Analysis of brain tumour and stroke prediction using selected machine learning algorithms

Abstract

In this work, brain tumour detection and stroke prediction are studied by applying techniques of machine learning. To address challenges in diagnosing brain tumours and predicting the likelihood of strokes, this work developed a machine learning-based automated system that can uniquely identify, detect, and classify brain tumours and predict the occurrence of strokes using relevant features. The system utilizes advanced algorithms to analyze medical data and images to extract meaningful patterns and relationships that can assist in accurate prediction. The machine learning algorithms used for the brain tumour prediction is the convolutional neural network (CNN), whilst for the stroke prediction, support vector machine, random forest, decision tree and logistic regression were used for purposes of comparison. The dataset was partitioned into 80 % for training and 20 % for testing the brain tumour images using the machine learning programming language Python. It was observed that when the CNN model was trained up to 100 epochs, it achieved an overall accuracy of 95.78 %. The precision of the simulated model was 96.70 %, recall was 96.65 %, and F1-score was 96.17 %. For the stroke images, it was partitioned into 90 % for training and 10 % for testing and the decision tree algorithm gave the most accurate results among the four machine learning algorithms with an area under the curve (AUC) score of 0.97 on the original dataset and a value of 1.00 after hyperparameter tuning.