# Chest X-ray (xr/chest)

> Chest radiograph read: urgent flags, findings with boxes, device tips, CTR and interval change.

Upload a chest radiograph study as DICOM (CR/DX) or as PNG/JPEG. The run checks that the film is a chest X-ray, reads the frontal film (PA or AP) and returns: a critical / urgent / routine triage flag; the normal-or-abnormal call and a 37-finding panel (effusion, consolidation, atelectasis, cardiomegaly, nodule or mass, interstitial shadowing, fractures, free air and more); pneumothorax with side and outline; airspace opacity consistent with pneumonia with its outline; endotracheal tube, nasogastric tube, central venous catheter and Swan-Ganz catheter, each with its RANZCR position category, the endotracheal tube tip, the carina and the tip-to-carina distance in millimetres. Outlines return as DICOM SEG on your image, findings and the distance as a TID 1500 SR.

Model page: https://medrun.ai/models/xr/chest
Status: beta · version 1.0.0 · viewer: image

## Input

A chest X-ray: the frontal film (PA or AP), with the lateral if you have it. DICOM gives millimetres for the tube distance; PNG/JPEG exports work too (distances in pixels). Adults (18 years and older).

Accepted: DICOM files or folder, CD export (DICOMDIR), zip, PNG / JPEG images.

## Pricing

- `xr/chest/interpret`: $0.05 per request
- Draft report (opt-in, `options.report`): +$0.02 per run

## Result sections

- `#triage` Triage (available): Worklist priority: critical for pneumothorax, a malpositioned endotracheal or nasogastric tube, free air or pneumomediastinum; urgent for effusion, consolidation, a widened mediastinum, a malpositioned central line or a nasogastric tip off the film; routine otherwise. Every flag points to its finding.
- `#quality` Image quality (available): Projection (PA, AP or lateral) of the read film, rotation of the stored image, calibration and exposure flags; rotation, inspiration and penetration of the patient join later.
- `#findings` Findings (available): Normal or abnormal, and a 37-finding panel in Fleischner glossary terms: the 17 findings with a radiologist reference are always stated (present, absent, or for consolidation, nodule or mass and cardiomegaly also not excluded); the others are listed when present. Boxes for local findings join with the VinDr-trained detector.
- `#pneumothorax` Pneumothorax (available): Pneumothorax present or absent, with side and outline on the film. Rim size by the BTS and ACCP conventions and chest-drain detection join later.
- `#lines-tubes` Lines and tubes (available): Endotracheal tube, nasogastric tube, central venous catheter and Swan-Ganz catheter: presence, outline and the RANZCR position category (normal, borderline, abnormal; nasogastric tip not on the film); the endotracheal tube tip, the carina and the tip-to-carina distance in millimetres when the film carries a pixel spacing.
- `#pneumonia` Pneumonia (available): Airspace opacity consistent with pneumonia, with side and outline; lobe joins later.
- `#tuberculosis` Tuberculosis (planned): TB abnormality score 0–100 with a locally set threshold for screening programmes (WHO CAD), after a bacteriologically confirmed evaluation.
- `#lung-nodules` Lung nodules (planned): Nodule and mass boxes with diameter and a CT recommendation.
- `#cardiomegaly` Cardiothoracic ratio (planned): Heart and thoracic widths and the cardiothoracic ratio on PA films.
- `#effusion` Pleural effusion (planned): Effusion side and size grade (blunting, meniscus, half the hemithorax or more).
- `#edema` Pulmonary oedema (planned): Oedema severity 0–3 and the RALE score.
- `#fractures` Fractures (planned): Rib, clavicle, scapula and vertebral fractures with boxes and rib numbers.
- `#segment-anatomy` Anatomy segmentation (planned): Lungs, heart, clavicles, scapulae, mediastinum, aorta, diaphragm and spine outlines (lungs and heart already come with every read as evidence).
- `#compare` Comparison with prior (planned): Per-finding change against a prior film (improving, stable, worsening), new and resolved findings, device moves.
- `#cvd-risk` Cardiovascular risk (planned): Ten-year cardiovascular risk from the frontal film.
- `#lung-cancer-risk` Lung cancer risk (planned): Long-term lung-cancer risk and the screening suggestion.
- `#report` Draft report (available): Opt-in draft report (findings, impression, limitations) written from the structured result; every number is validated against it.

## Endpoint `xr/chest/interpret` (run)

The full read of the chest radiograph study: every film is checked and routed, the frontal film is read and the result contains all sections the study supports.

Docs: https://medrun.ai/models/xr/chest/api · OpenAPI: https://api.medrun.ai/v1/models/xr/chest/interpret/openapi.json

### Input schema

- `files` (array; required): The chest radiograph study: DICOM files, a folder, a CD export or a zip; PNG/JPEG images are read too. `file_…` ids from /v1/uploads or https URLs.
- `prior_files` (array): Optional earlier chest film of the same patient (the comparison section joins in a later version).
- `clinical_context` (object): Optional. Age is used when the DICOM header has none; never names, IDs or dates.
  - `age_years` (integer): Age (years)
  - `sex` (string; one of female, male, other, unknown): Sex
  - `indication` (string): For example "dyspnoea after line insertion" or "ICU morning film, intubated".
- `options` (object): Options
  - `report` (boolean; default false): Adds a draft chest radiograph report written from the structured result (English). +$0.02 per run.
  - `return` (array; default ["json","dicom-sr","dicom-seg"]): JSON is always returned. DICOM objects reference your original UIDs.

### Example (cURL)

```bash
response=$(curl --request POST \
  --url https://api.medrun.ai/v1/run/xr/chest/interpret \
  --header "Authorization: Bearer $MEDRUN_KEY" \
  --header "Content-Type: application/json" \
  --data '{
  "input": {
    "files": [
      "file_01K7XRCHEST000000000000000"
    ],
    "options": {
      "report": false
    }
  }
}')
REQUEST_ID=$(echo "$response" | jq -r .id)
```

## Output

The MedRun result envelope: https://medrun.ai/docs/result-envelope