Differential Quadri-Partitioned Neutrosophic Interval-Valued Polynomial Attention-Based Deep CNN For Brain Tumor Detection

Abstract

 Brain tumor image classification is a vital part of the medical image area. Early treatment diagnosis
 of brain tumors is challenging through the Magnetic Resonance Imaging (MRI) in clinical neuroradiology.
 Brain tumor detection is a development of vital significance where Convolutional Neural Networks (CNN) find
 application. However, the accuracy and time required to detect brain tumors is a large challenge. To address
 the issue, the proposed Differential Quadri-partitioned Neutrosophic Interval-valued Polynomial Attention
based Deep CNN (DQNI-PADCNN) is introduced for brain tumor detection. Initially, the region of interest
 (RoI) detection is performed by using Differential Quadripartitioned Neutrosophic Sets. After RoI detected
 brain tumor images, classification is carried out via Interval-valued Quadripartitioned Neutrosophic Polynomial
 Attention-based Deep Convolutional Neural Network. The designed model includes convolutional layers, pooling
 layers, and fully connected layers. In the convolutional layer, feature maps consider RoI brain tumor images.
 Feature maps are sampled down and offered as input to the pooling layers. In this layer, the ELU function
 is estimated. Finally, the brain tumor detection result is obtained in the fully connected layer with higher
 accuracy and less time. Experimental evaluation is carried out by using the brain tumor dataset with different
 factors, such as the PSNR, the brain tumor detection accuracy, the brain tumor detection time, sensitivity, and
 specificity. The results confirm that the proposed technique achieves higher accuracy of the PSNR and disease
 detection with a minimum of time and space complexities than the conventional classification methods.

DOI: 10.5281/zenodo.15467692