Diagnosing Sarcopenia with AI-Aided Ultrasound (DINOSAUR)-A Pilot Study

Area of Science:

• Gerontology
• Medical Imaging
• Artificial Intelligence

Background:

• Sarcopenia negatively impacts surgical outcomes, increasing complication and readmission risks.
• Current sarcopenia diagnosis lacks muscle quality assessment and validated ultrasound cut-offs.
• Artificial intelligence (AI) can enhance ultrasound's objectivity in assessing muscle quality.

Purpose of the Study:

• To evaluate AI-aided ultrasound for reliable muscle quality assessment.
• To determine the fidelity of AI-ultrasound in diagnosing sarcopenia in surgical patients.
• To establish validated cut-offs for sarcopenia diagnosis using ultrasound-derived parameters.

Main Methods:

• Prospective cohort study of 36 surgical patients diagnosed with sarcopenia per AWGS 2019 guidelines.
• Rectus femoris muscle ultrasound analyzed by AI software (MuscleSound®) for intramuscular adipose tissue (IMAT) index.
• Receiver operative characteristic (ROC) curve analysis for diagnostic accuracy; intra- and inter-rater reliability assessed.

Main Results:

• AI-derived IMAT index showed an Area Under the Curve (AUC) of 0.727 for sarcopenia diagnosis.
• An optimal IMAT index cut-off of 4.827%/cm² was identified.
• Excellent intra-rater (ICC=0.938) and good inter-rater (ICC=0.776) reliability for IMAT index was demonstrated.

Conclusions:

• AI-aided ultrasound, specifically the IMAT index, is a reliable and reproducible tool for sarcopenia screening and diagnosis.
• The proposed IMAT index cut-off facilitates point-of-care assessment in community settings.
• Further research could integrate additional ultrasound parameters for enhanced surgical risk prediction.