Brain tumor segmentation and classification using MRI: Modified segnet model and hybrid deep learning architecture with improved texture features.
Journal:
Computational biology and chemistry
PMID:
40020564
Abstract
Brain tumors are quickly overtaking all other causes of death worldwide. The failure to perform a timely diagnosis is the main cause of increasing the death rate. Traditional methods of brain tumor diagnosis heavily rely on the expertise of radiologists, making timely and accurate diagnosis challenging. Magnetic Resonance Imaging (MRI) has emerged as the primary modality for brain tumor detection, but manual interpretation of MRI scans is time-consuming and error-prone. To address these challenges, an automated approach for brain tumor segmentation and classification (BTS&C) using MRI scans is proposed in this work. This work suggests a brain tumor classification scheme using MRI. Initially, the input images T1, TIC, t2 and t2 flair are fused via an improved fusion method. Then, Median Filtering (MF) is applied to preprocess the fused image. Also, the Modified Segnet model is proposed with a new pooling operation to do the segmentation process. Features like Improved local Gabor Binay pattern Histogram Sequence (ILGBPHS), Weber Local descriptor (WLD), and Tetrolet waveform are extracted from the segmented image. Finally, classification is done with HDLA that combines Bi-LSTM and Modified Linknet models. When TD= 90 %, the proposed method achieves a higher accuracy of 98 % which is compared to other methods like Bi-LSTM, Link Net, LeNet, Squeeze Net, Efficient Net, HHOCNN and CNN-SVM.