Design and Engineering of Clinical C-SPECT Patient and Task Specific Slits for Optimizing Sensitivity

Objectives: Today SPECT is widely available for qualitative noninvasive diagnostic assessment of coronary artery disease. However, low sensitivity & poor spatial resolution limits its clinical accuracy. An approach to elevate the sensitivity without sacrificing spatial resolution is a C-SPECT system, which is a cardiofocal stationary imager that consists of 14 modular NaI(Tl) pixelated detectors & task-specific variable slits-slats collimators arranged around the thorax. A conveyor system that exchanges different slits based on the imaging task & patient. Slits provide transverse collimation, & adjustable slats provide axial collimation. Since the system is nearing completion, we have updated the slit design, which was previously based on a population of cardiac-imaging patients, but considered only the general geometry of the system assuming essentially a continuous detector model. With the conveyor system complete & the detectors in place, we now present a design approach that considers the gaps between detector modules & directs the center of each projection near the center of each module for best performance. Methods: Design & engineering work is based on 378 previously acquired & analyzed SPECT imaging patients. That work identified two cylindrical predetermined imaging volumes (PIVs) that contain the heart for small & large patients, & a common ellipse that encompasses the patient body contours for 90% of patients. This patient ellipse is fixed relative to a separate ellipse that models the position of the conveyor links. There is some flexibility for positioning the PIVs within the body-contour ellipse, but slit locations need to be close to the slit ellipse to allow conveyor motion while providing shielding. This leads to the additional constraint that all slits need to be housed within a common holder for fabrication efficiency. With a specific resolution in mind, we then vary the position of the PIV & the link phase (i.e, the position of the conveyor link along its track) to determine the number of counts from a uniform cylinder filling the PIV & passing through aperture i onto detector i (Si) (i.e., good counts) & the number of counts onto detector i from adjacent slits (Oi) (i.e., multiplexed counts). We then maximize the sum of (Si + Oi)2 / (Si + 2Oi) with respect to PIV location within tolerance of its ideal location from patient data & the link phase. Results: The optimization leads to a practical & stable approach, projecting the center of