Consistency-Based Patient-Motion Correction in Pinhole and Conventional SPECT – SPECT-Motion-eDCC

Single Photon Emission Computed Tomography (SPECT) refers to a technique for 3D imaging of a radioactive tracer that has been administered to a patient, in order to track certain biological functions. SPECT remains principally a diagnostic tool, but is now gaining increased usage for radionuclide-based therapeutic procedures. However, the long acquisition time (10-40 minutes) makes SPECT prone to patient motion artifacts in the reconstructed images which can render correct diagnosis or quantification impossible. Manual monitoring and correction procedures, and even re-scanning, are currently required, which can be difficult for patients and can impact clinical workflow and costs. The goal of SPECT-Motion-eDCC is to investigate and develop image-based methods to detect, identify and compensate for patient motion. We will investigate the use of exponential data consistency conditions (eDCC) to automatically detect and quantify patient motion that occurs during the course of the scan prior to SPECT reconstruction. The estimated motion will then be corrected for in the reconstruction of the SPECT image.

Images acquired by a SPECT scanner prior to reconstruction can be mathematically modeled by the attenuated Radon transform or, assuming that the emission occurs in a region with homogeneous attenuation, the exponential Radon transform. They are then referred to as “exponential projections”. Exponential data consistency conditions are equations, found from the range of the exponential Radon transform, that express the small redundancy of information between any two exponential projections. These mathematical conditions are known for parallel projections but have not yet been applied to correct patient motion in conventional SPECT. For divergent projections, no eDCC are yet known.

SPECT-Motion-eDCC is a collaborative project between groups with complementary expertise. The CREATIS team will develop tools, using Monte Carlo techniques, to realistically simulate parallel and pinhole SPECT cameras, for a selected variety of patient models. Patient motion will be included in the models, as well as the capability to ‘switch off’ various physical effects to generate ideal data. Detection and identification of patient motion will be performed by the TIMC team using eDCC, and tested initially on ideal data, successively adding more realistic components to explore whether the method breaks down in the presence of real physical effects and, if so, for which. Experiments will be carried out for validation with real data measured on conventional rotating clinical SPECT/CT scanners in Lyon (such as the GE Discovery 670) and on clinical multi-pinhole stationary SPECT scanners in Ottawa (such as the GE Discovery NM530c). The main deliverable will be an automated method to detect patient motion in real scans, thereby avoiding the current operator inspection of SPECT images. The endpoint will be a documented advanced understanding of the sensitivity of eDCC motion detection and how well motion can be identified and corrected in a completely automatic fashion. Mathematical deliverables will include new eDCC range conditions.