Phenotyping TPF via Self-Supervised Learning: A Label-Agnostic Framework with Expert Validation

2026-06-15

AI Summary

This study addresses the substantial inter-observer variability inherent in conventional tibial plateau fracture classification systems—such as Schatzker and AO/OTA—which can cause supervised models to learn human disagreement rather than true morphological patterns. To overcome this limitation, the authors propose the first fully unsupervised self-supervised learning framework that leverages knee radiographs to automatically discover imaging-derived fracture phenotypes. The approach integrates RadImageNet-pretrained ResNet-50, SimCLR contrastive learning, UMAP dimensionality reduction, and k-means clustering. Clinical expert blind review confirmed that the four identified phenotypes exhibit high cohesion (silhouette coefficient: 0.511; bootstrap-adjusted Rand index [ARI]: 0.319) and strong clinical interpretability. Notably, one phenotype was consistently interpreted as comminuted fracture and proved nearly orthogonal to Schatzker classification (ARI = 0.013), revealing a critical morphological dimension overlooked by existing taxonomies.

Abstract

The full potential of artificial intelligence in tibial plateau fracture characterisation remains unrealised, constrained by a fundamental dependency on labelled datasets whose consistency cannot be guaranteed: conventional classification schemes such as Schatzker and AO/OTA suffer from inter-observer variability, causing supervised models to learn human disagreement rather than stable fracture morphology. We design, implement, and validate a label-agnostic framework that eliminates this constraint by learning fracture representations directly from imaging data without observer-assigned labels. A RadImageNet-pretrained ResNet-50 encoder is fine-tuned on 154 cleaned knee radiographs using the SimCLR contrastive objective, preceded by a data cleaning protocol and followed by UMAP dimensionality reduction and k-means clustering to discover four imaging-derived phenotypes. Phenotype validity is assessed through a blinded expert review protocol administered to two independent clinicians. The four phenotypes demonstrate robust stability (bootstrap ARI = 0.319 +/- 0.041), strong internal cohesion (silhouette = 0.511), and coherence ratings of 3-5/5 from both reviewers under blinded conditions; one phenotype was unanimously identified as exhibiting comminution -- a high-complexity feature isolated without any supervisory signal. Inter-partition comparison against Schatzker labels yields ARI = 0.013, confirming orthogonality to conventional classification boundaries. Notably, expert reviewers anchored to established classification vocabularies perceived imaging-derived groups as heterogeneous precisely where Schatzker alignment was lowest, suggesting that Schatzker-trained perception and label-agnostic embedding geometry measure orthogonal dimensions. These findings establish label-agnostic SSL phenotyping as a reproducible and clinically interpretable complement to conventional classification.

Problem

• tibial plateau fracture
• inter-observer variability
• label dependency
• fracture phenotyping
• classification inconsistency

Innovation

• self-supervised learning
• label-agnostic phenotyping
• tibial plateau fracture
• contrastive learning
• expert validation