Gradient energy valley optimization enabled segmentation and Spinal VGG-16 Net for brain tumour detection

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Neuro-oncology

Background:

• Anomalous enlargement of brain cells, known as Brain Tumour (BT), poses significant risks due to potential damage to vital neural and vascular structures.
• Precise and early detection of brain tumours is critical for mitigating severe health consequences and improving patient outcomes.

Purpose of the Study:

• To introduce an advanced deep learning model, SpinalNet Visual Geometry Group-16 (Spinal VGG-16-Net), for the early and accurate detection of brain tumours.
• To enhance the performance of brain tumour detection through a combination of image processing techniques and a novel neural network architecture.

Main Methods:

• Magnetic Resonance Imaging (MRI) data underwent denoising using a bilateral filter.
• Brain tumour regions were segmented using entropy-based Kapur thresholding, with optimal threshold values determined by Gradient Energy Valley Optimization (GEVO).
• The Spinal VGG-16-Net model, integrating SpinalNet and VGG-16 Net, was employed for feature extraction and final brain tumour detection after image augmentation.

Main Results:

• The proposed Spinal VGG-16-Net model demonstrated a maximum accuracy of 92.14%.
• High performance metrics were achieved, including a True Positive Rate (TPR) of 93.16%, True Negative Rate (TNR) of 91.35%, Negative Predictive Value (NPV) of 89.73%, and Positive Predictive Value (PPV) of 92.13%.

Conclusions:

• The Spinal VGG-16-Net model shows significant promise for early and accurate brain tumour detection.
• The developed method offers a robust approach for brain tumour diagnosis, outperforming existing schemes in key performance indicators.