Scholarlink

Background <p>Parkinson's disease (PD) and multiple system atrophy with parkinsonian type (MSA-P) share various motor and nonmotor symptoms, complicating early differential diagnosis, although prognosis and levodopa response differ.</p> Objectives <p>To develop a machine learning model using balance analysis to differentiate early-stage PD and MSA-P.</p> Methods <p>We enrolled 22 healthy controls (HC), 20 PD, and 17 MSA-P patients within three years of onset. Participants stood for 60&#xa0;s on dual force-plates under eyes open (EO) and eyes closed (EC). Seven center of pressure (COP) parameters per condition and EC–EO differences were analyzed. Group differences were assessed with ANOVA. Feature selection was performed using LightGBM (LGBM), followed by model training with fourfold cross-validation.</p> Results <p>Mean ages were 63.8 (HC), 64.2 (PD), and 68.8&#xa0;years (MSA-P). Disease durations were 2.4&#xa0;years in PD and 1.8&#xa0;years in MSA-P. Except for mean distance in the anteroposterior direction with EO and 95% confidence ellipse area with EC, all parameters showed significant group differences. Post-hoc analysis revealed significant differences between HC and PD/MSA-P in EO, whereas in EC, differences were more pronounced between MSA-P and PD/HC. The LGBM model achieved 81.4% accuracy in distinguishing PD from MSA-P, 94.9% in distinguishing healthy controls from patient groups, and 84.9% for three-group classification.</p> Conclusion <p>Balance biomarker, which analyzes the balance parameters with machine learning model could differentiate early-stage PD and MSA-P with high accuracy.</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

Balance biomarker for early differentiation of Parkinson’s disease and multiple system atrophy with parkinsonian type

  • 期刊论文

  • 发表日期：

    2026年3月6日

查看全文
  • Hee Jin Chang,
  • Jong Ho Kim,
  • Han-Wook Song,
  • Sanghyun Lee,
  • Eunjin Kwon,
  • Seong-Hae Jeong,
  • Eungseok Oh

摘要

Background

Parkinson's disease (PD) and multiple system atrophy with parkinsonian type (MSA-P) share various motor and nonmotor symptoms, complicating early differential diagnosis, although prognosis and levodopa response differ.

Objectives

To develop a machine learning model using balance analysis to differentiate early-stage PD and MSA-P.

Methods

We enrolled 22 healthy controls (HC), 20 PD, and 17 MSA-P patients within three years of onset. Participants stood for 60 s on dual force-plates under eyes open (EO) and eyes closed (EC). Seven center of pressure (COP) parameters per condition and EC–EO differences were analyzed. Group differences were assessed with ANOVA. Feature selection was performed using LightGBM (LGBM), followed by model training with fourfold cross-validation.

Results

Mean ages were 63.8 (HC), 64.2 (PD), and 68.8 years (MSA-P). Disease durations were 2.4 years in PD and 1.8 years in MSA-P. Except for mean distance in the anteroposterior direction with EO and 95% confidence ellipse area with EC, all parameters showed significant group differences. Post-hoc analysis revealed significant differences between HC and PD/MSA-P in EO, whereas in EC, differences were more pronounced between MSA-P and PD/HC. The LGBM model achieved 81.4% accuracy in distinguishing PD from MSA-P, 94.9% in distinguishing healthy controls from patient groups, and 84.9% for three-group classification.

Conclusion

Balance biomarker, which analyzes the balance parameters with machine learning model could differentiate early-stage PD and MSA-P with high accuracy.