AI-Driven Brain Tumor Prediction using Feature Extraction and Machine Learning Classifiers

Abstract

Effective diagnosis and treatment planning for brain tumors depend largely on early and correct prediction. Conventional methods employed by MRI and histopathological analyses mandate an expert’s interpretation and are time-consuming. This study presents an AI based brain tumor classification system which utilizes Binary Patterns Pyramid Filter (BPPF) and Auto color correlogram filter (ACCF) for feature extraction along with IBK, Decision table and decision stump for prediction. A total of 7, 023 MRI pictures from Kaggle Brain Tumor dataset were utilized, which include glioma, meningioma, pituitary tumor, and no tumor class. As this experiment shows, IBK and BPPF integration has the best classification result of 92.98% and recall of 0.92, ROC-AUC = 0.94. In terms of computational effectiveness, ACCF coupled with Decision Table needed minimal execution time of 0.25 seconds hence appropriate for real-time applications. On the other hand, Decision Stump classifiers provided the lowest accuracy (~81%) and recall – which means low diagnostic reliability. The outcomes confirm that the synergies of BPPF with IBK produce greater diagnostic accuracy, and that for effective executions, the combination of ACCF and Decision Table is better, thus both combinations are ideal for Clinical Decision Support systems.

Topics & Concepts