Application of Machine Learning in Stroke Prediction: A Systematic Review

A brain stroke is a dangerous condition in which there is insufficient blood flow to a part of the brain, frequently as a result of brain haemorrhage or clogged arteries. Long-term oxygen deprivation results from this, which damages the brain irreversibly and kills brain cells. Stroke patients should receive treatment as soon as possible since restoring blood flow might lessen the extent of the damage. In the world, stroke ranks as the second most common cause of death. Algorithms for machine learning can assist in both patient rehabilitation and the detection of brain strokes. Large datasets are being analysed using ML techniques to spot trends that might suggest a person's risk of having a stroke. This is a growing area of use in stroke prediction. Key methods include NB, SVM, GB, RF, DT, KNN, Clustering, and Ensemble Learning. By combining predictions from several trees, these methods can help pinpoint important risk factors, understand how variables interact, and increase model accuracy. Subgroups and patterns in the data are also made easier with their assistance. Using the stack ensemble model, which employed three 1D convolutional models using neural networks from the survey done on the ECG dataset, the maximum accuracy of 99.70^% was achieved.