Temporal Super-Resolution in T1 Mapping: Pioneering Speed and Detail Preservation for Multimodal Consumer Health Data

Cardiac T1 mapping imaging is clinically significant as it provides valuable information for early diagnosis, disease classification, treatment planning, and prognostic evaluation of cardiac diseases. However, its long scan time and susceptibility to motion artifacts limit its widespread application in clinical practice. Therefore, it is necessary to install T1 mapping temporal super-resolution imaging technology on consumer electronic devices such as tablets to facilitate the assessment of cardiac function by medical professionals. This paper proposes a planar inter-frame acceleration method aimed at improving the speed of T1 mapping imaging, named Cross-frame Attention-based Temporal Super Resolution Network (CATSRNet). The CATSRNet achieves temporal super-resolution through multi-scale deformable interpolation in the Feature Temporal super-resolution (FTSR) module. Additionally, the Deformable Spatiotemporal Attention Enhance (DSAE) module utilizes deformable techniques to aggregate multimodal data, thereby better exploring the potential connections between multiple frames of data. Our model performs excellently on the CMRXRecon dataset, producing images that retain clear cardiac details and fewer artifacts. Transferring the cross-frame adaptive enhancement model to consumer electronic products can offer a more convenient and efficient solution for clinical diagnosis, ensuring consumer health, and promoting technological advancement and innovation in the field of healthcare.