Implant geometry as a patient-specific identifier in breast brachytherapy: leveraging electromagnetic tracking to...

Patient or treatment plan mix-ups are among the most serious patient-specific human errors in brachytherapy. However, many brachytherapy departments rely only on review by a second, independent person, which does not eliminate the risk of human failure. In this work, we developed and retrospectively evaluated an automated patient identification method based solely on the geometry of the interstitial implant in breast cancer patients.

The implant geometry is assessed using an electromagnetic tracking (EMT) system that provides real-time positional data of each catheter with sub-millimetre accuracy. The measured implant geometry is rigidly registered to the CT-based implant geometry associated with the clinical treatment plan. To quantitatively compare them, a similarity metric based on a distance-to-agreement (DTA) criterion (3-10 mm) and a pass rate threshold (50-95%) was used. The implants of 80 patients were included in the evaluation, resulting in 6400 patient-treatment plan combinations.

The classifier reliably identified patients with an area under the receiver operating characteristic (ROC) curve close to 1, highlighting an overall excellent discriminative performance. At the optimal decision threshold under the requirement of a false positive rate of 0%, it achieved a sensitivity between 94.8% and 97.5% depending on the DTA and pass rate thresholds, and an overall accuracy of 99.9%.

Interstitial implants in breast brachytherapy are virtually unique, so determining their geometry prior to each fraction is a viable option for patient identification. The EMT-based automated technique has proven to be effective in detecting patient or treatment plan mix-ups with near-perfect accuracy.