Pancreatic Cancer Classification Using Multimodal Imaging

Pancreatic cancer, as one of the most aggressive malignancies, poses a significant challenge to the medical community due to its complex heterogeneity and late-stage detection. This chapter presents a novel approach for pancreatic cancer detection and classification using techniques such as single image super-resolution, image segmentation, image registration based on the region of interests, image fusion, and classification based on global shape and local texture analysis of pancreatic region. Different types of convolution neural networks have been utilized in spatial and frequency domains. A good balance between accuracy and complexity at each stage of the proposed approach allows to hope for its active practical application. Experiments were done using the collected data set from Indian hospitals and two public data sets with good classification results achieved an accuracy of 97%.