Atlas-based Aorta Deformation Heatmaps for Improved Characterization of Thoracic Aortic Disease.

Background: Thoracic aorta disease (TAD) is common [1,2] and can result in high-risk complications, including aneurysm and dissection. The current risk stratification paradigm for TAD relies on frequent surveillance imaging with computed tomography angiography (CTA) or magnetic resonance angiography (MRA), followed by aorta diameter measurements at guideline-based anatomical locations. However, aorta risk prediction is based on primitive aortic diameter thresholds, which are known to have poor predictive value for aortic complications [3,4]. There is a need for more reliable techniques to assess abnormal aortic anatomies that fully leverage the 3D information from MRA or CTA data. To address this limitation, we developed a framework for generating personalized aorta deformation (AD) heatmaps to visualize and quantify deviations of aortic anatomy in TAD patients from a normative aorta atlas, derived from age and sex-matched healthy control populations. We hypothesize that AD heatmaps can detect regional abnormal aorta surface area deviations and reliably differentiate between healthy and TAD aortas.

Methods: Aortic atlases were generated from 278 healthy 3D aorta segmentation volumes (females: 139, males: 139, ages 20-79) of 4D Flow-derived PC-MRA images. A total of 12 atlases were generated for 6 age groups (20-29, 30-39, 40-49, 50-59, 60-69, and 70-79 yr) and male/female sex (23 ± 5 aortas/group, range: 14-33). For each sex and age-decade combination, a 2-stage spatial registration process was used (Fig. 1).

To generate patient-specific AD heatmaps, the triangular surface mesh of the aorta atlas was spatially registered to the patient's 3D aorta volume. The normal surface area ratio (TAD-N_ratio) was calculated for each mesh triangle to generate AD heatmaps visualizing the extent of regional aortic dilatation (Fig. 2). The appropriately matched atlas was selected to derive the AD heatmaps for each patient. The group average of mean TAD to normal surface area ratios (TAD-N_ratio) was assessed in 25 TAD patients with mid-ascending aortic dilatation (diameter ≥ 4.0 cm [5]) and 25 healthy controls.

Results: AD heatmap visualizations show regions of aortic dilatation in yellow and red (ratio ≈ 2-3) and normal aorta in blue and green (ratio ≈ 1-1.5) (Fig. 3D and E). The group average of mean TAD-N_ratio was significantly higher in TAD patients compared with controls (1.76 vs 1.15, p < 0.001) (Fig. 3C).

Conclusion: The findings of this feasibility study suggest that personalized AD heatmaps based on a normative aorta atlas concept provide a reliable approach for visualizing and quantifying aorta surface area deviations from healthy control populations. The AD heatmap analysis also extends beyond diameter measurement-based TAD assessment and could reliably distinguish between healthy and TAD aortas. Future work will focus on quantifying potential variance in the normative aorta atlas and addressing the false positive AD heatmap deviations attributed to the aorta arch branches in 3D volumes.

Fig. 1: All 3D aorta volumes of each group (A) were resampled to the same voxel size = 1x1x1 mm3 (B). For the first stage registration (C), one aorta data set was randomly designated as a rigid registration target for all other aortas. The second stage registration (E) used the resultant ensemble average aorta (D) as the affine registration target for all aortas in the age and sex group. The final atlas 3D geometry (F) was then converted to a triangular surface mesh (G).
Fig. 2: The aorta atlas was spatially registered to the TAD patient's 3D aorta volume using deformable registration (A). Using the same transformation (B), the triangular surface mesh of the atlas (Fig. 1G) was transformed (C). The TAD to normal surface area ratio (TAD-Nratio) was calculated for each mesh triangle by dividing the surface area of the transformed mesh (TAD aorta) by that of the original mesh (normal aorta) (Fig. 3A). Color-coded TAD-Nratio were mapped onto the TAD patient aorta surface to generate an AD heatmap (D), visualizing the extent of regional aortic dilatation.
Fig. 3: The TAD-Nratio is defined (A) and color-coded (B). The group average of mean TAD-Nratio is significantly higher in TAD patients compared with controls (C). TAD aortas have regions of high TAD-Nratio (D), while healthy aortas appeared normal (E).