Multilevel thresholding technique with Archery Gold Rush Optimization and PCNN-based childhood medulloblastoma classification using microscopic images

Childhood medulloblastoma (CMB) is an aggressive type of cancerous brain tumor that often affects children between the ages of 12 to 14, which significantly increases the mortality rate. This disease is considered the most common malignant tumor in the Central Nervous System (CNS). Thus, precise and early analysis of medulloblastoma is essential to obtain accurate treatment by enhancing the mortality rate. Moreover, various conventional approaches are exploited for the diagnosis, but the manual diagnosis process consumes more time, is subjective, and has many errors. Therefore, new techniques are required, which aim to improve diagnostic accuracy and convergence speed in clinical environments. A novel technique of Archery Gold Rush Optimization with Parallel Convolutional Neural Network (AGRO_PCNN) is developed, which is the fusion of Archery Algorithm (AA) and Gold Rush Optimization (GRO). Initially, the microscopic images from the IEEE Data Port are considered as the input, and the enhancement of the image is done by Contrast-Limited Adaptive Histogram Equalization (CLAHE). The cell segmentation is done by multilevel thresholding entropy based on Kapur’s method. Features, such as the Gray Level Co-occurrence Matrix (GLCM) and Pyramid Histogram of Oriented Gradients (PHOG) are extracted in the feature extraction phase. Finally, CBM is classified as a normal brain tissue cell and CMB cell by the AGRO_PCNN model. The experimental results assured that the AGRO_PCNN obtained the best accuracy, True Positive Rate (TPR), True Negative Rate (TNR), precision, and F1-score of 91.52%, 92.52%, 90.93%, 91.277%, and 91.897% at the training samples 90%.