D2IP: Deep Dynamic Image Prior for 3D Time-sequence Pulmonary Impedance Imaging

--Unsupervised learning methods, such as Deep Image Prior (DIP), have shown great potential in tomographic imaging due to their training-data-free nature and high generalization capability. IP introduces three key strategies--Unsupervised Parameter Warm-Start (UPWS), T emporal Parameter Propagation (TPP), and a customized lightweight reconstruction backbone, 3D-FastResUNet--to accelerate convergence, enforce temporal coherence, and improve computational efficiency. IP enables fast and accurate 3D time-sequence Electrical Impedance T omography (tsEIT) reconstruction. IP delivers superior image quality--with a 24.8% increase in average MSSIM and an 8.1% reduction in ERR--alongside significantly reduced computational time (7.1 faster), highlighting its promise for clinical dynamic pulmonary imaging. ULMONARY imaging plays a critical role in the early diagnosis, monitoring, and management of respiratory diseases such as pulmonary edema, chronic obstructive pulmonary disease (COPD), and acute respiratory distress syndrome (ARDS) [1]-[4]. Among available techniques, tomo-graphic imaging modalities--including Computed Tomography (CT) [5], [6] and Magnetic Resonance Imaging (MRI) [7], [8]--are widely used in clinical pulmonary imaging due to their ability to produce high-resolution anatomical images. Hao Fang, Hao Y u, Sihao Teng, Zhe Liu and Y unjie Y ang are with the SMART Group, Institute for Imaging, Data and Communications, School of Engineering, The University of Edinburgh, Edinburgh, UK. (Correspondence authors: Y unjie Y ang and Zhe Liu; Email: y.yang@ed.ac.uk and zz.liu@ed.ac.uk). Tao Zhang is with the Department of Intensive Care Unit, Tianjin Huanhu Hospital, Tianjin, China.