Machine learning-based hemodynamics quantitative assessment of pulmonary circulation using computed tomographic...

Area of Science:

• Cardiology
• Radiology
• Medical Imaging

Background:

• Pulmonary hypertension (PH) is a severe condition affecting pulmonary circulation.
• Right heart catheterization (RHC) is the standard for assessing pulmonary hemodynamics.
• Noninvasive methods for quantitative hemodynamic evaluation are limited.

Purpose of the Study:

• To develop and validate a noninvasive method for quantitative evaluation of pulmonary hemodynamics.
• To assess the performance of a radiomic feature model derived from CTPA compared to RHC.
• To determine the accuracy and reproducibility of the radiomic model in predicting PH and hemodynamic parameters.

Main Methods:

• Patients with CTPA and RHC within two weeks were included.
• A radiomic feature model and a 2D feature model were constructed.
• Model performance was evaluated using mean squared error, ICC, AUC-PR, and Bland-Altman analysis.

Main Results:

• The radiomic feature model, integrating 2D and 30 radiomic features, showed consistency with RHC results.
• The radiomic model outperformed a 2D feature model.
• The radiomic model demonstrated good reproducibility (ICC=0.87) and accurately identified PH (AUC-PR=0.99).

Conclusions:

• CTPA-derived radiomic features offer a noninvasive approach for quantitative pulmonary hemodynamics assessment.
• This method shows potential as an alternative to RHC for evaluating PH.
• Further validation is needed before widespread clinical adoption.