A Hierarchical Dual-Branch CNN with Feature Concatenation for Brain Tumor Classification

This paper presents a hierarchical dual-branch convolutional neural network (Hierarchical Dual-Branch CNN with Feature Concatenation) for brain tumor classification; this model uses a feature concatenation structure and is trained by a brain tumor MRI dataset.The model's ability to efficiently extract image features improves the accuracy of recognizing various brain tumors and also for non-tumor cases.The performance of the model is evaluated and compared with transfer learning methods based on popular architecture such as MobileNetV2, and the structure-based model proposeds in previous researches.According to experimental results, the proposed method performs better on the test dataset across a range of assessment metrics than the other models.