1267 - Investigation of Intra-Fractional Tumor Stability with AI-empowered Target Decomposition Technique in Lung DIBH SBRT Treatment
Presenter(s)
B. Huang1, W. Cai2, W. Harris2, P. Zhang2, N. Shaverdian3, J. J. Cuaron1, L. I. Cervino2, J. M. Moran2, T. Li2, Y. Fu2, and X. Li2; 1Memorial Sloan Kettering Cancer Center, New York, NY, 2Department of Medical Physics, Memorial Sloan Kettering Cancer Center, New York, NY, 3Department of Radiation Oncology, Memorial Sloan Kettering Cancer Center, New York, NY
Purpose/Objective(s):
In lung stereotactic body radiotherapy (SBRT), deep inspiration breath-hold (DIBH) is a frequently employed motion management technique used to reduce respiratory-induced tumor motion and spare the organ at risk and healthy tissue. While external surrogates are often used to monitor breath-hold consistency, the tumor's internal stability within and between breath-holds remains unclear. Previously, we developed an AI-empowered target decomposition technique for real-time markerless lung tumor tracking using triggered X-ray projection images in our clinic. This study aims to quantify intrafractional tumor motion stability with this technique during lung DIBH SBRT.Materials/Methods:
A retrospective single-institution analysis was conducted of a cohort of 23 lung cancer patients undergoing DIBH SBRT. Internal tumor motion was quantified using our target decomposition technique with a patient-specific deep learning model that transforms raw triggered kV projection images into synthetic decomposed target images (DTIs) with enhanced tumor contrast. Markerless tumor tracking was performed via template matching between the DTI image and the reference digitally reconstructed radiograph (DRR) generated from the region of interest in the planning CT images. A total of 23 patients with 80 treatment fractions and 549 individual breath holds were analyzed. Stability was evaluated by measuring the Superior-Inferior (SI) and in-plane left-right (IPLR) tumor excursion within each hold (intra-BH) and the systematic drift between sequential holds (inter-BH).Results: For all patients in this cohort, the intra-BH tumor motion ranges were 1.8 mm ±1.2 (std) mm in the SI direction and 1.4mm ± 0.9 (std) mm in the IPLR direction. Among all individual breath holds, the percentages of tumor displacement within a 5mm clinical margin were 88.9% and 97.8% in the SI and IPLR directions, respectively; the percentages of tumor motion amplitudes exceeding 3mm and 5mm within each breath hold were 19.3% and 9.5%, respectively. Furthermore, the inter-BH tumor drifts within each treatment fraction were 1.7 mm ± 2.0 mm in the SI direction and 1.5mm ± 1.5 mm in the IPLR direction were observed
Conclusion:
Our results demonstrate that even under DIBH, lung tumor motion can exceed the 5mm clinical margin, and direct markerless tumor tracking, rather than using an external surrogate, is essential for high-precision lung SBRT treatment, especially if a further treatment margin reduction is desired.