Classification and Pixel Change Detection of Brain Tumor Using Adam Kookaburra Optimization-Based Shepard...

Area of Science:

• Medical Imaging and Artificial Intelligence
• Computational Neuroscience
• Oncology

Background:

• Brain tumors require early detection for improved patient survival rates.
• Accurate identification of tumor regions and classification is challenging due to variations in size, shape, and appearance.
• Magnetic Resonance Imaging (MRI) is crucial for diagnosing neurological conditions, including brain tumors.

Purpose of the Study:

• To develop an efficient and accurate method for brain tumor classification and pixel change detection.
• To introduce a novel optimization algorithm for enhancing the performance of deep learning models in medical image analysis.
• To improve the diagnostic capabilities for brain tumors using advanced computational techniques.

Main Methods:

• Development of an Adam kookaburra optimization (AKO)-based Shepard CNN (ShCNN) for classification and pixel change detection.
• Integration of kookaburra optimization algorithm (KOA) with Adam optimizer to create AKO.
• Pre-processing and segmentation of MRI scans using U-Net++, tuned by the bald Border collie firefly optimization algorithm (BBCFO).

Main Results:

• The AKO-based ShCNN achieved high performance metrics on the Brain Images of Tumors for Evaluation (BITE) database.
• Achieved accuracy of 93.78%, true positive rate (TPR) of 93.60%, true negative rate (TNR) of 92.26%, and positive predictive value (PPV) of 89.91%.
• AKO demonstrated faster convergence and higher classification accuracy compared to conventional optimization algorithms.

Conclusions:

• The proposed AKO-based ShCNN is effective for brain tumor classification and pixel change detection.
• The novel optimization strategy significantly enhances the performance of deep learning models in medical image analysis.
• This approach holds promise for improving early diagnosis and treatment monitoring of brain tumors.