ESTRO 2024 - Abstract Book

S3914

Physics - Image acquisition and processing

ESTRO 2024

2358

Mini-Oral

White matter integration in RT planning: towards tract sparing dosimetry with seamless MRI workflow

Sophie BOCKEL, Nathan BENZAZON, François DE-KERMENGUY, Cristina VERES, Leo DAUTUN, Vjona CIFLIKU, Charlotte ROBERT, Eric DEUTSCH

Gustave Roussy, Radiation Oncology, Villejuif, France

Purpose/Objective:

Understanding the white matter (WM) anatomy has been facilitated by recent advances in neuro-imaging, such as Magnetic Resonance Diffusion Tensor Imaging (DTI). Anatomo-functional atlases of the WM tracts (WMT) and their associated functions have been elaborated based on awake surgery thanks to brain mapping through direct electrical stimulation[1,2]. This underscores the crucial importance of preserving WMT in order to maintain patients neurocognition and quality of life. At the same time, stereotactic brain radiotherapy (SBRT) has dramatically changed the outcome of brain metastasis (BM) patients. While focal control and extended progression-free survival have been achieved, decreasing the neurocognitive side effects of these patients is of major importance[3,4]. Despite these recent knowledges, WM structural connectivity is not implemented in SBRT treatment planning of BM. This can be explained by the lack of accessible resources to facilitate such an approach. Here we introduce an automated method that is easily accessible, and straightforward, enabling the visualization of WM tracts on individual patient pre-SRT imaging (MRI and CT scan) and planning dose maps, using an atlas-based approach [5].

Material/Methods:

We focused on BM patients treated for small (<1.5cm) lesion(s) with one fraction SBRT in our institution. All pre-SBRT brain images (T1 gadolinium enhanced MRI and CT scan) were preprocessed as follow: DICOM to NIFTI conversion, correction for the bias field effect using the N4ITK algorithm as implemented in Advanced Normalizations Tools for Python (ANTSPy) ecosystem [6] for MRIs, resampling on a 1 mm × 1 mm × 1 mm grid and brain extraction (HD-BET tool). Using ANTSPy, an affine registration to the ICBM 2009a nonlinear asymmetric space was performed, follow by a nonlinear registration (Symmetric diffeomorphic image registration (SyN)). The accuracy of the final registration was evaluated with DICE from cortical auto segmentation (ANTS Atropos [6] for MRI and SPM CTseg for CT scan [7]). Then, we applied the transformation matrix from affine registration and deformation field from nonlinear registration to the dose maps.

Results:

We implemented a fully Python scripted pipeline allowing the visualization of brain pre-SBRT MRI and CT scan, dose plan and WM tracts from an open-source tractography atlas in the same reference space (Figure 1 & 2). The accuracy of the registration of our model has been verified through DICE index of brain cortex (WM plus gray matter) in a