Participants
In this retrospective study, a total of 30 patients with coronary artery disease were included from our registered trial (ChiCTR1800019741) to develop an OAT for cardiac viability 18F-FDG imaging. Another group of patients with coronary artery disease for cardiac viability 18F-FDG imaging was randomly selected from our database and approved by the ethics committee of our hospital (registration number 106). All patients signed an informed consent form before imaging with insulin loading preparation. The patients fasted ≥ 6 h before blood glucose measurement. In the development group, routine intravenous insulin (dosage (IU) = blood glucose (mmol/L) − 2) was administered approximately 24 min before 18F-FDG injection with a 10% incremental dose for diabetic patients [14]. In the validation group, intravenous regular insulin (dosage (IU) = k × [blood glucose (mmol/L) − 2] × weight (kg); where k = 0.02 and 0.023 for non-diabetic and diabetic patients, respectively) was administered approximately 20 min before 18F-FDG injection with or without 250 mL of whole milk [18].
Cardiac 18F-fluorodeoxyglucose positron emission tomography/computed tomography data acquisition and reconstruction
Cardiac PET data were acquired with a three-dimensional list mode, 200 × 200 matrix, using PET/computed tomography (Biograph mCT Flow64, Siemens, Malvern, PA, USA) with 15 and 10 min/bed approximately 90 and 50 min after 18F-FDG (3.7 MBq/kg) injection in the development and validation group, respectively. Attenuation-corrected cardiac PET images were retrospectively reconstructed with iterative TrueX (Siemens, Malvern, PA, USA; 3 iterations, 24 subsets) into 900, 360, 180, 90, and 45 s durations in the development group, although they were retrospectively reconstructed into 30 and 600 s and optimal acquisition duration in the validation group.
Qualitative and quantitative image analyses
Standardized uptake value (SUV) measurements were performed with TrueD (Siemens, Malvern, PA, USA). A volume of interest with approximately 1.6 cm diameter was placed in the right atrium to measure the mean SUV of the blood (BloSUV). Moreover, 41% maximal myocardial SUV (MyoSUVmax) was set as the cutoff value to measure the mean SUV of the myocardium (MyoSUV) [17]. Segmental uptake (SU) percent was automatically analyzed using the quantitative perfusion single-photon emission computed tomography 2012 version (Cedars-Sinai Medical Center, Los Angeles, CA, USA) and displayed on a 17-segment polar map. SUVs and SU from 900 and 600 s image durations were set as true values in the development and validation groups, respectively. Bias of SUV and SU describes the relative percentage difference of their estimated values from true values. Bias within ±0.10 was acceptable [19]. The OAT was defined to obtain more than 16/17 segments with biases within 0.10 of the minimum acquisition time.
Statistical analyses
Continuous variables are expressed as mean ± standard deviation. Categorical variables are expressed as percentages. Between-group comparisons were performed using equivalence paired t-tests and chi-squared tests as appropriate. The equivalent limitation values of the SUV of the blood pool and myocardium were set as ± 0.1 and ± 0.5, respectively. The power and significance were set at 90% and 5%, respectively. Because biases of SU were affected by counts in the volume of interest, the product of MyoSUVmax and SU and acquisition time (MSAT) was set as the independent variable. A receiver operating characteristic curve analysis was performed, and the cutoff value of MSAT was determined, with a specificity ≥ 16/17.
In the validation group, the sample size was calculated according to the bias of MyoSUV obtained from a 90-s image duration in the development group. The OAT was calculated as MSAT divided by MyoSUV obtained from a 30-s image duration. The SUV and SU of the left ventricular myocardium on the image reconstructed with optimal time were compared with those on images acquired at 600 s.