Sarcopenia Assessment Using Fully Automated Deep Learning Predicts Cardiac Allograft Survival in Heart Transplant...

Area of Science:

• Cardiology
• Radiology
• Artificial Intelligence

Background:

• Sarcopenia, low muscle mass, is linked to poor outcomes in end-stage heart failure patients.
• Manual quantification of muscle mass from CT scans is labor-intensive and prone to errors.
• Predicting heart transplant success requires reliable methods to assess patient condition.

Purpose of the Study:

• To evaluate if deep learning-based automated assessment of radiographic sarcopenia predicts outcomes after heart transplantation.
• To establish a faster and more consistent method for sarcopenia evaluation in heart transplant candidates.

Main Methods:

• Retrospective analysis of 164 adult heart transplant recipients (2013-2022).
• A deep learning tool automatically calculated skeletal muscle area from pre-transplant chest CT scans.
• Radiographic sarcopenia defined by lowest sex-specific quartile of skeletal muscle index (SMI).

Main Results:

• Radiographic sarcopenia at T11 was associated with significantly lower cardiac allograft survival (83% vs. 97% at 3 years).
• Automated sarcopenia assessment predicted increased risk of allograft loss or death (HR 3.86).
• Patients with sarcopenia experienced longer hospital stays post-transplant.

Conclusions:

• Automated radiographic sarcopenia assessment using CT scans is a viable predictor of heart transplant outcomes.
• This AI-driven approach offers a reliable, efficient alternative to manual methods.
• Improved candidate selection and patient outcomes are potential benefits for heart transplantation.