Beyond Static Assessment: A Proof-of-Concept Evaluation of Functional Data Analysis for Assessing Physiological...

Area of Science:

• Sports Science
• Biostatistics
• Physiology

Background:

• Traditional recovery metrics overlook temporal continuity and individual differences.
• Physiological recovery analysis often uses discrete, time-point-independent data.
• Interindividual heterogeneity in recovery is masked by conventional methods.

Purpose of the Study:

• Assess Functional Data Analysis (FDA) for characterizing individual response dynamics post-functional threshold power (FTP) test.
• Evaluate FDA's efficacy in classifying physiological recovery phenotypes.
• Explore FDA's ability to capture temporal continuity and interindividual heterogeneity in athlete recovery.

Main Methods:

• Collected physiological time-series data (lactate, heart rate, blood pressure, glucose) from 21 trained cyclists.
• Represented data as functional objects using FDataGrid without basis expansion or smoothing.
• Employed unsupervised (K-means, Fuzzy K-means) and supervised (KNN, functional QDA) classification models with cross-validation.

Main Results:

• Lactate and diastolic blood pressure showed significant discrimination across classifiers.
• Heart rate had modest discriminative value; glucose showed intermediate performance.
• Unsupervised analysis revealed distinct lactate recovery profiles and graded membership for hemodynamic/metabolic variables.

Conclusions:

• Functional Data Analysis (FDA) provides a feasible and informative method for classifying recovery phenotypes.
• FDA preserves the temporal structure of physiological data, highlighting individual response dynamics.
• Findings are promising but require external validation in larger cohorts before clinical application.