目的: 采用随机森林算法构建基于早期康复数据的脑卒中患者吞咽功能恢复进展的预测模型,并验证其预测效能。方法: 本研究为一项前瞻性队列研究,共纳入88例急性脑卒中患者。收集患者的基础信息、神经功能评分、吞咽功能评估、康复训练数据及生理生化指标。使用随机森林、支持向量机（support vector machine,SVM）、逻辑回归等机器学习算法,构建预测模型并进行交叉验证,评估各模型在康复预测中的表现,包括准确率、灵敏度、特异性及受试者工作特征曲线下面积（AUC）。结果: 吞咽功能评分在第12周显著改善（平均分从入组时的42.58增加至68.47,P<0.001）,神经功能评分显著改善。随机森林模型在预测吞咽功能恢复方面表现最优,测试集AUC为0.809,显著优于支持向量机（AUC=0.774）和逻辑回归（AUC=0.733）（P<0.05）,显示了其在捕捉康复数据复杂非线性关系方面的优势。结论: 基于早期康复数据的随机森林模型能够较好地预测脑卒中患者吞咽功能恢复进展,为临床制定个性化康复治疗方案提供科学依据。
Abstract
Objective To construct a prediction model for the recovery progress of swallowing function in stroke patients based on early rehabilitation data using the random forest algorithm, and validate its predictive performance. Methods This study is a prospective cohort study involving a total of 88 patients with acute stroke. Baseline information, neurological function scores, swallowing function assessments, rehabilitation training data, and physiological and biochemical indicators were collected for each patient. Machine learning algorithms, including Random Forest, Support Vector Machine (SVM), and Logistic Regression, were used to construct predictive models. Cross-validation was performed to evaluate the performance of each model in predicting rehabilitation outcomes, including accuracy, sensitivity, specificity, and the area under the receiver operating characteristic curve (AUC). Results The swallowing function score showed significant improvement at Week 12(with an average score increasing from 42.58 at enrollment to 68.47, P<0.001). Additionally, there were notable improvements in neurological function scores, as indicated by the modified Rankin Scale (mRS, P=0.021), Barthel Index (BI, P=0.008), and NIHSS (P=0.036). The Random Forest model exhibited the best performance in predicting swallowing function recovery, with an AUC of 0.809 in the test set. This significantly outperformed both the Support Vector Machine (AUC=0.774) and Logistic Regression (AUC=0.733, P<0.05), demonstrating its advantage in capturing complex nonlinear relationships within rehabilitation data. Conclusion The Random Forest model based on early rehabilitation data can effectively predict the progress of swallowing function recovery in stroke patients, providing a scientific basis for clinical formulation of personalized rehabilitation treatment plans.
关键词
脑卒中 /
吞咽功能 /
恢复进展 /
随机森林 /
预测模型
Key words
stroke /
swallowing function /
recovery progression /
random forest /
prediction model
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 田娣, 刘燕. rTMS联合吞咽训练治疗脑卒中后恢复期吞咽障碍患者对吞咽功能和脑微循环的影响[J]. 黑龙江医学, 2024, 48(20): 2525-2527.
[2] 刘静娜, 赵艳艳. 脑卒中后吞咽障碍的全面评估与预后预测的新方法及趋势[J]. 全科护理, 2024, 22(20): 3819-3823.
[3] 于文静, 毛炅平. 间歇鼻饲、吞咽手法训练配合神经肌肉电刺激治疗仪对脑卒中吞咽障碍患者功能恢复的影响[J]. 临床医学研究与实践, 2024, 9(28): 147-150.
[4] 刘祎祎. 神经肌肉电刺激治疗对脑卒中后恢复期吞咽障碍患者吞咽功能及营养状况的影响[J]. 航空航天医学杂志, 2024, 35(09): 1035-1037.
[5] 陈娟, 彭瑾, 谢珊, 等. 跨理论模型教育护理联合精密摄食训练对脑卒中后吞咽功能障碍病人吞咽功能康复的影响[J]. 蚌埠医学院学报, 2024, 49(09): 1214-1219.
[6] 张亮, 刘学春, 胡小山, 等. 吞咽治疗联合舌压抗阻反馈训练对脑卒中吞咽障碍患者摄食能力、吞咽功能的影响[J]. 海军医学杂志, 2024, 45(08): 817-822.
[7] 周升霞, 张佳, 王祖萍, 等. Logistic回归模型和XGBoost模型对急性缺血性脑卒中患者发生吞咽障碍的预测价值[J]. 新疆医科大学学报, 2024, 47(08): 1179-1185.
[8] 李晓丹, 周颖, 刘胜锋, 等. 超声定量评估在脑卒中后吞咽功能障碍患者中应用价值分析[J]. 中国医学装备, 2024, 21(07): 71-75.
[9] 王轩, 邓颖. 老年脑卒中患者吞咽功能障碍研究进展[J]. 中西医结合护理, 2024, 10(3): 2-8.
[10] CHEN Y, FEI Z, CHUNQING X, et al.Community-based group rehabilitation program for stroke patients with dysphagia on quality of life, depression symptoms, and swallowing function: a randomized controlled trial[J]. BMC Geriatrics, 2023, 23(1): 876.
[11] 陈菁, 陶钰, 杨灵梅, 等. 不同算法的老年吞咽功能障碍患者吸入性肺炎风险预测模型比较[J]. 广西医科大学学报, 2023, 40(11): 1829-1835.
[12] SYLVIA I S, TINA H, ELLA R, et al.Concurrent and predictive validity of the Mann Assessment of Swallowing Ability in Belgian acute stroke patients based on a one-year follow-up study[J]. Folia phoniatr Logop, 2024, 76(2): 206-218.
[13] CHARULATA L, AASHISH A, SANGEETA P, et al.Swallowing Dysfunction after Acute Stroke: The Incidence, Predictors and Outcome[J]. J Assoc Physicians India, 2023, 71(8): 11-12.
[14] YANJUN R, JIHONG W, SHUANG L, et al.[Preliminary exploration of clinical prediction model of severe swallowing disorder after acute ischemic stroke based on nomogram model] [J]. Zhonghua wei zhong bing ji jiu yi xue, 2023, 35(4): 371-375.
[15] 鹿茜茜. 脑卒中吞咽功能障碍患者误吸现状及影响因素分析与预测模型的构建[J]. 中西医结合心血管病电子杂志, 2023, 11(09): 8-11.
[16] 林书阳, 陈妙, 周荣, 等. 洼田吞咽功能评分联合荧光吞咽透视检查评分在神经性吞咽障碍患者营养不良中的预测价值[J]. 中国现代医生, 2022, 60(31): 25-30.
[17] TANG A, CHEN X, MA J, et al.Characteristics of submental muscles function and hyoid bone movement in patients with dysphagia after stroke[J]. Clin Biomech, 2022, 100: 105772.
[18] CUNHA K M, FERNANDES V O D, CARVALHO L P O D, et al. Combined conventional speech therapy and functional electrical stimulation in acute stroke patients with dyphagia: a randomized controlled trial[J]. BMC Neurology, 2022, 22(1): 231.
[19] LIN Q, LI XY, CHEN LL, , et al.[Effect of acupuncture for dysphagia after stroke based on fiberoptic endoscopic swallowing function evaluation] [J]. Zhongguo Zhen Jiu, 2022, 42(5): 486-490.
[20] WANG T, TAI J, HU R, et al.Effect of Tongue-Pressure Resistance Training in Poststroke Dysphagia Patients With Oral Motor Dysfunction: A Randomized Controlled Trial[J]. Am J Phys Med Rehabil, 2022, 101(12): 1134-1138.
[21] REN X, HUANG L, WANG J, et al.Efficacy of systematic voice training combined with swallowing function exercises for the prevention of swallowing dysfunction in stroke patients: a retrospective study[J]. Ann Transl Med, 2022, 10(4): 195.
基金
江苏大学2021年度临床医学科技发展基金（JLY2021146）
PDF(3573 KB)
