Enhanced Brain Tumor Classification with Inception V3 and Xception Dual-Channel CNN

Brain tumor is a major concern related to the medical community due to its increasing prevalence and the challenges associated with manual diagnosis. According to the scientific community, a brain tumor is an unusual growth in brain cells, some of which may develop into cancer. Manually diagnosing the severity of brain tumors is not only a challenge but also error-prone. To mitigate the issues with manual diagnosis, deep learning can be proposed as a potential solution to aid brain tumor diagnosis through medical imaging. Since the past decade, several deep learning-based models are developed for diagnosing brain tumor. In this article, a joint training technique which is a Dual-Channel Convolutional Neural Network (DCCNN) is proposed for the effective classification of brain tumor by leveraging the strength of two pre-trained models: Inception V3 and Xception. The Inception V3 model is known for its potentiality in extracting features from multiple scales while the Xception model is known for its superior performance in image classification tasks. Combining the representations learned by the two models, the joint training algorithm boosts the learning process and provides better generalization. With the incorporation of Inception V3 and Xception, the proposed model yielded better accuracy in classifying the type of brain tumor when compared to the existing deep learning-based models.