A tumor in the brain is an urgent deadly condition brought on by unnatural and unregulated cell division. These tumors differ immensely in terms of their locations, textures, and sizes. (MRI) is a vital tool in finding for tumors that are cancerous. Yet, recognizing brain tumors by manual can be challenging and takes a lot of time that could result in errors. The proposed approaches utilize deep convolutional neural networks to extract Sequential data that comes from several frames and features gathered through axial slices are subsequently supplied to a neural network for classification. To address this, an enhanced brain tumor detection system uses the ResNet and DenseNet models for the precise identification of meningioma, tumors and pituitary gland cancers. To enhance the accuracy and efficacy of tumor detection, the present research introduces a novel hybrid model which incorporates the best attributes of the ResNet and DenseNet architectures. The proposed strategy makes use of DenseNet’s dense connections for optimal information flow and ResNet’s residual connections to lessen the vanishing gradient issue. The hybrid model captures detailed patterns and features in brain images through the combination of these complementing elements. The obtained outcomes indicate how successful the two suggested models are in letting the transfer learning model achieve ResNet accuracy of 93.15% and DenseNet accuracy of $\mathbf{9 1. 2 4 \%}$.
Effective Classification and Identification of Brain Tumors using Hybrid Model of ResNet and DenseNet
2024-11-06
630456 byte
Conference paper
Electronic Resource
English