Scholarlink

<p>Migraine is a complex neurological disorder with diverse clinical phenotypes and a multifaceted pathophysiology, which poses substantial challenges for accurate diagnosis, subtype differentiation, and biomarker discovery. Machine learning (ML) techniques have emerged as promising tools for classifying migraine patients and uncovering the underlying neurobiological mechanisms that differentiate migraine types and subtypes. This systematic review identifies current ML classification models for migraine types and subtypes, evaluating the quality, reproducibility, and clinical utility of published studies. The findings demonstrate that current ML models, particularly support vector machines and linear discriminant analysis, can accurately classify migraine patients based on structural and functional neuroimaging features with accuracies ranging from 75 to 98%. However, quality assessment revealed significant methodological heterogeneity across studies, including inconsistent reporting of model performance, insufficient patient phenotyping, small and imbalanced datasets, and limited external validation. These limitations hinder the global generalizability and reproducibility of these studies. We propose a roadmap for future research emphasizing well-characterized clinical subgrouping, standardized data acquisition and feature engineering protocols, transparency in model development and reporting, and collaborative multicentric designs to enable large-scale validation. Furthermore, this review stresses the importance of incorporating real-world phenotypic data, such as treatment response, comorbidities, and digital phenotyping metrics, to enrich ML models and support the transition toward precision medicine in migraine care. Ultimately, this review highlights the urgent need for methodological rigor in migraine ML classification studies to bridge the gap between experimental success and clinical applicability.</p> Graphical Abstract <p></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

Application of machine learning in migraine classification: a call for study design standardization and global collaboration

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年10月2日

查看全文
  • Igor Petrušić,
  • Roberta Messina,
  • Lanfranco Pellesi,
  • David Garcia Azorin,
  • Chia-Chun Chiang,
  • Adriana Della Pietra,
  • Woo-Seok Ha,
  • Alejandro Labastida-Ramirez,
  • Dilara Onan,
  • Raffaele Ornello,
  • Bianca Raffaelli,
  • Eloisa Rubio-Beltran,
  • Ruth Ruscheweyh,
  • Claudio Tana,
  • Doga Vuralli,
  • Marta Waliszewska-Prosół,
  • Wei Wang,
  • William David Wells-Gatnik,
  • Paolo Martelletti,
  • Alberto Raggi

摘要

Migraine is a complex neurological disorder with diverse clinical phenotypes and a multifaceted pathophysiology, which poses substantial challenges for accurate diagnosis, subtype differentiation, and biomarker discovery. Machine learning (ML) techniques have emerged as promising tools for classifying migraine patients and uncovering the underlying neurobiological mechanisms that differentiate migraine types and subtypes. This systematic review identifies current ML classification models for migraine types and subtypes, evaluating the quality, reproducibility, and clinical utility of published studies. The findings demonstrate that current ML models, particularly support vector machines and linear discriminant analysis, can accurately classify migraine patients based on structural and functional neuroimaging features with accuracies ranging from 75 to 98%. However, quality assessment revealed significant methodological heterogeneity across studies, including inconsistent reporting of model performance, insufficient patient phenotyping, small and imbalanced datasets, and limited external validation. These limitations hinder the global generalizability and reproducibility of these studies. We propose a roadmap for future research emphasizing well-characterized clinical subgrouping, standardized data acquisition and feature engineering protocols, transparency in model development and reporting, and collaborative multicentric designs to enable large-scale validation. Furthermore, this review stresses the importance of incorporating real-world phenotypic data, such as treatment response, comorbidities, and digital phenotyping metrics, to enrich ML models and support the transition toward precision medicine in migraine care. Ultimately, this review highlights the urgent need for methodological rigor in migraine ML classification studies to bridge the gap between experimental success and clinical applicability.

Graphical Abstract