Session Item

Precise target definition is a keystone of the promising cardiac radioablation (CR) technique for the treatment of ventricular tachycardia (VT). This step of treatment planning relies on multimodal data integration (cardiac CT scans, electro-anatomical mapping (EAM), PET...), as many anatomic and functional information must be exploited to locate the arrhythmogenic substrate. CR target definition is a difficult task, and prone to imprecisions due to the numerous data integration processes and to visualization limitations.

The objective of this work was to propose a workflow for multimodal data integration to improve the robustness of CR target definition.

A target definition workflow was developed to generate a 3D mesh on which were fused all descriptors of interest extracted from multimodal images of a given patient.

The left ventricle and myocardium were automatically segmented from cardiac Computed Tomography (cCT) image using a deep learning approach. The 3D mesh of the left ventricle was then built using the marching cubes algorithm.  Myocardium thickness was computed on each point of the cCT mesh, as its distance to the closest non-ventricle, non-myocardium point.

Other datasets were then registered to the cCT, in order to project useful multimodal information on this mesh.

EAM data was exported from the CARTO system as a mesh and a list of measurements. The EAM mesh was automatically registered to the cCT mesh using ICP registration algorithm.

PET information, if available, was also integrated into the workflow through registration of the associated CT image to the cCT, and reporting of the PET values at the coordinates of each cCT mesh point.

These processes allowed the representation of the information on one unique mesh. A tool was integrated to delineate the target by geodesic path picking on its surface. The resulting target was finally propagated to the whole width of the myocardium in order to generate the clinical target volume which can be directly exploited by treatment planning systems.

Examples of one patient data representations are illustrated in figure 1. The path picking process is presented in figure 2.

Our workflow was tested on the data of 4 patients. Cardiologists using this workflow reported being more confident about their target definition than when using standard treatment planning tools. The three-dimensional representation especially was considered helpful to identify the optimal target zone.

All operations were applied automatically with potential manual corrections. It represented a gain of time, compared to a slice-by-slice delineation on cCT, that was evaluated up to 50%.

The developed workflow enables to fuse multimodal information to improve the robustness of target definition in CR of VT. Preliminary experiments show very positive feedback from clinicians and a decrease of delineation time. Further studies will be needed to confirm these results on a larger cohort of patients.