Clustering of shoulder movement patterns using K-means algorithm based on the shoulder range of motion
This article has been Reviewed by the following groups
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
- Evaluated articles (PREreview)
Abstract
Context
Categorization in medicine is used to enhance understanding of a disease or syndrome and apply it to treatment and is based on human clinical experience or theory. Cluster analysis using the K-means algorithm is an unsupervised machine learning method that classifies clusters based on numerical data. The purpose of this study was to classify subjects into clusters using K-means algorithm based on shoulder range of motion (ROM) and identify the characteristics of the clusters.
Design
Cross-sectional study
Methods
551 data samples measured in the 5th Size Korea Anthropometric Survey (2003∼2004) were used. Clustering was performed using the K-means algorithm, and the appropriate number of clusters was determined using the elbow curve and silhouette score. The characteristics of the clusters were analyzed by comparing the average values of shoulder ROM in the clusters.
Results
The appropriate number of classifications of clusters according to the shoulder ROM was 8. Clusters 1 and 5 had the lowest flexion range, and clusters 7 and 8 had low internal rotation and shoulder horizontal adduction ranges. Clusters 2 and 6 exhibited the highest flexion and overall high flexibility. Clusters 3 and 4 showed moderate flexion ranges but low horizontal adduction ranges. Shoulder movement patterns were classified into a total of 8 clusters according to the shoulder ROM.
Conclusion
Based on this clustering system, it was possible to identify the pattern of shoulder movement in ordinary people, and it could be used as basic data to identify and treat diseases or syndromes according to the pattern.
Article activity feed
-
This Zenodo record is a permanently preserved version of a PREreview. You can view the complete PREreview at https://prereview.org/reviews/19198817.
Short Summary of Main Findings In this 2023 medRxiv preprint, researchers applied unsupervised K-means clustering to shoulder range of motion (ROM) data from 541 young Korean adults (mean age ~24 years) drawn from a national anthropometric survey. Six ROM measures (flexion, extension, horizontal adduction/abduction, internal/external rotation) were used. The optimal number of clusters was determined as 8 based on the highest silhouette score. Distinct patterns emerged:
Clusters 1 & 5: markedly limited flexion (~143°).
Clusters 7 & 8: notably reduced internal rotation (lowest in cluster 8 at ~37°) and horizontal adduction.
Clusters 2 & 6: highest overall flexibility (flexion up to ~203°).
Clus…
This Zenodo record is a permanently preserved version of a PREreview. You can view the complete PREreview at https://prereview.org/reviews/19198817.
Short Summary of Main Findings In this 2023 medRxiv preprint, researchers applied unsupervised K-means clustering to shoulder range of motion (ROM) data from 541 young Korean adults (mean age ~24 years) drawn from a national anthropometric survey. Six ROM measures (flexion, extension, horizontal adduction/abduction, internal/external rotation) were used. The optimal number of clusters was determined as 8 based on the highest silhouette score. Distinct patterns emerged:
Clusters 1 & 5: markedly limited flexion (~143°).
Clusters 7 & 8: notably reduced internal rotation (lowest in cluster 8 at ~37°) and horizontal adduction.
Clusters 2 & 6: highest overall flexibility (flexion up to ~203°).
Clusters 3 & 4: moderate flexion but restricted horizontal adduction.
These data-driven clusters showed similarities to established clinical movement impairment syndromes (e.g., Sahrmann's classification) and potential links to subacromial pain/impingement risk.
How This Work Has Moved the Field Forward It demonstrates a purely objective, machine-learning-based approach to classifying shoulder movement patterns in the general population without relying on symptomatic or theoretical categorization. This provides quantitative baseline profiles of shoulder ROM variability and offers a foundation for future studies linking specific clusters to shoulder pathologies, risk stratification, or personalized rehabilitation in physical therapy and sports medicine.
Major Issues
Still an unreviewed preprint with no identified peer-reviewed publication.
Cross-sectional data from healthy young adults (low disease prevalence); no actual shoulder disorder diagnoses or clinical outcomes, so clusters remain descriptive only.
No statistical testing of differences between clusters (e.g., ANOVA); interpretations rely on mean comparisons.
Limited generalizability (Korean population, young age group, data from 2003–2004).
Minor Issues
Elbow curve showed no clear inflection point; reliance on silhouette score alone for choosing 8 clusters.
Age and gender were collected but not incorporated into the clustering model.
Clinical implications discussed speculatively without supporting longitudinal or patient data.
Competing interests
The author declares that they have no competing interests.
Use of Artificial Intelligence (AI)
The author declares that they did not use generative AI to come up with new ideas for their review.
-
