ESTRO 2020 Abstract book

S126 ESTRO 2020

OC-0224 Fully automated heart dose calculation from cine MV images recorded during breast cancer treatments S.S. Andreasen 1 , C. Muurholm 2 , S. Skouboe 2 , H. Spejlborg 3 , B. Offersen 2,4 , P.R. Poulsen 2,4 1 Aarhus University, Department of Physics and Astronomy, Aarhus, Denmark ; 2 Aarhus University Hospital, Department of Oncology, Aarhus, Denmark ; 3 Aarhus University Hospital, Department of Medical Physics, Aarhus, Denmark ; 4 Aarhus University Hospital, Danish Center for Particle Therapy, Aarhus, Denmark Purpose or Objective Heart irradiation during breast cancer radiotherapy can lead to late cardiac morbidity. Even with deep-inspiration breath-hold (DIBH), daily heart position variations can give higher heart doses than planned. For tangential treatments, the heart position during treatment can be monitored in continuous portal images (cine MV images). The heart position is, however, only a surrogate for the heart dose, which is the clinically relevant quantity. This study develops fully automated heart dose calculation based on cine MV images. Material and Methods For ten left-sided breast cancer patients receiving DIBH radiotherapy, 7.7Hz cine MV images of the two main tangential fields were recorded at 153 fractions. No patient had internal mammary lymph node involvement. The heart edge was segmented automatically in the MV images by utilizing that pixels on the heart edge change their intensity with heartbeat frequencies. The segmentation was used to determine the daily horizontal mean shift of the heart in beam’s eye view (BEV) relative to the planned position. A patient-specific relationship between BEV heart shifts and mean heart doses (MHD) was obtained by exporting the 3D dose matrix and heart contour from the treatment planning system to an in-house developed program that calculated the heart dose for a range of left, anterior and combined left-anterior heart shifts. The dose distribution was assumed to be invariant to the heart shifts, which allowed for full automation without the need for dose recalculation for each heart shift. The relationship between BEV heart shifts and MHD was used to calculate the MHD at each fraction from the auto-detected shifts. The accuracy of the MHD calculation was quantified by comparison with more accurate manual MHD calculations in the TPS with dose recalculation with water density assigned to the shifted heart contour. Results The heart was frequently exposed in both tangential fields for six patients, who had a BEV mean heart shift relative to the plan of 5.3mm. Figure 1 shows the automatically calculated MHD for a variety of heart shifts in BEV for these patients. The relationship between BEV heart shifts and MHD was nearly independent of the 3D heart shift direction (left, anterior or left-anterior), and a single patient- specific second-degree polynomial fit was therefore used for all shifts (Figure 1). The mean RMS error of the MHD fit was 0.08Gy. The automated MHD calculation at each fraction (black bars in Figure 2) on average overestimated the MHD with 0.08Gy compared to the more accurate manual recalculation with water density assignment to the shifted heart structure (white bars). The fractions with heart exposure had an average MHD of 1.2Gy while the planned MHD was 0.6Gy.

Conclusion Fully automated heart dose calculation from cine MV images of tangential breast fields was proposed, developed and shown to be highly accurate. It may be used in a surveillance program with automated heart dose monitoring of all breast cancer treatments in a clinic.