Assessment of inter-operator and inter-software variability (PMOD and CARIMAS) in MBF quantification with 13NH3 and 3D-PET

Aim: To assess the inter-operator and inter-software variability (PMOD and CARIMAS) in MBF quantification with 13NH3 and 3D-PET. Materials and methods: 22 patients with different kind of pathologies (hypertension, CAD, hypertrophic and dilative cardiomyopathy) underwent a rest and dypiridamole stress study by using 13NH3 and 3D-PET. The integrated PET/CT system used in this work was the Discovery-690 (GEMS, Milwaukee, USA). The acquisition protocol consisted in an injection of 370 MBq of 13NH3 and a 3D-PET dynamic scan with frames of: 9x10s, 6x15s, 3x20s, 2x30s, and 1x900s. 3D-PET raw data were reconstructed by using a 3D-Reprojection (3D-RP) algorithm with the following parameters: image matrix: 256, reconstructed Field-Of-View: 30 cm, Post-Filter Hanning 3D: cut-off 4.3 mm FWHM (Nyquist limit). Quantification was performed using two different softwares, a commercial one PMOD (v3.1, PMOD technologies, Zurich) and CARIMAS (v 2.4 Turku PET centre, Finland) a freely available research software. The analysis were carried out by two different experienced operators. Inter-software and inter-operator variabilities were assessed by Bland Altman plots. Reported variabilities are standard deviations. Results: For operator 1 (OP1) the mean difference was a +4.4% increase in MBF using PMOD with respect to CARIMAS. For operator 2 (OP2) it was +3.9% . The inter-software variability for OP1 was 6.4% at rest and 13.5% at stress, 7.1% at rest and 13.0% at stress for OP2. The mean difference for inter-operator analysis were zero on both software. The inter-operator variability for PMOD was 7.8% at rest and 11.4% at stress. For CARIMAS the inter-operator variability was 6.0% at rest and 8.2% at stress. Conclusions:Quantification of myocardial blood flow with 13NH3, 3D-PET and De-Grado 1-compartment model is similar between the two software, the variabilities show a marked increase at stress. Possible explanations for the mean differences could be: 1) the drawing of tissue VOIs, full thickness with CARIMAS and center line method with PMOD 2) the free positioning of the input function VOI with PMOD and the fixed position at the base of the LV cavity with CARIMAS. The increase in variability at stress can be attributed to higher heart rate and patient motion. The inter-operator and inter-software variabilities resulted limited to about 10% (average rest and stress) thus making both software reliable and robust for clinical use.