Machine learning analysis of motor evoked potential time series to predict disability progression in multiple sclerosis.
- 作者列表："Yperman J","Becker T","Valkenborg D","Popescu V","Hellings N","Wijmeersch BV","Peeters LM
BACKGROUND:Evoked potentials (EPs) are a measure of the conductivity of the central nervous system. They are used to monitor disease progression of multiple sclerosis patients. Previous studies only extracted a few variables from the EPs, which are often further condensed into a single variable: the EP score. We perform a machine learning analysis of motor EP that uses the whole time series, instead of a few variables, to predict disability progression after two years. Obtaining realistic performance estimates of this task has been difficult because of small data set sizes. We recently extracted a dataset of EPs from the Rehabiliation & MS Center in Overpelt, Belgium. Our data set is large enough to obtain, for the first time, a performance estimate on an independent test set containing different patients. METHODS:We extracted a large number of time series features from the motor EPs with the highly comparative time series analysis software package. Mutual information with the target and the Boruta method are used to find features which contain information not included in the features studied in the literature. We use random forests (RF) and logistic regression (LR) classifiers to predict disability progression after two years. Statistical significance of the performance increase when adding extra features is checked. RESULTS:Including extra time series features in motor EPs leads to a statistically significant improvement compared to using only the known features, although the effect is limited in magnitude (ΔAUC = 0.02 for RF and ΔAUC = 0.05 for LR). RF with extra time series features obtains the best performance (AUC = 0.75±0.07 (mean and standard deviation)), which is good considering the limited number of biomarkers in the model. RF (a nonlinear classifier) outperforms LR (a linear classifier). CONCLUSIONS:Using machine learning methods on EPs shows promising predictive performance. Using additional EP time series features beyond those already in use leads to a modest increase in performance. Larger datasets, preferably multi-center, are needed for further research. Given a large enough dataset, these models may be used to support clinicians in their decision making process regarding future treatment.
背景: 诱发电位 (EPs) 是中枢神经系统电导率的量度。它们用于监测多发性硬化症患者的疾病进展。以前的研究仅从EPs中提取了几个变量，这些变量通常进一步浓缩为单个变量: EP评分。我们对运动EP进行机器学习分析，使用整个时间序列而不是几个变量来预测两年后的残疾进展。由于小的数据集大小，获得该任务的实际性能估计一直是困难的。我们最近从比利时Overpelt的康复和MS中心提取了一个EPs数据集。我们的数据集足够大，可以首次获得包含不同患者的独立测试集的性能评估。 方法: 利用高度比较性的时间序列分析软件包，从运动EPs中提取大量的时间序列特征。与目标的互信息和Boruta方法用于发现包含未包括在文献中研究的特征中的信息的特征。我们使用随机森林 (RF) 和逻辑回归 (LR) 分类器来预测两年后的残疾进展。检查添加额外功能时性能提升的统计显著性。 结果: 与仅使用已知特征相比，在运动EPs中包括额外的时间序列特征导致统计学上显著的改善，尽管效果在幅度上是有限的 (对于RF，Δ auc = 0.02，对于LR，Δ auc = 0.05)。具有额外时间序列特征的RF获得最佳性能 (AUC = 0.75 ± 0.07 (平均值和标准差))，考虑到模型中生物标志物的数量有限，这是良好的。RF (非线性分类器) 优于LR (线性分类器)。 结论: 在EPs上使用机器学习方法显示出有希望的预测性能。使用已经使用的EP时间序列特征之外的附加EP时间序列特征导致性能的适度提高。需要更大的数据集，优选多中心，用于进一步研究。给定足够大的数据集，这些模型可以用于支持临床医生在关于未来治疗的决策过程中。
METHODS:PURPOSE:The aim of the study was to assess dual-task cost to spatio-temporal gait parameters in people with multiple sclerosis and a matched control group. METHOD:The multiple sclerosis group was composed of 17 participants with a diagnosis of multiple sclerosis and an Expanded Disability Status Scale score of less than 6. A total of 17 healthy participants were allocated to the control group by stratification. Controls were matched on the basis of age, sex, sociocultural habits, and body structure. Dual-task cost was determined by within-group repeated-measures analysis of variance. Participants were instructed to ambulate under normal conditions and perform a discrimination and decision-making task concurrently. Then, between-group analysis of variance was used to assess differences in mean dual-task cost between groups and determine dual-task cost differential. Testing was performed using three-dimensional photogrammetry and an electronic walkway. RESULTS:Based on dual-task cost differential, gait cycle time increase (-5.8%) and gait speed decrease (6.3%) because of multiple sclerosis-induced impairment. CONCLUSIONS:During single- and dual-task conditions, gait speed was lower in multiple sclerosis participants, because of a shorter step length and increased swing time. Increased gait time might be the result of compensatory mechanisms adopted to maintain stability while walking specially during the double-support phases.
METHODS:OBJECTIVE:The aims of the study were to compare mobility in multiple sclerosis, Parkinson disease, and stroke, and to quantify the relationship between mobility and participation restrictions. DESIGN:This is a multicenter cross-sectional study. Included were compliant subjects with Parkinson disease, multiple sclerosis, and stroke seen for rehabilitation, with no comorbidities interfering with mobility. Functional scales were applied to each subject to investigate gait speed (10-meter walking test), balance while maintaining body position (Berg Balance Scale), dynamic balance and mobility (Timed Up and Go and Dynamic Gait Index), and participation (Community Integration Questionnaire). RESULTS:Two hundred ninety-nine patients (111 multiple sclerosis, 94 Parkinson disease, and 94 stroke) were enrolled. Stroke had the slowest gait speed (mean gait speed = 0.9 m/sec) compared with Parkinson disease (1.1 m/sec), and multiple sclerosis (1.2 m/sec) (P < 0.001). Multiple sclerosis was more limited than Parkinson disease and stroke in dynamic balance both in the Timed Up and Go Test (multiple sclerosis = 16.7 secs, Parkinson disease = 11.4 secs, stroke = 14.0 secs; P < 0.001) and Dynamic Gait Index (multiple sclerosis = 11.6 points, Parkinson disease = 12.9 points, stroke = 13.6 points; P = 0.03); ability to maintain balance and body position (Berg Balance Scale) was more affected in stroke and Parkinson disease than multiple sclerosis (multiple sclerosis = 42.6 points, Parkinson disease = 39.4 points, stroke = 39.7 points; P = 0.03). Balance disorders were associated with participation restrictions but not gait speed. CONCLUSIONS:Neurological conditions have differing impacts on gait and balance, leading to different levels of participation restriction.
METHODS:OBJECTIVES:Cerebrospinal fluid (CSF) and blood neurofilaments (NFLs) are markers of axonal damage and are being investigated, mostly in relapsing-remitting (RR) MS, as a marker of disease activity and of response to treatment, while there are less data in progressive MS patients. Primary aim was to measure NFL in plasma samples of untreated patients with primary (PP) and secondary (SP) progressive MS and to correlate them with disability, disease severity, and prior/subsequent disability progression. MATERIALS AND METHODS:Neurofilament concentrations were measured using SIMOA (Single Molecule Array, Simoa HD-1 Analyzer; Quanterix). RESULTS:Neurofilament concentrations were measured on plasma samples of 70 progressive (27 PP and 43 SP), 21 RRMS patients, and 10 HCs. Longitudinal plasma NFL (pNFL) concentrations (median interval between sampling: 25 months) were available for nine PP/SP patients. PNFL concentrations were significantly higher in PP/SP compared to RRMS patients. They correlated with EDSS and MS Severity Score values. There was no difference in pNFL levels between PP/SP patients with EDSS progression in the preceding year (14% of patients) or during a median follow-up of 27 months (41%). In the longitudinal sub-study, pNFL levels increased in all patients between sampling by a mean value of 23% while EDSS mostly remained stable (77% of cases). CONCLUSION:In PP/SP progressive MS patients, pNFL levels correlate with disability and increase over time, but are not associated with prior/subsequent disability progression, as measured by EDSS, which may not be a sufficiently sensitive tool in this context.