Pancreatic Cancer Classification Based on Deep Learning

Abstract

Due in large part to its early asymptomatic character and the dearth of accurate diagnostic techniques, pancreatic cancer continues to rank among the most difficult cancers to identify and treat. Despite the fact that early and precise pancreatic cancer classification can greatly enhance patient outcomes, traditional diagnostic methods frequently have low sensitivity and specificity. In order to improve diagnostic performance, we provide a unique hybrid deep learning framework in this study for the classification of pancreatic cancer. This framework combines machine learning classifiers with sophisticated convolutional neural networks (CNNs). To identify pertinent anatomical features, the suggested approach starts with thorough preprocessing and segmentation of medical imaging data, such as computed tomography (CT) and magnetic resonance imaging (MRI).  Pre-trained deep learning models, such ResNet and EfficientNet, are used for feature extraction. These models have been refined on pancreatic datasets to capture complex patterns linked to cancers. To increase accuracy and generalization by 95%, particularly when data is scarce, these extracted features are fed into conventional classifiers like Support Vector Machines (SVM) and Random Forests (RF) rather than depending exclusively on end-to-end deep learning classification. According to experimental results, the hybrid approach performs better than solo deep learning models in terms of precision, recall, and F1-score, indicating that it has the potential to be a potent tool for clinical decision support in the diagnosis of pancreatic cancer.