Classifying individuals with and without patellofemoral pain syndrome using ground force profiles - Development of a...

Area of Science:

• Biomechanics
• Sports Medicine
• Data Science

Background:

• Current prognostic models for patellofemoral pain syndrome (PFPS) rely on scalar predictors, limiting their ability to incorporate functional physiological data.
• Traditional modeling techniques fail to leverage the full predictive potential of functional variables in PFPS assessment.

Purpose of the Study:

• To evaluate the classification performance of a statistical model using functional ground reaction force (GRF) time-series data for PFPS status.
• To determine if functional GRF data can effectively discriminate between individuals with and without PFPS.

Main Methods:

• Thirty-one participants (17 control, 14 PFPS) performed maximal countermovement jumps recorded by force plates.
• Three-dimensional GRF profiles were time-normalized and utilized as functional predictors within a functional data boosting (FDboost) model for binary classification.
• Area Under the Receiver Operating Characteristic Curve (AUC) was employed to assess the model's discriminative ability.

Main Results:

• The statistical model achieved an average out-of-bag AUC of 93.7% using three GRF waveform predictors.
• Specific GRF patterns at different jump cycle percentages were identified as significant predictors for PFPS classification.
• Analysis revealed how variations in medial, posterior (AP direction), and vertical GRF influence the odds of being classified with PFPS.

Conclusions:

• Functional GRF variables, analyzed with FDboost, provide a clinically interpretable and highly accurate method for classifying PFPS status.
• This approach demonstrates excellent classification performance, suggesting FDboost is a valuable tool for prognostic studies involving both scalar and functional data.