Real-time GB pattern convolution neural network-based brain image classification.

Medical image classification has been identified as a critical task in several medical solutions. Brain image classification has been identified as a challenging task that has been approached with different methods. The existing methods use different features like shape, texture, and gray to classify brain images. However, the methods need to improve performance in categorizing the brain image and detecting the presence of tumors. An efficient GB Pattern Convolution Neural Network (GBP-CNN) is presented to handle this issue. The method involves preprocessing the brain image given with Adaptive Feature Distribution Normalizer (AFDN) algorithm. The Grey Canvas Segmentation (GCS) technique was used to segment the normalized image. The feature vector is created by extracting the binary and grayscale patterns from the image that has been segmented. A convolution neural network is trained to recognize the features that were extracted. Two convolution layers in the CNN's design are used to convolve the image's features. Additionally, a pooling layer is being incorporated into the network's design, which reduces the data to a single dimension. The neuron predicts support values for several attributes during the assessment stage. The approach then calculates the Class-Based Fitness value (CBFV) in order to categories the image against various tumour classes. The proposed GBP-CNN algorithm improves the performance of brain image classification with less time complexity. [ABSTRACT FROM AUTHOR]