Abstract Background: Stroke is a significant health problem in India, with an uneven prevalence and high early mortality rates. Worldwide, stroke is the second leading cause of death. In the initial months post-stroke, motor impairment is a primary concern, in addition to various other deficits. Predicting recovery after a stroke is crucial for optimizing resource allocation. Utilizing machine learning offers the potential to enhance therapeutic decision-making and predict individual recovery outcomes in stroke rehabilitation. Objectives: We aimed to employ machine learning algorithms to predict functional recovery after stroke by estimating Barthel Index scores, identifying patterns and correlations within the dataset, and determining the most effective machine learning model among the five algorithms that were tested. Materials and Methods: Participants were screened for eligibility before enrollment, and demographic information, stroke characteristics, and Barthel Index scores were recorded. The dataset was split into training and testing subsets for analysis. Five machine learning algorithms were trained using the initial dataset to develop predictive models. Results: High alcohol and tobacco use potentially influenced Barthel Index scores and stroke recovery. The recovery process varied based on stroke type, with ischemic and hemorrhagic strokes. The Random Forest model exhibited the highest predictive accuracy among the models. Conclusion: The study highlights the role of demographics, lifestyle habits, comorbidities, and stroke type in functional recovery poststroke. The Random Forest model demonstrated the most reliable predictive capability, indicating artificial intelligence’s potential in stroke recovery prediction. Furthermore, research studies are needed to develop and evaluate robust prediction models.
Read full abstract