Optimization method for 3D visualization of mobile medical images based on user interaction prediction

With the advancement in mobile device performance and the growing demand for 3D visualization of medical images, achieving efficient 3D visualization on mobile devices has become paramount. In this study, we dynamically select rendering methods and quality based on different interaction methods and process dynamics during the user exploring volume data, aiming to swiftly address user interaction requests. Additionally, we employ a trajectory prediction model by leverage of LSTM+MDN to discern inherent patterns in virtual camera parameter sequences from users' historical interaction statistics, predicting virtual camera motion trajectories for pre-rendering and further reducing waiting time for remote rendering. Through comprehensive experimental evaluations encompassing various network conditions, medical image data, and devices, we validate the efficacy of the proposed rendering optimization strategy and prediction model. The experimental outcomes demonstrate a 10% reduction in overall network transmission volume while preserving image quality and response time, thus enhancing user satisfaction with high-quality visualization and interactive exploration of mobile medical images.