Cameras and study design
The study was designed to validate the clinical feasibility of the novel digital SPECT system (Valiance X12 prototype) and compare the overall clinical performance of the digital system to that of analog systems routinely used.
The analog cameras used for comparison were the Discovery NM/CT 670 and Infinia Hawkeye 4 (GE Healthcare) fitted with low-energy high-resolution (LEHR) collimators. Images were reconstructed using the standard protocols used in our department, with the vendor-provided ordered subsets expectation minimization (OSEM) iterative reconstruction (all except brain studies) or filtered back-projection (FBP) reconstruction (brain studies).
The digital Valiance X12 prototype (Fig. 1) is comprised of CZT-based elongated detectors, 16 by 64 pixels each, covering approximately 4 by 16cm, mounted on a ring-shaped gantry. The detectors are electronically pixelated and are equipped with high-sensitivity tungsten parallel-hole collimators, with a wider opening solid angle than the analog SPECT LEHR collimator (about two times wider). The square collimator holes are in registration with the pixel array. The system’s architecture provides radial and swivel motion of the detector, which allows for targeted region of interest (ROI) centered imaging. Images are produced using a proprietary iterative reconstruction engine, based on an ordered subsets maximum a-posteriori (OS-MAP) algorithm, which accurately models the system and acquisition geometry.
Thirty-five patients (21 male, 14 female, mean age 50 ± 17 years, range 19–74 years) were scanned. All patients underwent planar and SPECT (or SPECT/CT) imaging in the analog SPECT camera, immediately followed by SPECT in the Valiance X12 prototype. Overall, a total of 36 analog SPECT scans were available (30 Infinia Hawkeye 4, 6 Discovery NM/CT 670).
Twenty patients underwent 99mTc MDP SPECT scans of 21 regions on both analog and digital SPECT. Skeletal regions evaluated included the knees (n = 8), lumbar spine, pelvis and hips (n = 5), ankles and feet (n = 5), skull (n = 1), cervical spine (n = 1), and tibia (n = 1). One patient had two skeletal regions evaluated. The same patient also underwent both standard and targeted ROI-centric scans on the digital SPECT camera in the same session resulting in 23 bone scans acquired by digital SPECT and 21 by analog SPECT.
Additional SPECT studies included 99mTc-ECD brain perfusion (n = 5), 99mTc-MAA lung perfusion (n = 3), 99mTc MIBI Parathyroid (n = 3), 99mTc DMSA renal (n = 2), 99mTc -nanocolloid breast sentinel node (n = 1), and 99mTc HMPAO-labeled leukocyte study (n = 1).
The study was approved by the institutional review board (IRB), and all subjects signed an informed consent form.
Image acquisition
Patients were scanned in the analog scanner according to the clinical referral and procedure guidelines used regularly in the department. Digital SPECT studies were performed 30 to 60 min after completion of the analog study. The actual scan times used were comparable between the analog SPECT and digital SPECT.
Image acquisition workflow in the Valiance X12 prototype is as follows: The patient is positioned on an imaging table inside the scanner bore. The detectors move toward the patient until close proximity is achieved, after which the system calculates a contour containing the targeted body region. The system continues to acquire a rapid (approximately 1–2 min) low-resolution preview SPECT image to be used for the calculation of the patient-specific sampling scheme. The sampling scheme includes the order, number, duration, and angles for the detector positions, which are optimized to reduce acquisition bias resulting from insufficient statistics. To further refine the image acquisition, a target-specific ROI may be interactively placed on the preview SPECT images such that inactive areas are excluded (e.g., the imaging table). Optionally, a second ROI can also be placed within the first one, yielding a scan design in which a greater portion of imaging time is allocated to acquiring radiation emanating from the specific region in focus. Scans performed using this methodology are termed ROI-centric or ROI-focused scans. An illustration of the scan planning methodology is presented in Fig. 2.
Following the completion of the planning stage, the study includes acquisition of multiple views at each gantry position (median 180, range 100 to 500), using a dedicated mechanism for rotating the detection modules, each around its own axis (swivel motion). Once a gantry position acquisition is completed, the gantry rotates to the next position and the process repeats itself until scanning has been completed for all the planned gantry positions and full 360° angular coverage is achieved.
Image reconstruction
Apart from brain studies, analog SPECT data was reconstructed using OSEM reconstruction [12] routinely used in the department, with 64 Gantry positions, 2 iterations, and 10 subsets, and was filtered using a Butterworth filter with cutoff frequency in the range of 0.33–0.5 and power of 10.
Analog brain studies were reconstructed using filtered back projection with Chang attenuation correction [13] and were filtered by a Metz filter with a point spread of 4 mm and power of 3.
Digital CZT SPECT data was reconstructed using the proprietary OS-MAP iterative reconstruction algorithm as previously described, with median root prior (MRP) [14].
We used 6–10 gantry positions (typically 8), 3 iterations, and 40 subsets.
For brain studies, similar assumptions to Chang-based attenuation correction, typically used for FBP reconstruction (i.e., uniform linear attenuation coefficient), were incorporated into the system model and used for attenuation correction in the iterative reconstruction.
Interpretation of findings
All images (analog and digital) were evaluated by two experienced Nuclear Medicine physicians who were blinded to the origin of images and were graded on a scale of 1 (poor) to 4 (very good) for sharpness, contrast, uniformity, overall quality, and diagnostic confidence. When there was disagreement, the results were obtained by consensus.
Results were tested for statistical significance using the single-tailed paired T test.
Increased uptake of 99mTc-labeled tracers (MDP, MIBI, WBC, nanocolloid) was classified as diffuse or focal, and location of the finding was noted. Reduced uptake of 99mTc-labeled MAA, DMSA, and ECD was similarly described as focal or diffuse, and their location noted.
Findings from conventional SPECT and digital SPECT studies were considered congruent when both sets of images demonstrated the same findings at the same anatomical location. Discrepancies were characterized as “additional/missing” pathology (seen on Valiance X12 prototype), using the conventional SPECT study as reference.