Predicting Hypertension Persistence in Coarctation of the Aorta: A Feasibility Study

Area of Science:

• Cardiovascular Medicine
• Medical Imaging
• Computational Biology

Background:

• Hypertension (HTN) is a frequent complication following endovascular repair of coarctation of the aorta (CoA).
• Predictors and optimal management strategies for post-repair HTN remain unclear.
• Existing treatments often fail to fully resolve this challenging condition.

Purpose of the Study:

• To develop and assess a feasibility workflow integrating statistical shape analysis (SSA), computational hemodynamics, and machine learning (ML).
• To investigate patient-specific geometric and hemodynamic factors as predictors of persistent HTN after endovascular CoA repair.
• To establish a novel approach for identifying individuals at risk for long-term hypertension post-intervention.

Main Methods:

• Utilized computed tomography angiography (CTA) data from a randomized controlled trial cohort (n=29) with paired baseline and 3-year follow-up scans.
• Employed deep-learning segmentation for patient-specific aortic geometry reconstruction and derived statistical shape modes (SSMs).
• Computed computational fluid dynamics (CFD)-based hemodynamic indices and applied a stacking ensemble ML classifier to predict HTN persistence.

Main Results:

• The integrated workflow demonstrated variable but promising predictive performance, with cross-validation accuracies ranging from 71.9% to 93.8% and AUC-ROCs from 0.74 to 0.95.
• Statistical analysis identified specific hemodynamic variables as potential biomarkers associated with persistent post-treatment HTN.
• The feasibility of combining SSA, hemodynamics, and ML for predicting HTN was supported by the findings.

Conclusions:

• The developed workflow integrating SSA, computational hemodynamics, and ML is feasible for exploring shape- and flow-related factors in post-CoA repair HTN.
• This approach holds potential for identifying key predictors of persistent hypertension, guiding future management strategies.
• Further research can refine this methodology for improved patient outcomes in coarctation of the aorta treatment.