Patient Body Model Reconstruction and Clinical Applications in PET Imaging and Mixed Reality

Abstract

Patient-specific body models reconstructed from medical volumetric images, such as CT and PET, provide 3D representations of the patient’s body. These models offer valuable insights for clinical applications, including the direct measurement of patient body habitus measurements (BHMs), such as body volume and surface area, and surgical visualisation.

Despite their potential, several challenges limit the broader adoption of body models in medical contexts. The first challenge arises from missing body regions in the reconstructed models, caused by the scanner’s limited field of view and variations in patient positioning. This results in incomplete body models. The second challenge is the lack of reliable approaches for deriving patient-specific BHMs and incorporating them into clinical workflows, such as calculating the radiotracer dosage required for optimal PET imaging. Addressing this requires novel methods to adopt recovered full body models for automated BHM extraction and validate their use in dosage calculation. The third challenge involves integrating body models into mixed reality (MR) surgical applications, where alignment between holographic and physical bodies is essential but hindered by positioning differences and a lack of shared surface features.

This thesis addresses these challenges by proposing novel methods and an integrated framework: (i) recovering missing regions using a pipeline that combines parametric models with optimisation and geometric deformation, (ii) automating BHM derivation from full models for PET imaging optimisation, and (iii) aligning holographic models with patients in MR settings, accounting for position variation and clinical workflow integration.

The proposed methods were evaluated using imaging datasets and compared with state-of-the-art methods. Results demonstrate their effectiveness and potential to support the integration of body models into imaging and surgical systems for better workflows and intraoperative guidance.