A deep learning approach for brain tumor classification using hybrid efficient net and inception networks in MRI images

BackgroundA brain tumor is a medical pathology, whereas biomedical imaging is a key element used to visualize the condition. Brain tumor classification is an essential imaging task for early pinpointing and treatment. Accurate and timely tumor characterization supports appropriate treatment planning, prognosis assessment and better patient outcomes.ObjectiveThis paper aims to develop an efficient hybrid deep learning model for early detection of brain tumors from MRI (Magnetic Resonance Imaging) images.MethodsTo classify tumors, this work proposes a hybrid EfficientNet B7 and Inception V4 (ENETV4) network. The Inception network interprets spatial features for improving classification accuracy, while the EfficientNet B7 framework increases the computational efficiency to maintain high feature extraction ability. A unified feature is created by concatenating the retrieved features from the two models. To improve classification performance, a Figshare dataset comprising glioma, meningioma and pituitary tumor MRI images was used. Optimized hyperparameters enhanced generalization across the 70:20:10 train-test-validation split of the multi-sequence MRI dataset.ResultsThe proposed hybrid ENETV4 networks achieved 98.53% accuracy using an independent dataset, outperforming existing methods when evaluated on the dataset. These results accurately differentiate various kinds of tumors. This paper compares the hybrid ENETV4 model with the traditional models, reaffirming that the hybrid ENETV4 method enables effective classification of brain tumors from medical images, helping clinicians select suitable treatment strategies.ConclusionThe proposed hybrid methods provide accurate tumor classification in an early stage, supporting timely clinical decisions and treatment planning.