Enhanced brain tumor classification in MRI using an optimized deep random graph dilated diffusion convolutional...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Biology

Background:

• Brain tumors (BT) significantly disrupt brain function, necessitating early detection for effective treatment.
• Accurate and timely diagnosis via MRI is critical for successful intervention in brain tumor cases.

Purpose of the Study:

• To develop a revolutionary brain tumor categorization approach using deep learning and optimization.
• To enhance tumor identification accuracy through a novel deep random graph dilated diffusion convolutional attention network (DR2DCAN) with a crested porcupine optimizer (CPO).

Main Methods:

• MRI preprocessing using a hybrid fast conventional bilateral filter (HFCBF) for noise reduction and edge preservation.
• Tumor region segmentation and feature extraction using DeepLabV3+ and multi-discrete Laguerre wavelet transforms.
• Classification using DR2DCAN with a random graph diffusion attention mechanism, optimized by CPO.

Main Results:

• The framework was tested on diverse MRI datasets including gliomas, pituitary tumors, and meningiomas.
• Achieved superior performance over existing methods with 98.7% accuracy, 98.4% precision, 98.8% recall, and 98.6% F1-score.
• Statistical analyses confirmed significant performance improvements.

Conclusions:

• The developed framework demonstrates high accuracy in brain tumor classification.
• It shows promise as a valuable tool for clinical applications and early diagnosis of brain tumors.