Deep Inception Based Hybrid Machine Learning Framework for Binary Classification of Brain Tumor MRI Scans

Accurate and early detection of Brain Tumors (BT) is pivotal to improve treatment planning and increase survival rates. A significant diagnostic system for the identification of brain disorders is Magnetic Resonance Imaging (MRI). In this study, a unique hybrid framework is developed by integrating InceptionV3, a deep learning model, with three machine learning models: AdaBoost, Random Forest (RF), and Logistic Regression (LR). High-dimensional spatial characteristics are extracted from pre-processed MRI data using the deep Inception model. Binary classification is then carried out by feeding these deep features into machine learning classifiers. Two hundred MRI images were used, half of which contained tumors and the other half of which did not. To ensure the reliability of the results, 50 distinct data splits and 10-fold cross-validation were employed. With an accuracy rate of 98.2% and an Area Under Curve (AUC) of 0.999, LR was the most successful. Next was RF, which had an accuracy of 94.6% and an AUC of 0.98. AdaBoost got an AUC of 0.874 and an accuracy of 87.4%. Experimental results prove that the hybrid technique achieves better classification accuracy and fewer false positives. The proposed framework is thus appropriate for clinical decision assistance since it strikes a compromise between learning depth and decision interpretability through the combination of deep feature representations and classifiers.