Deep Learning for Classification and Localization of tumor in pancreas

This study presents an integrated approach for the classification of pancreatic tumors using a combination of advanced image processing and deep learning techniques. Initially, image enhancement is performed using Contrast Limited Adaptive Histogram Equalization (CLAHE) to improve the visual quality and highlight key features within MRI or CT scan images. Following enhancement, tumor regions are accurately isolated using the Attention U-Net segmentation architecture, which incorporates attention mechanisms to enhance focus on relevant areas while suppressing irrelevant background information. For classification, a Convolutional Neural Network (CNN) is employed, trained using the Stochastic Gradient Descent with Momentum (SGDM) optimizer to distinguish between normal and tumor-affected pancreatic tissues. The model outputs a binary classification result—either "Tumor" or "Normal." Performance evaluation is carried out using multiple quantitative metrics including the confusion matrix, accuracy, precision, recall, F1-score, and Peak Signal-to-Noise Ratio (PSNR). These metrics provide comprehensive insights into the model's classification reliability and image quality after preprocessing. The proposed framework demonstrates the potential of combining enhancement, segmentation, and deep learning classification to improve diagnostic accuracy in pancreatic tumor detection, offering a promising tool for supporting medical professionals in early and accurate diagnosis.

Keywords: Pancreatic Tumor, Image Enhancement, Attention U-Net, Convolutional Neural Network, Medical Image Classification, Deep Learning.