Thoracic Aortic Aneurysm Risk Assessment: A Machine Learning Approach.

Area of Science:

• Biomedical engineering
• Computational biology
• Cardiovascular research

Background:

• Traditional thoracic aortic aneurysm (TAA) risk assessment relies on size, which is an unreliable predictor of complications.
• Aneurysm biomechanical function may offer better risk prediction but is challenging to assess non-invasively.
• Novel methods are needed to accurately assess TAA risk and prevent adverse events.

Purpose of the Study:

• To investigate the utility of machine learning (ML) models for assessing TAA biomechanical function.
• To correlate ML-derived energy loss measurements with clinical and genetic data.
• To develop a more reliable method for TAA risk stratification.

Main Methods:

• Biaxial tensile testing on patient-resected TAA tissue to calculate energy loss.
• Collection of clinical data including medical scans and genetic paneling as input parameters.
• Training and evaluation of four ML algorithms, including Gaussian process regression, using Matlab.

Main Results:

• The best ML model achieved a strong correlation with tissue energy loss (R² = 0.63), outperforming aortic diameter (R² = 0.26) and indexed size (R² = 0.32).
• Incorporating an echocardiogram-derived stiffness metric improved the ML model's correlative performance (R² = 0.62).
• The study included 158 patients (46% with bicuspid aortic valve) and 11 healthy controls.

Conclusions:

• Preliminary ML models demonstrate potential for predicting TAA tissue mechanical function.
• Clinical data can be utilized by ML algorithms to enhance TAA risk stratification.
• This approach offers a promising tool for improving patient management and outcomes in TAA.