS = 0.48 vs. 0.28; 2D: rS = 0.83 vs. 0.29). Correlations with clinical status were slightly lower (CNN: rS = 0.38, p < 0.001; conventional: (rS = 0.26, p < 0.025)) Longitudinal analyses demonstrated stable, monotonic changes over time, and qualitative review revealed fewer boundary misclassifications.

Conclusion

The nnU-Net enables fast, reproducible, and clinically meaningful segmentation of demyelinated white matter in MLD. It generalizes across MRI protocols, correlates with motor function, and offers a scalable tool for standardized biomarker extraction in clinical trials and other leukodystrophies.