We evaluated 504 patients treated with breath hold. Moderate deep inspiration breath hold was audio-guided. Medial and lateral large tangential field segments were delivered in a single breath hold and movieloops of these fields were acquired with an EPID. The thoracic wall position was automatically detected in each frame and the full range of thoracic wall motion (RTWM) was determined. If the RTWM >4 mm more than 3 times, the patient was excluded from breath hold treatment if further coaching did not yield improvement.
<h4 id="absSec_3">Resultsh4>Unstable breath hold was observed in 2.8% of the patients. However, this frequency dropped from 9.5% in the first 6 months to 1.6% in the subsequent 16 months. The 97% of patients with proper breath hold showed excellent stability: the average RTWM was 0.9 ± 0.5 mm. The reproducibility of the breath hold depth was confirmed by (1) the small difference between the thoracic wall positions in the medial and lateral fields within one fraction and (2) the setup errors of breath hold patients showed no significant differences with those of right-sided breast patients.
<h4 id="absSec_4">Conclusionsh4>We have developed and clinically applied an imaging tool to automatically determine stability of breath holds in each treatment fraction during beam delivery.