Optimizing Deep Inspiration Breath-Hold Patient Selection in Left Breast Cancer Radiation Therapy: A Rapid Screening Strategy

Purpose: The deep inspiration breath-hold (DIBH) technique for left breast cancer reduces cardiac radiation exposure but requires additional DIBH computed tomography (CT) scans to confirm patient suitability, increasing radiation dose and costs. We propose an innovative screening approach using high-performance DIBH-CT synthesis and quantitative dose prediction. Methods and Materials: A 2-stage model was first developed: (1) a synthetic DIBH-CT (sDIBH-CT) was generated from free-breathing CT (FB-CT) via an attention U-Net, and (2) the synthetic model was retrained using FB-CT and patient-specific DIBH orthogonal projections of test patients to obtain a refined sDIBH-CT. The FB-CT was then registered to the refined sDIBH-CT and the consequent deformation vector field was used to obtain the final predicted DIBH-CT (pDIBH-CT) and associated contours. A dose prediction model estimated the mean heart dose (MHD) for pDIBH-CT. Registration accuracy was quantified by the Dice similarity coefficient, while image quality was assessed using the mean absolute error. Furthermore, MHD reduction thresholds were established to stratify DIBH-suitable patients: patients with FB-MHD >4 Gy required ≥30% MHD reduction with DIBH, whereas those with FB-MHD ≤4 Gy required ≥50% reduction. The retrospective study was evaluated using 178 patients treated with left-sided breast radiation therapy. Results: The mean absolute error between FB-CT and DIBH-CT decreased from 178.49 ± 41.95 Hounsfield units (HU) to 55.19 ± 14.10 HU after generating pDIBH-CT. The Dice similarity coefficient increased from 0.50 ± 0.19 to 0.87 ± 0.05 for CTV, 0.59 ± 0.17 to 0.89 ± 0.04 for PTV, 0.79 ± 0.05 to 0.92 ± 0.01 for the heart, and 0.81 ± 0.04 to 0.96 ± 0.01 for the lungs. The accuracy of the radiation therapy strategy screening reached 86.7% (13/15), with the predicted outcomes closely matching the actual DIBH treatment results. Conclusions: We propose a rapid screening strategy for assessing DIBH suitability for patients treated with left-sided breast radiation therapy. This strategy has the potential to greatly optimize clinical workflow and reduce the burden on both patients and health care providers.