Department of Physical Rehabilitation, National Center Hospital, National Center of Neurology and Psychiatry, Tokyo, Japan.
Department of Rehabilitation Medicine, Tokyo Bay Rehabilitation Hospital, Chiba, Japan.
PLoS One. 2023 May 26;18(5):e0286269. doi: 10.1371/journal.pone.0286269. eCollection 2023.
Stepwise linear regression (SLR) is the most common approach to predicting activities of daily living at discharge with the Functional Independence Measure (FIM) in stroke patients, but noisy nonlinear clinical data decrease the predictive accuracies of SLR. Machine learning is gaining attention in the medical field for such nonlinear data. Previous studies reported that machine learning models, regression tree (RT), ensemble learning (EL), artificial neural networks (ANNs), support vector regression (SVR), and Gaussian process regression (GPR), are robust to such data and increase predictive accuracies. This study aimed to compare the predictive accuracies of SLR and these machine learning models for FIM scores in stroke patients.
Subacute stroke patients (N = 1,046) who underwent inpatient rehabilitation participated in this study. Only patients' background characteristics and FIM scores at admission were used to build each predictive model of SLR, RT, EL, ANN, SVR, and GPR with 10-fold cross-validation. The coefficient of determination (R2) and root mean square error (RMSE) values were compared between the actual and predicted discharge FIM scores and FIM gain.
Machine learning models (R2 of RT = 0.75, EL = 0.78, ANN = 0.81, SVR = 0.80, GPR = 0.81) outperformed SLR (0.70) to predict discharge FIM motor scores. The predictive accuracies of machine learning methods for FIM total gain (R2 of RT = 0.48, EL = 0.51, ANN = 0.50, SVR = 0.51, GPR = 0.54) were also better than of SLR (0.22).
This study suggested that the machine learning models outperformed SLR for predicting FIM prognosis. The machine learning models used only patients' background characteristics and FIM scores at admission and more accurately predicted FIM gain than previous studies. ANN, SVR, and GPR outperformed RT and EL. GPR could have the best predictive accuracy for FIM prognosis.
逐步线性回归（SLR）是预测脑卒中患者日常生活活动能力的最常用方法，其使用功能性独立性测量（FIM）作为预测指标，但嘈杂的非线性临床数据会降低 SLR 的预测准确性。机器学习在医学领域越来越受到关注，因为它可以处理此类非线性数据。先前的研究报告称，机器学习模型（回归树（RT）、集成学习（EL）、人工神经网络（ANN）、支持向量回归（SVR）和高斯过程回归（GPR））对这种数据具有鲁棒性，可以提高预测准确性。本研究旨在比较 SLR 与这些机器学习模型对脑卒中患者 FIM 评分的预测准确性。
本研究纳入了接受住院康复治疗的亚急性脑卒中患者（N=1046）。仅使用患者的背景特征和入院时的 FIM 评分，通过 10 折交叉验证，分别构建 SLR、RT、EL、ANN、SVR 和 GPR 预测模型。比较实际和预测的出院 FIM 运动评分和 FIM 增益的决定系数（R2）和均方根误差（RMSE）值。
机器学习模型（RT 的 R2=0.75、EL 的 R2=0.78、ANN 的 R2=0.81、SVR 的 R2=0.80、GPR 的 R2=0.81）比 SLR（R2=0.70）更能预测出院 FIM 运动评分。机器学习方法预测 FIM 总分增益的准确性（RT 的 R2=0.48、EL 的 R2=0.51、ANN 的 R2=0.50、SVR 的 R2=0.51、GPR 的 R2=0.54）也优于 SLR（R2=0.22）。
本研究表明，机器学习模型在预测 FIM 预后方面优于 SLR。与以往研究相比，机器学习模型仅使用患者的背景特征和入院时的 FIM 评分，能更准确地预测 FIM 增益。ANN、SVR 和 GPR 的预测效果优于 RT 和 EL。GPR 可能对 FIM 预后具有最佳的预测准确性。
