Imaging the human body
Understanding the arrangement of bone, muscle, and fat tissue in the human body is essential for the diagnosis of diseases and the development of personalised digital twins. While imaging modalities, such as CT and MRI give accurate measurements inside the body, these modalities require bulky and expensive equipment, expert operators, and time-consuming procedures.
Optical scanning of the body is a modality that is becoming cheaper and of higher spatial resolution, however it is limited to capturing the anatomy of the skin surface only. By combining optical scanning techniques with statistical models of the human anatomy, based on large datasets of internal body scans, it may be possible to perform rapid and cheap scanning of individuals to estimate their internal anatomy.
We have developed a custom optical 3D scanner that consists of an array of 24 high-resolution video cameras surrounding a human participant in a standing position. The acquisition hardware can image with all 24 cameras at 15 frames per second, enabling state-of-the-art 3D dynamic imaging of the entire human body.
Novel GPU-accelerated algorithms reconstruct a dynamic 3D point cloud of the body surface from the 24 video sequences. We are fitting scaffold-derived finite element meshes to the 3D point clouds, thereby constructing dynamic subject-specific body models.
Additional measurements of ECG and spirometry have been integrated with the optical scanner, enabling physiological measurements combined with 3D body geometry. For example, the temporal volume change of the torso is compared with spirometer measurements. This model-based analysis will be coupled with computational models of oxygen demand under different physiological conditions.
The scanner and subsequent model fitting workflow in this work provides a tool for quickly and cheaply estimating subject-specific anatomy. Such a tool has the potential for providing personalised anatomical landmarks to register the 3D coordinate system of whole-body physiological models (or digital twins).