323. Evaluation of confidence intervals of time-integrated activity curve in molecular radiotherapy based on Monte Carlo method

Abstract

Purpose In molecular radiotherapy (MRT) calculating biokinetic parameters plays a crucial role to assess the absorbed dose to lesions or organs at risk. The assessment of the time-integrated activity curve is affected by fitting model, number of acquired points and sampling timing. The main purpose of this work was to introduce a methodology which is able to estimate the error associated to Time-Integrated Activity Coefficient (TIAC). Methods The Monte Carlo method is implemented in matlab and relied on the generation of a large number of data-set based on root mean square error (RMSE) associated to the fit model. Biokinetic data (time acquisition window of about 50 h) of 35 patients affected by metastatic differentiated thyroid cancer and administered with two different radionuclides (123I and 131I) were analysed, by fitting all data sets. Two classes of data sets were considered: Low Number of Points (LNP) and High Number of Points (HNP) time-activity data, consisting of 4 and 20 points respectively. The correlation between RMSE and bandwidths was also investigated. Results The bandwidths assessed ([mean-lower, mean-upper]) was [−14.3, +17.1]% for LNP and [−9.3, +10.6]% for HNP. The RMSE is demonstrated to be correlated to the bandwidths with two different power laws for LNP and HNP. For LNP the R2 was 0.98 and 0.97 for lower and upper band respectively while for HNP the R2 was 0.94 and 0.97 for lower and upper band. Conclusion The method is very simple to be implemented and it has the advantage of high flexibility when there is need to impose physical constraints to the fitting. As expected, the bandwidths decrease when considering data with a larger number of points. Moreover, having our data similar time window acquisition, the RMSE itself has been observed as a good predictor of the bandwidths as computed with the present method.