Revolutionizing acute ischemic stroke care: a systematic review of machine learning and deep learning applications in diagnosis and outcome prediction

Abstract Acute ischemic stroke (AIS) represents a major global health burden, with incidence projected to reach 89.32 per 100,000 people by 2030. This systematic review examines how artificial intelligence (AI), particularly machine learning and deep learning, can enhance AIS care through improved diagnostic accuracy and outcome prediction. We synthesize evidence on AI applications in lesion segmentation and functional outcome forecasting, with emphasis on clinical translation and relevance for diverse healthcare settings. Current literature demonstrates that AI-assisted approaches achieve clinically meaningful performance, with segmentation models frequently showing Dice coefficients exceeding 0.85 and outcome prediction models achieving area under the curve values above 0.80. Integration of multimodal data, combining imaging features with clinical parameters such as Barthel Index, Modified Rankin Scale, National Institutes of Health Stroke Scale, and Functional Independence Measure consistently enhances predictive accuracy. However, significant challenges persist, including demographic biases in training data, limited generalizability across populations, and reliance on small, single-center datasets. Ethical considerations around algorithmic fairness and the need for explainable AI in clinical decision support are crucial for equitable implementation. Successful clinical translation requires addressing workflow integration, validation in real-world settings, and development of approaches suitable for resource-limited environments. This review highlights the transformative potential of AI in stroke care while emphasizing the need for robust clinical validation and equitable deployment to ensure improved patient outcomes across diverse healthcare contexts.