An Enhanced Diagnostic System using Deep Learning for Early Prediction of Pancreatic Cancer
Kumbham Bhargavi, M Lakshmi Prasad, G Arun, Pundru Chandra Shaker Reddy, S. Yuvalatha, Mogalagani Triveni · 2024
With a dismal 5-year survival-rate of approximately 7%, pancreatic-cancer(PC) remains one of the world's most lethal tumors. In order to improve patient survival rates, early detection of PC is essential. Computerized-tomography(CT), MRI combined with MRCP, or a biopsy is necessary for the diagnosis of PC. The steps involved in the suggested CAD design process are as follows: picture preprocessing, segmentation, feature-extraction & classificatiion. Color conversion and the isotropic diffusion filter method are utilized for preprocessing. The next step is the segmentation processes' usage of the suggested Fuzzy K-NN Equality code. A classification tool that makes use of Deep Learning is feature extraction. Using the characteristics gathered from the pancreatic sample, tumor cells are categorized. The image classification criterion includes both the train values and the testing datasets. To identify pancreatic cancer, an algorithm called DCNN_DBN is employed, which combines Deep-Convolutional Neural-Networks with Deep-Belief-Networks. The results of the experiments show that the present CAD system has great promise and is safe for the automated diagnosis of benign & malignant tumors, with an accuracy rate of 99.8 percent. The use of this classifier greatly reduces the computational complexity. More anomalies in pancreatic cancer cells could be detected with an improved version of the proposed method.