Dasari Yugandhar, M. S. R. Naidu & Anilkumar B
Pancreatic cancer is one of the deadliest cancers with high mortality rates as it is often diagnosed late, leading to limited treatment options. This demands an effective classification algorithm to accurately detect pancreatic cancer in its early stages. In this study, we proposed an innovative hybrid diagnostic model combining optimization and artificial intelligence algorithms using computed tomography (CT) images. The proposed work commences with the collection of CT images. Then, the images are pre-processed using the adaptive bilateral kernel filtering (ABKF) approach to enhance image quality by minimizing noise. Then, a hybrid segmentation algorithm named Lionized Red Deer Optimization-based Deconvolution Thresholding Algorithm (LRDO-DTA) was developed to segment the pancreatic disease images. Further, a ResNet-50 approach was employed for performing feature extraction, which enables the extraction of the most informative attributes from the segmented images. Finally, a golden eagle optimization-based deep belief multilayer perceptron (GEO-DBMPN) was utilized to identify pancreatic cancers. The experimental results depict that the proposed strategy achieved 97.5% Dice Similarity Coefficient (DSC), 90.5% Cancer Specificity Index (CSI), 95.19% recall, 98.12% accuracy, and obtained minimum execution time of 1.5 s.