Electromagnetic tracking for integrated quality assurance in breast interstitial brachytherapy: The impact of registration method

Purpose:
Electromagnetic tracking (EMT) has great potential as a quality assurance tool in interstitial brachytherapy. Since its clinical application in most cases comprises a comparison with brachytherapy plan data, EMT registration and plan data are crucial. Registration uncertainties influence EMT outcomes and further decision-making processes. The impact of registration using four algorithms on the resulting data was investigated.

Material and methods:
Registration abilities of four sub-algorithms, which share the same principle of an iterative closest point approach were considered. Sub-algorithms differ regarding underlying mathematical norms, and also include a non-geometric implementation focusing on stochastic optimization. This study focused on the registration of point clouds originating from data obtained during treatment with interstitial brachytherapy in breast cancer patients. Data were acquired from a hybrid device prototype, capable of EMT measurements and data extracted from treatment plan. Main criteria for good registration performance were good geometric agreement, including low Euclidean distances between registered data sets as well as short registration time. In addition, previously published patient data with different than employed in the original publication algorithm were re-analyzed to further investigate the impact of registration method on clinically influential results.

Results:
The stochastic algorithm presented much higher registration times, and earliest implementation of geometric approaches occasionally misaligned patient data. The remaining two algorithms performed almost identically resulted in sufficiently accurate registration performance. The re-analysis of previously published data demonstrated minor inconsistencies between former registration method and different algorithm; however, the overall trends with almost identical mean Euclidean distance of 2.22 mm (2.19 mm in the original work) could be reproduced.

Conclusions:
The choice of registration algorithm can influence the quality of registration. Therefore, it is advised to regularly examine the registration performance, especially in in-house programs.