Proximal peri-coronary adipose tissue radiomics score partially mediates the relationship between stenosis and coronary ischemia, and combined with stenosis degree in a random forest model achieves high efficacy in identifying FFRCT ≤ 0.75.
Key Findings
Results
PCAT Rad-Score partially mediates the impact of stenosis on ischemia in the left anterior descending artery.
Mediation analysis was performed using structural equation modeling.
The partial mediation effect of Rad-Score in the left anterior descending artery (LAD) was 11.43%.
253 vessels from 174 patients were included in the cross-sectional study.
Vessels were divided into FFRCT ≤ 0.75 (ischemia) and FFRCT > 0.75 (non-ischemia) groups.
Results
PCAT Rad-Score partially mediates the impact of stenosis on ischemia in the right coronary artery.
The partial mediation effect of Rad-Score in the right coronary artery (RCA) was 6.38%.
Mediation analysis was conducted via structural equation modeling.
The mediation effect was smaller in the RCA (6.38%) compared to the LAD (11.43%).
Results
A random forest model combining Rad-Score and stenosis degree achieved high efficacy in identifying FFRCT ≤ 0.75 in the LAD subgroup.
Area under the curve (AUC) was 0.879 in the LAD subgroup.
Sensitivity was 0.8 and specificity was 0.929 in the LAD subgroup.
The dataset was divided into a train set (70%) and test set (30%) for model construction.
Feature selection was used to derive the Radiomics Score (Rad-Score) formula from PCAT features.
Results
A random forest model combining Rad-Score and stenosis degree achieved high efficacy in identifying FFRCT ≤ 0.75 in the RCA subgroup.
AUC was 0.840 in the RCA subgroup.
Sensitivity was 0.889 and specificity was 0.778 in the RCA subgroup.
Random forest (RF) methods were used to construct both Rad-Score models and Rad-Score + stenosis degree models.
Methods
PCAT radiomics features were extracted from semiautomatic 3D segmentation and used to develop a Rad-Score.
PCAT was semi-automatically outlined using 3D Slicer software.
Radiomics features were extracted from the PCAT regions of interest.
Feature selection was applied to yield the Radiomics Score (Rad-Score) formula.
The study focused specifically on proximal PCAT relative to coronary stenosis.
Methods
A total of 253 vessels from 174 patients were included in this cross-sectional study.
Vessels were analyzed at the vessel level, not just the patient level.
Subgroup analyses were performed separately for the left anterior descending artery and right coronary artery.
FFRCT (fractional flow reserve derived from coronary computed tomography) was used as the reference standard for ischemia.