Brain Tumor Detection Based on Multimodal Information Fusion and Convolutional Neural Network

Aiming at the problem of low accuracy of traditional brain tumor detection, in this paper, a combination of multimodal information fusion and convolution neural network detection method of brain tumors, we call it a Multi-CNNs. First, this paper uses the extension of the 2D-CNNs to multimodal 3D-CNNs, and can obtain brain lesions under different modal characteristics of three-dimensional space. It can solve the 2D-CNNs raw input requires large neighborhood of faults, at the same time better to extract the modal of the differences between information. Then the real normalization layer is added between the convolution layers and pooling layer to improve the convergence speeds of the network and alleviate the problem of overfitting. In the end, the loss function was improved, and the weighted loss function was used to enhance the feature learning of the lesion area. The experimental results showed that the brain tumor detection method proposed in this paper could effectively locate tumor lesions, and better results were obtained in correlation coefficient, sensitivity, and specificity. Compared with two-dimensional detection network and single mode brain tumor detection methods, the detection accuracy is significantly improved.