Creative Commons source reading · Data quality and algorithm evaluation
A hybrid deep learning framework combining CTGAN and transformers for personalized injury mitigation in elite sports
Teng T; Zhang L
Originally published in Front Physiol · 2026-10-07
From the article
Selected source text. JournalOS removes markup and shortens excerpts where indicated. The full article provides the complete context.
Abstract
AbstractIntroductionOptimizing training in elite sports while managing injury risk requires models that can learn from heterogeneous and imbalanced athlete records without contaminating evaluation data. This study presents PTI-GANFormer (Personalized Training and Injury Risk Prediction using Generative Adversarial Network-Former), an integrated decision-support framework. Its contribution is not CTGAN or self-attention in isolation; rather, it is the leakage-controlled combination of training-only conditional augmentation, three functionally separated attention pathways, five domain-engineered indicators, and a transparent risk-to-action recommendation layer.MethodsAfter median imputation, plausibility-aware outlier handling, categorical embedding, scaling, and feature engineering, the data were stratified into training and untouched…
Result and discussion
This section describes the results of the experiment performed using the developed PTI-GANFormer framework for predicting injury risk among elite athletes. The evaluation of model performance has been done based on class distribution, classification metrics, confusion matrix analysis, ROC curve analysis, and performance comparison analysis. The outcomes have been analyzed to provide evidence on how effective the developed model can be in the classification of the risk levels for injury and prediction in training optimization.
Discussion
The results indicate that training-only CTGAN augmentation, five domain-engineered indicators, and complementary attention representations can improve classification under the present dataset and evaluation protocol. The contribution is integrative rather than a claim that CTGAN or self-attention is new. Same-split comparisons, corrected paired tests, confidence intervals, low-level ablations, and feature sensitivity analyses provide stronger evidence than point metrics alone. The constrained recommendation engine reduced simulated safety-constraint violations relative to alternative mappings, but this retrospective result does not demonstrate clinical effectiveness. The model should therefore be used to support, not replace, decisions by coaches, sports physicians, and physiotherapists.
Editorial reading guide
- What data created and tested the model, and was evaluation independent?
- What is the cost of a false positive, false negative or out-of-distribution decision?
