Scholarlink

Purpose <p>To evaluate the feasibility of machine learning (ML) models using preprocedural MRI-based Whole-Organ Magnetic Resonance Imaging Score (WORMS) and clinical parameters to predict treatment response after genicular artery embolization in patients with knee osteoarthritis.</p> Materials and Methods <p>This retrospective study included 66 patients (72 knees) who underwent GAE between December 2022 and June 2024. Preprocedural assessments included WORMS and Kellgren-Lawrence grading. Clinical response was defined as a ≥ 50% reduction in Visual Analog Scale (VAS) score. Feature selection was performed using recursive feature elimination and correlation analysis. Multiple ML algorithms (Random Forest, Support Vector Machine, Logistic Regression) were trained using stratified fivefold cross-validation. Conventional statistical analyses assessed group differences and correlations.</p> Results <p>Of 72 knees, 33 (45.8%) achieved a clinically significant response. Responders showed significantly lower WORMSs for cartilage, bone marrow, and total joint damage (<i>p</i> &lt; 0.05). The Random Forest model demonstrated the best performance, with an accuracy of 81.8%, AUC-ROC of 86.2%, sensitivity of 90%, and specificity of 75%. Key predictive features included total WORMS, ligament score, and baseline VAS. Bone marrow score showed the strongest correlation with VAS reduction (<i>r</i> = –0.430, <i>p</i> &lt; 0.001).</p> Conclusion <p>ML models integrating WORMS and clinical data suggest that greater cartilage loss, bone marrow edema, joint damage, and higher baseline VAS scores may help to identify patients less likely to respond to GAE for knee OA.</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

Machine Learning and MRI-Based Whole-Organ Magnetic Resonance Imaging Score (WORMS): A Novel Approach to Enhancing Genicular Artery Embolization Outcomes in Knee Osteoarthritis

  • 期刊论文

  • 发表日期：

    2025年8月4日

查看全文
  • Ali Dablan,
  • Hamit Özgül,
  • Mustafa Fatih Arslan,
  • Oğuzhan Türksayar,
  • Mehmet Cingöz,
  • Ilhan Nahit Mutlu,
  • Cagri Erdim,
  • Tevfik Guzelbey,
  • Ozgur Kılıckesmez

摘要

Purpose

To evaluate the feasibility of machine learning (ML) models using preprocedural MRI-based Whole-Organ Magnetic Resonance Imaging Score (WORMS) and clinical parameters to predict treatment response after genicular artery embolization in patients with knee osteoarthritis.

Materials and Methods

This retrospective study included 66 patients (72 knees) who underwent GAE between December 2022 and June 2024. Preprocedural assessments included WORMS and Kellgren-Lawrence grading. Clinical response was defined as a ≥ 50% reduction in Visual Analog Scale (VAS) score. Feature selection was performed using recursive feature elimination and correlation analysis. Multiple ML algorithms (Random Forest, Support Vector Machine, Logistic Regression) were trained using stratified fivefold cross-validation. Conventional statistical analyses assessed group differences and correlations.

Results

Of 72 knees, 33 (45.8%) achieved a clinically significant response. Responders showed significantly lower WORMSs for cartilage, bone marrow, and total joint damage (p < 0.05). The Random Forest model demonstrated the best performance, with an accuracy of 81.8%, AUC-ROC of 86.2%, sensitivity of 90%, and specificity of 75%. Key predictive features included total WORMS, ligament score, and baseline VAS. Bone marrow score showed the strongest correlation with VAS reduction (r = –0.430, p < 0.001).

Conclusion

ML models integrating WORMS and clinical data suggest that greater cartilage loss, bone marrow edema, joint damage, and higher baseline VAS scores may help to identify patients less likely to respond to GAE for knee OA.

Graphical Abstract