Skip to content

Bone and joint radiograph read: fractures, dislocations, bone age, osteoarthritis and alignment.

BetaCR · DXMSKPelvisRadiologyEmergencyv1.0.0 · $0.04 / request

What it does

Musculoskeletal X-ray reads any extremity, joint, hand or pelvis radiograph study, with every view of it. The run identifies the body part and projection of each image, looks for fractures and dislocations on all views together, and then adds the measurement that the region calls for: bone age on a child's hand, Kellgren–Lawrence grade on a standing knee, hip dysplasia lines on an infant pelvis, alignment on a full-length leg film.

Fractures are reported per view with a box, a study-level positive or negative answer, and AO/OTA location codes; physeal injuries in children carry a Salter–Harris type. Joint effusion and lipohaemarthrosis are reported with the fracture read because they point to fractures that are hard to see.

Intended use

Decision support for the clinician reading a trauma or joint film, most often an ED doctor, GP or orthopaedic resident. Missed fractures are among the most frequent diagnostic errors in emergency departments, and many sit on one view only: scaphoid, radial head, paediatric buckle fractures. The run reads every view and marks each fracture with its evidence.

Who it is for

Emergency physicians, orthopaedic surgeons and residents, radiologists, paediatric endocrinologists (bone age), rheumatologists (erosion scoring) and paediatric orthopaedists (DDH). Developers use the same run to add fracture detection or bone-age measurement to a PACS or clinic system.

Accepted input

  • DICOM CR or DX radiographs of any limb, joint, hand, foot or pelvis; upload the whole study, all projections.
  • Special films: left hand-wrist PA for bone age, standing AP knees, AP pelvis, full-length standing lower-limb films, and hand and foot series in rheumatoid arthritis.
  • PNG or JPEG: read, but uncalibrated, so lengths and joint-space widths are not given in millimetres.

Requirements

Requirement Why
All views of the study Many fractures are visible on one projection only
Pixel spacing and magnification factor Millimetre values; a calibration ball is detected on long-leg films, otherwise only ratios and angles are reported
Patient age or birth date Paediatric reading below 18 years: growth plates, ossification centres and Salter–Harris
Sex (for bone age) Bone-age standards differ by sex
Standing (weight-bearing) knee and leg films Joint-space width and the mechanical axis are assessed under load

Casts, implants and non-standard projections are recognised and noted.

Optional context

The clinical question (trauma mechanism, suspected scaphoid fracture, growth assessment, known RA) and prior studies can be given in clinical_context and prior_files.

The result

One JSON result with a section per question; sections that do not apply to the uploaded region are left empty.

Section Content
Fractures and dislocations Boxes per view, dislocation, joint effusion or lipohaemarthrosis, study-level positive or negative
Bone age Bone age in months with SD and z-score against chronological age
Knee osteoarthritis Per-knee KL grade 0–4, OARSI joint-space narrowing and osteophyte grades, minimum joint-space width
Hip osteoarthritis Per-hip Kellgren–Lawrence or Tönnis grade and minimum joint-space width
Hip dysplasia (DDH) Hilgenreiner and Perkin lines, acetabular index, Shenton line and IHDI grade
Leg alignment HKA angle, mechanical axis deviation, mLDFA, MPTA and leg-length discrepancy
Sharp/van der Heijde score Per-joint erosion and joint-space narrowing scores and the total score
Joint implants Implant type and manufacturer, periprosthetic lucency by zone, periprosthetic fracture
Bone lesions Lesion box or mask, benign or malignant appearance and Lodwick–Madewell grade
Bone fragility estimate Estimated T-score band from a hip, spine or chest film, with a DXA referral prompt
Draft report Opt-in (options.report): an English narrative that interprets the result for the reader (findings, impression, limitations); every number is checked against the findings

Standards

  • DICOM: TID 1500 Measurement Report for bone age, KL grade, angles and widths; GSPS overlays for boxes and lines; SEG for lesion masks; Encapsulated PDF.
  • Fracture terms: AO/OTA Fracture and Dislocation Classification Compendium (2018) and Salter–Harris.
  • Grading systems: Greulich & Pyle atlas, Kellgren–Lawrence, OARSI atlas, Tönnis, IHDI classification with age norms for the acetabular index, Paley malalignment nomenclature, Sharp/van der Heijde, Gruen and DeLee–Charnley zones, Vancouver classification, Lodwick–Madewell.
  • Bone fragility: WHO T-score categories and a FRAX referral prompt, with a referral to DXA.

Result sections

One run returns every section its input supports, at one price. Sections that cannot run say why (for example, not in input).

SectionWhat it returnsStandardsStatus

Fractures and dislocations

#fractures

Fracture boxes on every view with region and side, and a positive or negative verdict per exam (region and side). Covered in 1.0: children's wrists. Other regions, adults, patients of unknown age and views the router cannot place with confidence are reported as not assessed. AO/OTA bone and segment codes and Salter–Harris typing join in a later version.findings · classifications · artifactsDICOM SR TID 1500 (2026d)Since v1.0.0

Bone age

#bone-age

Greulich & Pyle-equivalent bone age in months from a child's left hand PA, with the z-score against chronological age (Brush Foundation SDs). Requires: A child's hand PA; sex (DICOM or clinical context) for the bone age, age for the z-score.measurementsGreulich & Pyle atlas (1959)Since v1.0.0

Draft report

#report

Opt-in draft report (findings, impression, limitations) written from the structured result; every number is validated against it.report—Since v1.0.0

Knee osteoarthritis

#knee-oa

Kellgren–Lawrence grade per knee, OARSI joint-space narrowing and osteophyte grades.classifications · measurementsKellgren–Lawrence (1957); OARSI atlas (2007)Planned · P1

Hip dysplasia (DDH)

#ddh

Hilgenreiner and Perkin lines, acetabular index and IHDI grade on an infant AP pelvis.measurements · classificationsIHDI classification (2015)Planned · P2

Bone lesions

#bone-tumors

Bone lesion box with benign versus malignant appearance.findings · classifications—Planned · P2

Validated performance

Numbers come only from MedRun's own evaluation harness (make bench (models/xr-msk)), with dataset, split and date.

SectionMetricValueDatasetDate
fracturesWrist fracture exam AUC (children)0.98795% CI 0.981–0.993GRAZPEDWRI-DX (held-out patients (Ju test)) · n=10762026-10-15
fracturesWrist fracture detection sensitivity (children)0.95395% CI 0.941–0.964GRAZPEDWRI-DX (held-out patients (Ju test)) · n=20292026-10-15
fracturesFracture image AUC (adults0.87295% CI 0.814–0.925FracAtlas (held-out test range) · n=4222026-10-15
bone-ageBone age MAE (months)4.1395% CI 3.7–4.6RSNA Pediatric Bone Age 2017 (official test) · n=2002026-10-15

Pricing

EndpointUnitPrice
xr/msk/interpretrequest$0.04
Draft report (opt-in, options.report)per run+$0.02
  • One price per study, whatever the number of views or regions.
  • Rejected inputs (for example, no radiograph in the upload) and failed runs are not charged.

Changelog

  1. v1.0.0Oct 15, 2026major
    • First MedRun build of the MSK radiograph read; fracture detection gated per region (children's wrists covered) and bone age.