Predictive Modeling for Ischemic Stroke Using Machine Learning

The main objective of this reserach work is to create a predictive model for ischemic strokes using machine learning techniques. The aim is to develop a reliable prediction system capable of identifying persons at risk. This work has involved three groups. Group 1 represents the Detection and Segmentation Network (DSN) which presents a unique architecture for automating ischemic stroke lesion segmentation on DWI and addressess subjectivity and data imbalance. Furthermore, the Multi-Plane Fusion Network (MPFN) improves accuracy, as demonstrated by significant findings from the ISLES2015 SSIS DWI dataset. Group 2 represents an approach that uses powerful machine learning algorithms, preprocessing, and feature selection to predict haemorrhagic stroke with high accuracy. The IST dataset is used to classify ischemic stroke subtypes using Extra Trees and Recursive Feature Elimination with Cross-Validation, which improves precision and contributes greatly to stroke research and therapeutic applications. The suggested method excels in precise ischemic stroke subtype classification, with high metrics such as accuracy 0.98, precision 0.94, recall 0.96, and F1 score 0.98, all with an efficient execution time of 0.16 s, making it useful for healthcare practitioners. In conclusion, the research confirms advances in ischemic stroke subtype categorization using approaches such as Extra Trees and RFECV, which enhance accuracy and interpretability for clinical decision-making. This is a crucial step toward verifying stroke research and treatment understanding.