Predicting post-lung transplant survival in systemic sclerosis using CT-derived features from preoperative chest CT...

Area of Science:

• Radiology and Imaging
• Transplant Surgery
• Rheumatology

Background:

• Survival prediction for lung transplant (LTx) recipients with systemic sclerosis (SSc) is challenging.
• Limited understanding of prognostic imaging biomarkers in this patient group.

Purpose of the Study:

• Identify novel CT-derived image features for predicting post-LTx survival in SSc patients.
• Develop and validate comprehensive prediction models integrating these features.

Main Methods:

• Retrospective analysis of 102 SSc patients undergoing LTx (2004-2020).
• Automated computation of CT-derived body composition (bone, muscle, adipose tissues) and cardiopulmonary features using 3D CNNs.
• Cox regression for survival analysis and model development; ROC-AUC for performance assessment.

Main Results:

• Muscle mass ratio, bone density, artery-vein volume ratio, muscle volume, and heart volume ratio were significant survival predictors.
• CT-derived feature models outperformed traditional clinical models.
• Integration of CT features improved 1, 3, and 5-year survival prediction accuracy (AUCs up to 0.90).

Conclusions:

• CT-derived imaging features offer significant prognostic value for post-LTx survival in SSc patients.
• Enhanced prediction models integrating CT features improve risk stratification.
• This tool aids clinicians in patient selection and management for lung transplantation.