Earlier Prognosis of Malignant Tumor in Cerebral Magnetic Resonance Images using Machine Learning

Adults are at risk of developing brain tumors, which are characterized by the rapid multiplication of abnormal brain cells and can cause considerable damage or even loss of organ function. These tumors can vary widely in dimensions, texture, and position. Magnetic resonance imaging (MRI) is critical in the detection of malignant tumors. However, detecting brain tumors manually is tedious, error-prone, and time-consuming. The proposed solution includes optimized feature selection and Machine Learning (ML) models for enhancing the accuracy of brain tumor detection. This method can identify meningioma, glioma, and pituitary gland malignancies with high accuracy. First, the Brain MRIs are collected and pre-processed to increase their visual quality. Second, the Gray Level Co-Occurrence Matrix (GLCM) is used to extract the features from images. Third, Genetic Algorithm (GA) and Fire-Fly Optimization (FFO) are used to optimize the created feature vectors, allowing for the selection of more appropriate features. Fourth, for classification, three ML models such as Support Vector Machine (SVM), Naïve Bayes (NB), and Random Forest (RF) are employed. Finally, evaluation is done to determine the best combination of feature selection technique and ML. SVM using FFO features exceeds all other approaches, with an accuracy score of 98.55%, sensitivity of 97.81%, specificity of 99.26%, precision of 97.95%, and F1 of 98.6%. The SVM with FFO features can help the physician in the early diagnosis of brain tumors.