Towards Smart Hybrid Feature Analysis and Deep Neuroevolutionary Brain Tumour Prediction Based on Transfer Learning

Abnormal cell growths within the brain or spinal canal are known as brain tumours, that can cause a range of symptoms. Early diagnosis and treatment are crucial for better outcomes. Deep learning techniques have shown impressive results in classifying brain tumours from medical imaging with minimal human involvement. MRI is the most commonly used imaging technique for detecting brain tumours. In this research study, an automated approach was proposed using greyscale MRI images for brain tumour detection. The approach includes initial enhancement to minimize colour variations, noise removal, and feature identification to locate the tumour region, threshold-based OTSU segmentation was used instead of colour segmentation. The study used exception, a novel framework that minimized the size and processing costs of deep neural networks for brain tumour diagnosis while also enabling better performance. Preliminary assessments of the Xception model using transfer learning showed good accuracy and prediction probability, with different probabilities for different layers when different layers were relearned. Overall, deep learning techniques in medical imaging can aid in the early detection and treatment of brain tumours, leading to better outcomes for patients.