A Strategic Approach to Patient Diagnosis and Prognosis of Pancreatic Cancer Using Deep Echo State Network With Catch Fish Optimization Algorithm

Problem: Patients with pancreatic ductal adenocarcinoma (PDAC) can have a much better prognosis if they receive an early and precise diagnosis. Numerous studies have created automated techniques that use a variety of medical imaging to forecast the development of PDAC. In this work, we suggest a methodical approach for tracking, categorizing, and identifying pancreatic cancers. The model combines fish optimization methods with Deep Echo State Network (DESN) model technology. Aim: This concept aims to successfully integrate the two technologies. Here, we use a new dataset of diagnostic cases and the Medical Segmentation Decathlon (MSD) dataset to test the effectiveness and dependability of the diagnostic process using deep learning methods. Method: The proposed work employed image processing techniques like Gaussian and median filters, as well as SSA-MLT (Multilevel Thresholding) image segmentation based on Salp Swarm Algorithm (SAL) to enhance CT images of pancreatic cancers. In this paper, we provide a model for PDAC images called VGG16 (VGG16-Backbone Feature Extractor). DESN-CFOA is used in classification processing to extract features. The segmentation and classification performance of pancreatic tumor images were used to assess model training and configuration goals. Result: A thorough comparative data analysis reveals that the suggested approach performs better than existing technology, as demonstrated by extensive simulations. The proposed work is anticipated to produce a better performance in terms of Dice coefficient, Jaccard index, precision, and recall.