Study Cohorts
Between December 2018 to December 2019, thirty-seven consecutive patients underwent a 99mTc-PYP thorax planar scan followed by SPECT and CT scans to diagnose suspected ATTR-CM. For each study subject, routine examinations were carried out to record comprehensive clinical data accordingly. This study was complied with the amended Declaration of Helsinki approved by the Institutional Review Board of Peking Union Medical College Hospital. All participants provided the informed written consent. According to the previous research study [18], patients were divided into three groups primarily based on clinical features, immunohistochemical or proteomics typing of amyloid, ECG, Perugini visual scores, genetic analyses and biopsy as the clinical routine for assessment of cardiac amyloidosis. Diagnosis of ATTR-CM included abnormal ECG finding and suggestive amyloidosis by visual grading of 99mTc-PYP planar images equal to 2 or 3 with absence of a detectable monoclonal protein despite serum/urine immunofixation electrophoresis (IFE) and serum free light chain (sFLC) assay. Group A: ATTR-CM (n=6) was based on clinical examination, ECG finding, positive 99mTc-PYP finding in planar images with Perugini visual scores >=2 and absence of abnormal serum/urine (IFE and sFLC). This diagnostic criterion identified one patient with ATTRwt and five patients with ATTRm. Heterogenous types of TTR mutations included Val50Gly (n=1), Val50Met (n=1), Gly73Glu (n=1), Asp38Asn (n=1) and Ala117Ser (n=1). Group B: AL-CM (n=10) was solely determined according to the presence of abnormal serum/urine (IFE or sFLC) as in lambda (λ)-light chain type (n=7) and kappa (κ)-light chain type (n=3). Group C: Others (n=21) that disqualified to fit into the diagnostic criteria of group A and group B. Several of them were ATTR mutation carriers from family history (n=13) as Ala117Ser (n=7), Val50Met (n=3), Ser97Tyr (n=2) and Asp38Asn (n=1) by genetic analyses but without an evidence of showing the burden of cardiac amyloidosis. The remaining patients included hypertrophic cardiomyopathy (n=2) and idiopathic cardiomyopathy (n=5). Patient characteristic of these three groups are listed in Table 1.
Phantom Experiment
For image quantitation, the experiment to derive the image conversion factor (ICF) to convert pixel value in quantitatively reconstructed SPECT images to 99mTc activity concentration was initially conducted by filling ~740 MBq of 99mTc water solution into a cylindrical phantom (radius 16 cm, height 20 cm). ICF was then calculated by the 99mTc true activity concentration divided by pixel value. To note, ICF maintains a constant value when physical interference from attenuation, scatter and statistical noise can be fully addressed in reconstructed images. To access partial volume effect (PVE) in myocardium (Myo), a standard cardiac insert phantom (Data Spectrum Corporation, Hillsborough, NC, USA) representing a 3-dimensional model of the left ventricle containing regions of the myocardial wall (~110ml) and ventricle (~60ml) was then utilized to measure PVE and to derive partial volume correction (PVC) factor under various activity concentration ratios (ACR) ofMyo and blood-pool (BP) in the ventricle cavity (0.15 to 10.0). PVE specified by the level of erroneous activity concentration was defined by the measured activity concentration divided by the true activity concentration in Myo. As reported, the degree of PVE considered as a function of wall thickness and Myo/Bp ACR remains approximately a constant level in the circumstance that Myo/BP ACR is over a certain threshold, and inversely, it rises rapidly with the decreased Myo/BP ACR [25]. The unique property of PVE provides an opportunity to derive and fit the PVC factor (1.0/PVE) as an analytic function of Myo/Bp ACR to recover the true activity concentration for the relatively unchanged wall thickness. In this study, acquisition parameters of SPECT scans for the phantom experiment were identical to those used in the patient scanning protocol as indicated in the next section.
Image Data Acquisition
Each study subject was intravenously injected with a ~740MBq 99mTc-PYP dose prepared by Beijing SHIHONG Pharmaceutical Center and calibrated by a radioactivity meter (CRC-25R, CAPINTEC, USA). Relevant parameters including injection dose, time and site were properly recorded. Post the 99mTc-PYP injection for one hour, a planar scan was performed in anterior and left lateral views for 10 minute and then followed by a SPECT scan in the thorax position on a dual-head SPECT camera (Discovery 630, GE Healthcare, Haifa, Israel). The SPECT camera consists of low-energy high-resolution collimator with 9.53 mm thickness of NaI(Tl) scintillation crystal. With patient’s heart positioned in the center field of view, planar images were acquired for a total of 750,000 counts with 256x256 matrix and 1.46 zoom factor. Imaging parameters for SPECT acquisition utilized 128x128 matrix, circular orbit (radius 30 cm), 180o arc, step-and-shoot, 30 steps at 40 secs/step, zoom=1.0 and multiple energy windows (126-154keV and 109–125keV). After the completion of SPECT acquisition, a low-dose free-breathing CT scan (120 keV, 35 mA, 12 sec) was separately acquired on a dedicated PET/CT scanner (Sinounion Polar Star m660, Beijing, China) for attenuation correction of SPECT images and image fusion. The patient positioning between two scans was optimally consistent to avoid non-translational misregistration.
Image Processing of Quantitative SPECT
In this study, image reconstruction and data analysis of quantitative SPECT were performed using a cardiac software package (MyoFlowQ, Taipei, Taiwan). This software incorporates image reconstruction and subsequent image analysis on a single platform to measure 99mTc or 99mTc-PYP activity concentration in regions of myocardial wall and ventricle cavity. For the quantitative image reconstruction of SPECT, projection data were pre-corrected for 99mTc isotope decay according to time points of rotation angles, and reconstructed by ordered subsets expectation maximization (OSEM) (4 iterations, 12 subsets) with full physical corrections for photon attenuation, scatter, collimator resolution and Poisson count-statistics as described previously [26-28]. Prior to the quantitative image reconstruction, a rapid image reconstruction with filtered back-projection (FBP) was preliminarily executed to provide quick SPECT images for the assessment of registration with CT images. SPECT-CT misregistration was verified visually and manually corrected by applying 3D translation to SPECT images. In the phantom experiment, a consistent region of interest (ROI) was drawn on SPECT images of the cylindrical phantom to count rate in pixel (unit: counts/sec/pixel) to 99mTc activity concentration (Bq/ml). To measure myocardial activity of the cardiac insert phantom, SPECT images were manually reoriented into the short-axis view. A threshold of 25% of peak activity was chosen to effectively differentiate between myocardial and ventricle regions. The myocardial centerline contour was automatically detected and refined by using an ellipsoid-approximated geometry with manually determined mitral valve plane to create the polar map. A consistent sampled region (1.0Í1.0Í2.0 cm3) was automatically placed in ventricle to measure the activity concentration of BP. PVC factor defined as the true 99mTc activity concentration divided by the measured 99mTc activity concentration from quantitative SPECT was presented in the scatter plot (y-axis=PVC factor, x-axis=measured Myo/BP ACR) and regressed with an exponential recovery model to derive analytic PVC factor as a function of measured Myo/BP ACR [29]. For the analysis of patients’ 99mTc-PYP SPECT images, the same processing steps, including image reorientation, myocardial centerline contour, creation of polar map and the placement of sampled region in ventricle cavity were performed identically to those processing steps of cardiac insert phantom. Under the situation when the determination of myocardial centerline contour was failed due to ultralow or no uptake of 99mTc-PYP in myocardium, a ROI (1.0Í1.0Í1.0 cm3) was manually placed in the insertion point between left and right ventricles on a transaxial plane of reconstructed images as the located joint of apex in the right ventricle with the apical septal of left ventricle. Post the recovery to absolute 99mTc-PYP uptake using PVC factor derived from the phantom experiment, standardized uptake value (SUV) was calculated with factors of injected 99mTc-PYP dose and patient’s body weight.
Interpretation of Planar Images and Semi-quantitative Measurement
Both anterior and lateral views of 99mTc-PYP planar images were evaluated by two consensus nuclear readers in nuclear cardiology to grade using the visual grading rule reported by Perugini and et. al. as: grade 0=cardiac uptake not visible; grade 1=mild cardiac uptake visible but inferior to skeletal uptake; grade 2=moderate cardiac uptake visible equal to or greater than skeletal uptake;grade 3=strong cardiac uptake with little or no skeletal uptake. The semi-quantitative analysis of planar images was performed by drawing a patient-specific circular ROI on the heart and mirror it to the contralateral chest in order to calculate the heart-to-contralateral (H/CL) ratio from the quotient of the mean counts [17].
Measurement of Intra- and Inter-operator Reproducibility
Correlations of image processing for semi-quantitative and quantitative parameters by the 1st operator (OP1) and the 2nd operator (OP2) were verified by linear regression. OP1 had twenty years of experience in image processing of nuclear cardiology, and OP2 encompassed three years of experience. To test the intra-operator reproducibility, OP1 processed all image data twice in four weeks apart. To test the inter-operator reproducibility, OP2 processed the same image data sets independently.
Statistics Analysis
All datasets were analyzed with a statistical software package (IBM SPSS Statistics version 25). Continuous variables were presented as mean±SD, whereas categorical variables were expressed as actual numbers and percentage. For the comparison between study subgroups, differences in continuous variables were analyzed using the one-way ANOVA with post hoc Bonferroni correction when the Levene's pre-test for homogeneity of variances meet the requirement, otherwise the one-way Welch ANOVA with post hoc Games-Howell correction was applied. Differences in categorical variables were analyzed using the χ2 test or Fisher exact test. The correlation of H/CL ratio and quantitative parameter (SUVmax) from OP1 and OP2 was obtained by the linear regression. Difference of correlation coefficient between two measurement was tested by the Z-test. Bland-Altman statistics were utilized to verify the systematic difference with a 95 % confidence interval (CI) for semi-quantitative and quantitative parameters. The repeatability coefficient (RPC) representing intra- and inter-operator reproducibility was calculated as: RPC=1.96×SD of difference between the two measurements [30]. All p values used were two sided with P<0.05 considered statistically significant.