Answers are grounded in this paper and cite their sources.
Early diagnosis is essential for the prognosis of brain tumors. Conventional methods of brain tumor classification involve biopsy through invasive brain surgery. Here we worked on the analysis of 3000 Magnetic Resonance Imaging (MRI) brain images consisting of glioma, meningioma, pituitary tumors and healthy brains to develop non-invasive strategies for the detection of tumors using a machine learning approach. This work included data augmentation to achieve equal numbers of tumor and non-tumor samples 1500 each. Seven methods were used for the classification purpose: Logistic Regression, SVC, KNN, Naïve Bayes, Neural Network, Random Forest, and cluster analysis through K-means. Basic evaluating parameters were used as the performance indicators including accuracy, precision, recall, F1-score, and AUC to determine the efficiency of each model. Out of the four algorithms tested Logistic Regression and Random Forest made the highest test accuracy of 96% they were closely followed by Neural Networks at 95% for tumor versus non-tumor classification. Based on these results, the use of non-invasive MRI-based machine learning as an accurate diagnostic method for tumor detection is highly emphasized, but it requires the enhancement of their diagnostic model to accomplish its high-level goal.
No takes yet. Share an insight, caveat, or question.
Govindegowda et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: