Construction of gender-specific regression models for aortic length estimation based on computed tomography images

The current study aimed to construct gender-specific regression models of aortic length using anthropometric and demographic parameters available in a routine clinical setting. In addition, our goal was to assess whether these models provide more accurate estimation of aortic length than the jugulum to symphysis distance and the male-specific regression models from a previous study by Rezai et al. Gender-specific regression models of aortic length were constructed using stepwise multiple regression analysis. The actual aortic length measured from the computed tomography images of 44 subjects (19 males, 25 females) was used as a dependent variable, with age, height, weight, body mass index and jugulum to symphysis distance as independent variables. The gender-specific models with maximal adjusted R-squared values (_aR²), where variables without a significant impact on the aortic length values were removed by backward approach, were selected by stepwise multiple regression analysis and validated using leave-one-out cross-validation. The accuracy of the models was assessed by root mean square error (RMSE). The constructed male-specific regression model based on age, height and BMI explains 50% of the variance in aortic length values (_aR² = 0.4957), and the female-specific model based on jugulum to symphysis distance and BMI covered 43% (_aR² = 0.4303). The constructed models showed the lowest RMSE comparing with other assessed methods. RMSE for our male-specific model, jugulum to symphysis distance and two models by Rezai et al. were 2.97 cm, 4.58 cm, 4.58 cm, and 3.85 cm, respectively. RMSE for our female-specific model and for jugulum to symphysis distance were 2.41 cm and 4.22 cm, respectively. More accurate aortic length estimation ensures an improvement in the accuracy of pulse wave velocity measurement, which is especially important in cardiovascular risk assessment.