EC-HDLNet: Extended coati-based hybrid deep dilated convolutional learning network for brain tumor classification

Abstract

Brain tumors are one of the most aggressive and dangerous forms of brain cancer, making their accurate and rapid detection critical for effective treatment. In this study, an innovative optimization driven hybrid deep learning model EC-HDLNet is proposed for classifying brain tumors in medical images. The model addresses limitations found in existing methods by minimizing pre-processing steps and optimizing deep learning models for better performance. The input images are pre-processed using Gaussian bilateral filtering (GBF), which effectively reduces noise while preserving edges. The Decouple SegNet module is then employed to segment the regions of interest, and deep features are extracted using the InceptionV3 model. For classification, the deep residual dilated convolution network (DResdiL) is introduced to enhance tumor classification accuracy. The proposed hybrid model presents a significant step forward in brain tumor classification, offering a more efficient, accurate, and practical solution for medical imaging applications. The experimental results show that EC-HDLNet outperforms existing state-of-the-art methods with an impressive accuracy of 99.78 %, precision of 99.65 %, recall of 99.72 %, and F1-score of 99.69 %. This method not only improves classification results but also reduces computational complexity and processing time by optimizing the model’s hyper parameters and integrating multiple advanced techniques.