Critical Diagnosis in Brain MRI Studies based on Image Signal Intensity and Supervised Learning

Abstract

The main objective of this investigation is to propose a new methodology for the detection of significantly critical findings related to the brain. To validate our method, we used magnetic resonance studies of 98 patients: 33 with healthy brains and 65 with brain pathologies. The proposed methodology was evaluated with five different machine learning classification models: KNN, Naive Bayes, Logistic Regression, Decision Tree and Random Forest. The supervised classification of these models shows outstanding results: the Naive Bayes model had the best results about the accuracy, kappa, and F-score, which was 100%. Due to its high performance in critical diagnosis classifications, it would allow prioritizing reading tasks, which could lead to a better clinical outcome for the patient.