Evaluating the accuracy of planar gated blood pool processing software using simulated patient studies

Abstract

Planar gated blood pool (GBP-P) radionuclide imaging is a valuable non-invasive technique for assessing left ventricular ejection fraction (LVEF). Serial cardiac imaging can be performed to monitor the potential decline in LVEF among patients undergoing cardiotoxic chemotherapy. Consequently, accurate LVEF determination becomes paramount. While commercial software programs have enhanced the LVEF values' reproducibility, concerns remain regarding their accuracy. This study aimed to generate a database of GBP-P studies with known LVEF values using Monte Carlo simulations and to assess LVEF values' accuracy using four commercial software programs. We utilised anthropomorphic 4D-XCAT models to generate 64 clinically realistic GBP-P studies with Monte Carlo simulations. Four commercial software programs (Alfanuclear, Siemens, General Electric Xeleris, and Mediso Tera-Tomo) were used to process these simulated studies. The accuracy and reproducibility of the LVEF values determined with these software programs and the intra- and inter-observer reproducibility of the LVEF values were assessed. Our study revealed a strong correlation between LVEF values calculated by the software programs and the true LVEF values derived from the 4D-XCAT models. However, all the software programs slightly underestimated LVEF at lower LVEF values. Intra- and inter-observer reliability for LVEF measurements was excellent. Accurate LVEF assessment is crucial for determining the patient's cardiac function before initiating and during chemotherapy treatment. The observed underestimation, particularly at lower LVEF values, emphasises the need for the accurate and reproducible determination of these values to avoid excluding suitable candidates for chemotherapy. The software programs' excellent intra- and inter-observer reliability highlights their potential to reduce subjectivity when using the semi-automatic processing option. This study confirms the accuracy and reliability of these commercial software programs in determining LVEF values from simulated GBP-P studies. Future research should investigate strategies to mitigate the underestimation biases and extend findings to diverse patient populations. Gated blood pool studies, left ventricular ejection fraction, Monte Carlo simulations, 4D-XCAT models.

Fig. 1

Flowchart illustrating the creation of the patient phantom database by generating GBP-P models and MC simulation of the GBP-P studies. Processing of the GBP-P studies is also shown.

Fig. 2

Sample transaxial images of the 4D-XCAT phantom's segmented activity map during the end-diastolic (ED) and end-systolic (ES) phases.

Fig. 3

Sample of MC simulated images of the ED- and ES-phase of the original 4D-XCAT model 1.

Fig. 4

Figures (a), (c), (e), and (g) present the linear regression of 4D XCAT LVEF values compared to the four processing software programs: Alfa Nuclear (AN), Siemens (SM), General Electric Xeleris (GE-X), and Mediso Tera-Tomo (M-TT). Bland-Altman analysis is shown in Figures (b), (d), (f), and (h) for the same four processing software programs: AN, SM, GE-X, and M-TT. The linear regression graphs contain the Pearson correlation coefficient (R) and the Standard Error of Estimation (SEE) for each of the software programs. The Bland-Altman graphs display a mean-, upper- and lower limit of the data.