Purpose: In targeted radiotherapy, there are many sources of uncertainty in the absorbed dose calculation process based on nuclear medicine (NM) images. In this work, we present a new method based on a bayesian approach to model the uncertainty of the different factors involved in the process of internal absorbed dose calculation.
Methods: A bayesian network (BN) was developed by identifying the relationships between the different variables involved in the evaluation of absorbed dose using the patient's images. This network is composed of four parts: (I) absorbed dose calculation, (II) volume estimation, (III) biokinetics determination, (IV) image processing. The BN structure was approved by clinical experts. The applicability of the BN was tested by calculating the probability distribution of absorbed dose in six lesions (thyroid metastases in the lungs) from hybrid data (4 planar images and one SPECT/CT) collected for a patient treated with iodine-131. The average doses obtained by the BN for the six lesions were compared with values determined without uncertainty from the same data. In addition, the coefficient of variation (COV) was calculated for the all lesions.
Results: The relative differences between the average absorbed doses obtained by the BN and the values determined without uncertainty were less than 0.11 for all lesions. The COV was between 0.35 and 0.40 for the 6 lesions. These results show that uncertainties in the calculated dose are important in the case of thyroid metastases in the lungs.
Conclusion: For the first time, a Bayesian network has been developed to evaluate the uncertainties on the absorbed dose in targeted radiotherapy. The study presented in this work demonstrates the effectiveness of bayesian networks in this field of application. Moreover, the results obtained for one patient show that the uncertainties can be quite large.
Dosimetry, Bayesian Statistics, Targeted Radiotherapy
IM/TH- Radiopharmaceutical Therapy: Dose estimation: MIRD/deterministic