Segmentation and classification of brain tumor using Taylor fire hawk optimization enabled deep learning approach.

Area of Science:

• Neurology
• Medical Imaging
• Artificial Intelligence

Background:

• Brain tumors arise from irregular cell growth, necessitating early diagnosis for effective treatment.
• Conventional diagnostic models face challenges in timely and accurate brain tumor detection and classification.
• Accurate segmentation and classification are critical for developing substantial treatment plans.

Purpose of the Study:

• To implement an optimized deep learning approach for enhanced brain tumor segmentation and classification.
• To introduce the Taylor Fire Hawk optimization (TFHO) algorithm for improving diagnostic accuracy.
• To overcome limitations of conventional models in identifying and categorizing brain malignancies.

Main Methods:

• Adaptive median filter for image de-noising.
• M-Net model, trained with TFHO, for image segmentation.
• Image augmentation and feature extraction.
• DenseNet classifier trained with TFHO for tumor classification.

Main Results:

• Achieved 94.86% accuracy in brain tumor classification.
• Demonstrated high diagnostic performance with a True Positive Rate (TPR) of 95.91% and F1-score of 90.98%.
• Reported a low False Negative Rate (FNR) of 4.37% and Positive Predictive Value (PPV) of 89.33%.

Conclusions:

• The TFHO-enhanced M-Net and DenseNet model significantly improves brain tumor segmentation and classification accuracy.
• This optimized approach offers a promising tool for early and precise diagnosis of brain malignancies.
• The method's high performance metrics suggest potential for clinical application in neuro-oncology.