Overview

Schema Version

https://atlas.rsna.org/schemas/2025-11/dataset.json

Name

Preoperative AP pelvic/hip radiographs, Mayo Clinic 1998-2018

Link

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9745445/

Indexing

Keywords: total hip arthroplasty, hip dislocation, anteroposterior pelvic radiograph, preoperative imaging, survival analysis, multimodal machine learning

Content: MK

SNOMED: 157265008

Author(s)

Bardia Khosravi

Pouria Rouzrokh

Hilal Maradit Kremers

Dirk R. Larson

Quinn J. Johnson

Shahriar Faghani

Walter K. Kremers

Bradley J. Erickson

Rafael J. Sierra

Michael J. Taunton

Cody C. Wyles

Organization(s)

Mayo Clinic

License

Contact

Corresponding author: Cody C. Wyles, Mayo Clinic, 200 First St SW, Rochester, MN 55905. Email: ude.oyam@ydoC.selyW

Funding

Supported by the Mayo Foundation Presidential Fund and the National Institutes of Health (R01AR73147 and P30AR76312).

Ethical review

Health Insurance Portability and Accountability Act compliant; Institutional Review Board approved with waiver of informed consent.

Comments

Retrospective single-institution cohort linking preoperative anteroposterior pelvic/hip radiographs and clinical variables to 5-year dislocation outcomes after primary total hip arthroplasty.

Date

Published: 2022-10-05

References

[1] Khosravi B, Rouzrokh P, Maradit Kremers H, et al.. "Patient-specific Hip Arthroplasty Dislocation Risk Calculator: An Explainable Multimodal Machine Learning–based Approach". Radiology: Artificial Intelligence. 2022-11-01. doi:10.1148/ryai.220067. PMID: 36523643. PMCID: PMC9745445.

Dataset

Motivation

To develop and evaluate a multimodal machine learning pipeline that predicts 5-year dislocation risk after primary THA using preoperative AP radiographs and clinical data, improving upon a prior clinical-only calculator.

Sampling

Retrospective cohort from Mayo Clinic total joint arthroplasty registry, surgeries 1998–2018; excluded patients without digitized preoperative AP pelvic/hip radiographs.

Partitioning scheme

10% holdout test set (n=1718 patients) with remaining data used for 10-fold patient-level cross-validation, stratified by dislocation incidence.

Missing information

Exact image file formats and full demographic breakdown not specified.

Relationships between instances

Multiple preoperative AP pelvic/hip radiographs per patient (median 4 images per patient); temporal interval to surgery recorded and used for weighting/feature input.

Noise

Outcome is rare (2% 5-year dislocation), leading to class imbalance.

Confidentiality

De-identified imaging; HIPAA compliant.

Re-identification

Images and data were de-identified; IRB waiver of consent obtained.

Sensitive data

Clinical variables include demographics, comorbidities, and surgical characteristics; PHI removed before analysis.