Personalization of 99mTc-sestamibi activity in SPECT/CT myocardial perfusion imaging with the cardiofocal SmartZoom® collimator

Study design

First, the variable “myocardial signal” (MS) was defined as the average count in the myocardial region of interest (ROI) in the MPI-SPECT image. Second, the MS in stress MPI-SPECT images of a large patient cohort was determined retrospectively. The rest images were not analyzed because the threefold additional activity that is injected at rest leads to high signal-to-noise ratios in any case. The analysis of the stress images yielded an equation that described the relationship between MS and patient weight (Eq. 1). Third, Eq. 1 is used to derive an equation that described the relationship between MS, patient weight, and injected activity over a constant scan time (Eq. 2). Fourth, the MS threshold that indicated acceptable image quality for perfusion-defect diagnosis was determined by asking seven nuclear physicians to evaluate the quality of a subset of the cohort stress images. Fifth, to display the utility of our approach, the MS threshold was plugged into Eq. 2 to generate a table listing the optimal activity (or scan time) for patients with specific body weights.

Patient selection and MPI-SPECT imaging

The cohort consisted of 294 adult patients who underwent an exercise or pharmacological stress test and 99mTc-sestamibi MPI-SPECT with a cardiofocal collimator-equipped camera. It was a consecutive series that included healthy patients and patients with perfusion defects. It was used to derive the equation between body weight and MS; a subset (n = 50) was also used to evaluate image quality.

In accordance with the clinical procedures in our institution, the 99mTc-sestamibi activity in all patients was ~ 260 MBq. Acquisition was performed with a conventional large-field dual-head Anger gamma-camera Intevo Bold (Siemens Healthineers) equipped with the SmartZoom® cardiofocal collimator [4] and the IQ-SPECT cardio-centered acquisition process [17]. The acquisition parameters are specified in Table 1. Scan time was fixed to 20 s per projection angle, using 17 projections. After manual reorientation of the input tomograms to achieve optimal separation of the left and right ventricle activity, the 4.8-mm cubic voxel size SPECT images were reconstructed with an iterative 3-dimensional ordered subsets expectation maximization (OSEM3D) algorithm (10 iterations, 3 subsets), a 10-mm full-width at half-maximum (FWHM) Gaussian filter, and a mask to exclude any surrounding activity (for example, that from gastrointestinal uptake). Only the uncorrected (UC) images were analyzed in this retrospective study. This was to avoid potential bias caused by the attenuation correction [18]. Moreover, it is likely that obtaining good quality UC images will further improve the corrected images. In addition, using the UC images is consistent with the clinical habits in our institution and others, which is to work with both UC and corrected images.

Table 1 Clinical acquisition parametersDefinition of myocardial signal and its relationship with patient weight or BMI

For each stress test UC MPI-SPECT image, an ROI corresponding to the myocardial area (i.e. the MS) was defined as the pixels whose counts exceeded 45% of the maximal pixel count in the whole image. Since a 10-mm FWHM Gaussian filter is applied to the reconstructed image, the maximal pixel is subject to very little noise and is an accurate representation of the actual tracer fixation in the myocardium. The 45% threshold was empirically chosen to obtain the best correlation between the ROI counts and patient weight or BMI. Figure 2 shows examples of these ROIs in three patients.

Fig. 2figure 2

Myocardial ROI in the uncorrected clinical SPECT images of three patients. The ROI was defined as the pixels whose counts exceeded 45% of the maximal pixel value

The MS of the 294-patient cohort was plotted as a function of patient weight or BMI, and the relationship between these variables was determined empirically by power fitting. To assess whether weight or BMI correlated better with MS at a statistical level, the means of the absolute relative error (%) in each model were compared by paired t-test (alpha level 0.05). To confirm that MS is associated significantly with patient weight, one-way analysis of variance (ANOVA) comparing the weight deciles in the total cohort in terms of MS was conducted, after verifying the following assumptions: independent and random observations, normality, and homoscedasticity.

Identification of a MS threshold for sufficient image quality for diagnosis

The relationship between MS and the quality of the clinical MPI-SPECT images was determined by extracting the UC stress test SPECT images of 50 patients from the 294-patient cohort. These patients were selected so that this subcohort had a high frequency of heavy patients (48% had BMI > 30 kg/m2), who had poor quality images due to gamma attenuation (Table 2). The images were subjected to independent visual evaluation by seven senior nuclear physicians without any additional information about the patient. After examining each image, each physician scored the quality of the image. The maximal score was 100, meaning that the image quality was perfect, while the lowest score was 0, which signified an unusable image. The 50 images were then categorized into good (score > 66), intermediate (score 33–66), and poor (score < 33) image-quality subgroups. To determine the MS threshold above which the image quality is sufficient, we studied the distribution of the MS in the poor/intermediate image quality images versus the good image quality images by constructing box-and-whisker plots. The first quartile of the good-quality images was taken as the optimal MS threshold because this meant that 75% of the good-quality images had an MS count that exceeded this threshold.

Table 2 Demographic and treatment characteristics of the patient cohorts

留言 (0)

沒有登入
gif