Prediction of clinical efficacy of acupuncture intervention on upper limb dysfunction after ischemic stroke based on machine learning: a study driven by DSA diagnostic reports data.

OBJECTIVE: To develop a machine learning-based model for predicting the clinical efficacy of acupuncture intervention in patients with upper limb dysfunction following ischemic stroke, and to assess its potential role in guiding clinical practice.

METHODS: Data from 1,375 ischemic stroke patients with upper limb dysfunction were collected from two hospitals, including medical records and Digital Subtraction Angiography (DSA) reports. All patients received standardized acupuncture treatment. After screening, 616 datasets were selected for analysis. A prediction model was developed using the AutoGluon framework, with three outcome measures as endpoints: the National Institutes of Health Stroke Scale (NIHSS), Fugl-Meyer Assessment for Upper Extremity (FMA-UE), and the Modified Barthel Index (MBI).

RESULTS: The prediction model demonstrated high accuracy for the three endpoints, with prediction accuracies of 84.3% for NIHSS, 77.8% for FMA-UE, and 88.1% for MBI. Feature importance analysis identified the M1 segment of the Middle Cerebral Artery (MCA), the origin of the Internal Carotid Artery (ICA), and the C1 segment of the ICA as the most critical factors influencing the model's predictions. Notably, the MBI emerged as the most sensitive outcome measure for evaluating patient response to acupuncture treatment. Additionally, secondary analysis revealed that the number of sites with cerebral vascular stenosis (specifically 1 and 3 sites) had a significant impact on the final model's predictions.

CONCLUSION: This study highlights the M1 segment, the origin of the ICA, and the C1 segment as key stenotic sites affecting acupuncture treatment efficacy in stroke patients with upper limb dysfunction. The MBI was found to be the most responsive outcome measure for evaluating treatment sensitivity in this cohort.