P13.07.A INTEGRATION OF WHITE MATTER TRACTS IN STEREOTACTIC BRAIN RADIOTHERAPY TREATMENT PLANNING WITH A FULLY AUTOMATED ATLAS-BASED PIPELINE

Abstract BACKGROUND Magnetic Resonance Diffusion Tensor Imaging (DTI), has enhanced our understanding of white matter (WM) anatomy and facilitated the creation of anatomo-functional atlases of WM tracts (WMT). Concurrently, systemic treatments and stereotactic brain radiotherapy (SBRT) have significantly improved outcomes for brain metastasis (BM) patients, with growing concern in decreasing neurocognitive side effects. Nevertheless, WM structural connectivity has not yet been integrated into SBRT treatment planning. We present an automated, and universally accessible method that enables visualization of WM tracts on individual pre-SBRT imaging and planning dose maps through an atlas-based approach. MATERIAL AND METHODS After basic image preprocessing steps, pre-SBRT brain CTs used for RT planning were coregistered using affine then non-linear transformation to the ICBM 2009a nonlinear asymmetric space, using the Advanced Normalizations Tools for Python (ANTSPy) ecosystem. This space accommodates a probabilistic atlas of WMT, derived from a sample of 1065 subjects. The composition of the affine transformation matrix and the non-linear deformation field was applied to the initial dose map to warp it to the ICBM space. The pipeline was implemented first on a retrospective cohort of 108 BM (< 2 cm) patients treated with single fraction SBRT. The anatomical accuracy of the registration was evaluated by computing DICE metric from cortical auto segmentations. Then, to evaluate the consistency of warped dose maps, clinically relevant dose statistics for the planning target volume (PTV) and organs-at-risk (OAR) were compared between the initial and transformed dose maps. Eight patients treated with single fraction SBRT were prospectively enrolled and beneficiated from a pre-SBRT DTI tractography (TractSeg DTI tool used for WMT reconstruction). Mean and maximum doses (Dmean and Dmax) on 14 main WMT as defined by the two methods (DTI / atlas-based strategies) were compared to validate the pipeline. RESULTS When applying the final registration to our retrospective cohort, mean DICE index on brain cortex was 0.89 (range: 0.62-0.92). The mean absolute dose differences did not exceed 0.03Gy for PTV Dmean and 0.1Gy for PTV Dmax when comparing dose statistics from the initial dose map to the warped one. For the OARs, differences were of the order of a cGy for brainstem, brain-PTV, chiasm and optic nerves Dmean and Dmax. On the prospective cohort, the mean absolute differences on Dmax and Dmean delivered to 14 main WMTs were 1.43Gy (standard deviation: 1.06, range: 0.05-4.2Gy) and 0.18Gy (standard deviation: 0.11, range: 0.02-0.47Gy). CONCLUSION This method is the first fully automated pipeline for WMT visualization on pre-SBRT images without performing DTI. This technique has the potential to improve SBRT treatment by enabling WM sparing dosimetry and reducing neurocognitive side effects.