Evaluation of an open-source pipeline to create patient-specific left atrial models: A reproducibility study

Abstract

This work presents an open-source software pipeline to create patient-specific left atrial models with fibre orientations and a fibrDEFAULTosis map, suitable for electrophysiology simulations, and quantifies the intra and inter observer reproducibility of the model creation. The semi-automatic pipeline takes as input a contrast enhanced magnetic resonance angiogram, and a late gadolinium enhanced (LGE) contrast magnetic resonance (CMR). Five operators were allocated 20 cases each from a set of 50 CMR datasets to create a total of 100 models to evaluate inter and intra-operator variability. Each output model consisted of: (1) a labelled surface mesh open at the pulmonary veins and mitral valve, (2) fibre orientations mapped from a diffusion tensor MRI (DTMRI) human atlas, (3) fibrosis map extracted from the LGE-CMR scan, and (4) simulation of local activation time (LAT) and phase singularity (PS) mapping. Reproducibility in our pipeline was evaluated by comparing agreement in shape of the output meshes, fibrosis distribution in the left atrial body, and fibre orientations. Reproducibility in simulations outputs was evaluated in the LAT maps by comparing the total activation times, and the mean conduction velocity (CV). PS maps were compared with the structural similarity index measure (SSIM). The users processed in total 60 cases for inter and 40 cases for intra-operator variability. Our workflow allows a single model to be created in 16.72 ± 12.25 min. Similarity was measured with shape, percentage of fibres oriented in the same direction, and intra-class correlation coefficient (ICC) for the fibrosis calculation. Shape differed noticeably only with users' selection of the mitral valve and the length of the pulmonary veins from the ostia to the distal end; fibrosis agreement was high, with ICC of 0.909 (inter) and 0.999 (intra); fibre orientation agreement was high with 60.63% (inter) and 71.77% (intra). The LAT showed good agreement, where the median ± IQR of the absolute difference of the total activation times was 2.02 ± 2.45 ms for inter, and 1.37 ± 2.45 ms for intra. Also, the average ± sd of the mean CV difference was -0.00404 ± 0.0155 m/s for inter, and 0.0021 ± 0.0115 m/s for intra. Finally, the PS maps showed a moderately good agreement in SSIM for inter and intra, where the mean ± sd SSIM for inter and intra were 0.648 ± 0.21 and 0.608 ± 0.15, respectively. Although we found notable differences in the models, as a consequence of user input, our tests show that the uncertainty caused by both inter and intra-operator variability is comparable with uncertainty due to estimated fibres, and image resolution accuracy of segmentation tools.

Keywords: Atrial imaging; Cardiac electrophysiology; Digital twins; Image analysis; Patient-specific modelling; Reproducibility.

Graphical abstract

Fig. 1

Overview of this study. Five independent users processed 20 cases at random from a pool of 50 cases. Processing was done using the CemrgApp pipeline developed for this work, which involves the creation of a simulation-ready mesh, fibres orientations, and a fibrosis map. Users processed some cases twice to test for intra-operator variability, whilst other cases were processed by two users to test for inter-operator variability. The pipeline was assessed for reproducibility at various stages: the surface area overlap, shortest distance of each points, and fibre orientations. Each of the 100 output cases (20 × 5 users) were used to run 3 electrophysiology simulations, where total activation time, absolute LAT differences, and correlation of PS map in universal atrial coordinates (UAC) were calculated.

Fig. 2

Example axial slices of input volumes. Each dataset consisted of an LGE-CMR scan (top) and its corresponding CE-MRA scan (bottom). Scans were screened to ensure the orientation is RAI (Right-to-left, Anterior-to-posterior, Inferior-to-superior). Scans were resampled to be isotropic, that is, each voxel had a resolution of $1{\mathrm{mm}}^3$.

Fig. 3

Overview of methodology to process a single MRA/LGE pair. Scans are processed in CemrgApp through a combination of embedded and external code called through docker containers. The pipeline processes the scans (a) from segmentation to a labelled mesh, which is then (b) refined using meshtool ([25]) and processed with the Universal Atrial Coordinates (UAC) docker container. The UAC docker container creates a standardised frame of reference for the mesh, and projects DTMRI fibres from an atlas onto the mesh. Finally, the user produces a fibrosis map from the LGE signal intensity. Outputs produced per case are: a labelled mesh, files with fibre orientations, and a fibrosis map. Electrophysiological simulations are then run on openCARP.

Fig. 4

Distance to closest point boxplots of the different metrics: (a) mean, (b) median, and (c) Hausdorff distance. Inter and intra-operator variability plots are shown in pink and blue, respectively. Different boxplots are presented to visualise the different structures: left atrial body (LA), left atrial appendage (LAA), as well as the pulmonary veins left superior (LSPV) and inferior (LIPV), and right superior (RSPV) and inferior Mean (a) and median (b) of the distance to closest point are overall under 1 mm.

Fig. 5

Fibrosis agreement. Inter- (left) and intra-operator (right) variability are presented by showing the different fibrosis scores. On both axes represent the fibrosis score ranging from 0 to 1. Different colours represent different thresholds of the IIR method, which is presented next to the ICC coefficient. Points close to the identity line show good agreement.

Fig. 6

Fibre orientation agreement distribution. (a) Inter- and (b) intra-operator histograms of the distribution of the absolute value of the dot product between two fibres orientations. The histograms corresponding to the different atrial layers (endocardium and epicardium) have been distinguished to show relative distributions. Values greater than $cos(22.5°)\approx 0.924$ correspond to angles between fibres between $\pm 22.5°$, these were considered in good agreement.

Fig. 7

Example of simulated Local Activation Time (LAT) Maps and Conduction Velocity histograms. Top row: comparison between operator A vs operator B for inter-operator variability. Bottom row: comparison between observation A vs observation B from one of the operators. Columns. (i) Local Activation Time Maps shows two observations of the same case, inter or intra depending on the row. (ii) LAT Comparison shows the LAT map from A with white contours, the contours of B’s LAT are superimposed in black. (iii) The distribution of Conduction Velocity (CV) is shown for both A and B. Ea, Early activation; La, Late activation; CV, Conduction Velocity; LAT, Local Activation Time.