A multimodal longitudinal generative adversarial network to discriminate high-risk cysts for the early detection of pancreatic cancer

There have been major advances in AI in the last five years, and their impact is already visible in the entertainment, automotive and manufacturing industries, but its transformative potential in early detection of pancreatic cancer is yet to be realized. PANC-CYS-GAN aims to address this bottleneck by developing innovative computational analytics, powered by deep learning and Generative Adversarial Network (GAN) to overcome the barriers of missing and noisy data and synthesise information from emerging data, longitudinal data points in an adaptable, continuous learning manner. This will allow to learn PC-related latent representation of messy and complicated distributions of longitudinal multimodal data to generate and validate plausible hypotheses with a functionality of learning to emphasize PC-linked variables/measurements focusing on cysts, particularly IPMN and MCN, as precursor lesions leading to a PC with the ability to counter unidentified challenges to ensure its safety and beneficial impact. The proposed solution is an explainable AI to interpret the results and aid decision-making. It aims to provide a range of short- and long-term measurable global benefits harnessing UK's positioning in health-informatics and artificial intelligence at the world stage to develop health technology emphasising holistic approach.