S81 Innovative methodology to assess regional quadriceps muscle oxygenation during exercise in post-hospitalised patients with long COVID and healthy participants

Abstract
Background
In the lungs, the concept of how ventilation is distributed relative to blood flow (Q) is well established. In the muscle, how Q is distributed relative to regional muscle metabolic rate (VO2) is important to overall muscle function. We have previously validated an approach using near-infrared spectroscopy (NIRS) to investigate the heterogeneity of VO2 to Q in healthy quadriceps muscle during exercise. Using the coefficient of variation (CV) of regional NIRS-derived oxygenation index (StiO2) from optodes placed on vastus lateralis, we have shown a tight matching between Q and VO2 during exercise in healthy participants.

Aims
To evaluate the degree of local quadriceps muscle VO2/Q heterogeneity and assess the performance of a machine learning (ML) model in distinguishing muscle oxygenation patterns between long COVID and healthy participants during exercise.

Methods
Twelve patients with long COVID and seven healthy individuals (age mean(SD): 54(9) and 57(10), respectively) undertook 4-min constant-load graded exercise bouts sustained at 20, 50 and 80% of peak work rate. Four pairs of NIRS optodes were placed on the vastus lateralis muscle. Regional muscle VO2/Q heterogeneity was calculated as the CV of StiO2 during the last 30 seconds of each workload. Two-way ANOVA compared muscle VO2/Q heterogeneity between cohorts across workloads. A linear regression ML model with a 5-fold cross-validation was implemented to distinguish raw StiO2 data between healthy and long COVID participants using each NIRS channel and workload as features; performance metrics included Cohen’s Kappa, F1 Score, Sensitivity, Precision, Accuracy, and Receiver Operating Characteristic Area Under the Curve (ROC-AUC).

Results
Regional quadriceps muscle VO2/Q heterogeneity did not differ (p=0.95) between cohorts across exercise workloads (figure 1a). The ML model demonstrated moderate discriminatory ability (AUC=0.71) between groups (figure 1b).

Conclusion
Regional quadriceps muscle VO2/Q heterogeneity analysis suggests that this cohort of long COVID patients does not exhibit impaired local muscle oxygen supply relative to metabolic demand across a range of exercise intensities. The ML model demonstrated a moderate ability to distinguish raw muscle StiO2 signals between the two cohorts, potentially presenting a novel approach to investigate the pattern of locomotor muscle oxygenation response to exercise in long COVID.