Two-Dimensional Aortic Size Normalcy: A Novelty Detection Approach.

Area of Science:

• Cardiology
• Medical Imaging
• Biostatistics

Background:

• Assessing thoracic aorta (TA) normalcy is crucial for diagnosing cardiovascular conditions.
• Current methods, like the Z-score, rely on linear regression models with specific assumptions.
• Limitations exist in accurately evaluating TA diameters, especially in diverse patient populations.

Purpose of the Study:

• To develop and validate novel tools for assessing thoracic aorta (TA) normalcy using echocardiography.
• To compare a machine learning-based Q-score with the traditional Z-score for TA assessment.
• To evaluate the performance of these tools in healthy individuals and patients with specific aortic conditions.

Main Methods:

• Measured TA diameters in 1112 healthy subjects (age 5-89 years).
• Developed Z-score and Q-score calculators considering sex, age, and body surface area.
• Compared calculator performance in 198 adults with TA > 40 mm and 466 patients with Marfan syndrome or bicuspid aortic valve (BAV).

Main Results:

• Q-score demonstrated superior diagnostic performance (AUC 0.989) compared to Z-score (AUC 0.955) in patients with TA > 40 mm.
• Q-score achieved higher sensitivity (97.5%) and specificity (95.4%) than Z-score (81.3%, 93.3%).
• The prevalence of TA dilatation was significantly lower using Q-score (<4%) versus Z-score (>2) in Marfan and BAV patients (50.09% vs. 73.4%).

Conclusions:

• The Q-score, a novel machine learning tool, provides a more robust assessment of thoracic aorta (TA) normalcy.
• Q-score requires fewer assumptions about variable distribution and captures complex relationships between aortic diameters and anthropometrics.
• This approach may reduce the misclassification of TA abnormality in patients at risk for aortic aneurysm, warranting further prognostic studies.