EP55 A fully automated approach to adult hip radiographs
Bibliographic record
Abstract
Abstract Purpose Plain radiographs are an essential, and often first line, tool employed in orthopedic practice for the evaluation of a wide range of hip pathologies and anatomic features. Their use ranges from the traumatic conditions such as fractures and dislocations to more chronic conditions such as hip dysplasia, and femoroacetabular impingement. Given this ubiquity, we propose a fully automated approach, or model, to studying these radiographs, employing convolutional neural networks (CNNs). We put forth that such an approach can produce measurements comparable with a trained individual and at a much faster rate. This can be useful in the clinical setting and in research where it enables the study of large datasets. Methods Our dataset includes hip radiographs from 82 patients. Anteroposterior (AP) and modified Dunn views are included and labelled by an orthopedic surgery resident highlighting anatomy of interest. A subset of this data is then used to train convolutional neural networks to perform the same task in a fully automated fashion. Custom developed code is then used to highlight landmarks of interest (such as the centre of the femoral head) and use these landmarks to calculate measurements used in evaluating adult hip anatomy. We tested the performance of this method against orthopedic surgery residents with respect to accuracy and speed. Results We compared the LCEA measurements performed by the automated model above to those of an orthopedic surgery resident as a first step. We show a mean difference of 5.75 degrees with a median of 3 degrees. Run time is approximately 800 milliseconds per radiograph on a consumer grade laptop. Conclusion Initial testing of the model developed suggests such an approach is feasible and can produce reliable results. Next steps are to include more data in model training to improve performance and then include more measurements on more views to achieve a complete evaluation of the adult hip. We then plan to use this model for studying large data sets and extract radiographic parameters with higher accuracy and reliability, and much higher speed than is done by human experts.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.002 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.014 | 0.006 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".