MIMIC-CXR Chest Radiograph Classification

Large-scale chest radiograph classification based on MIMIC-CXR (Johnson et al., Scientific Data 2019). Multi-label classification with 14 observations derived from radiology reports.

Configuration Fileconfig.yaml

# MIMIC-CXR Chest Radiograph Classification Configuration
# Based on Johnson et al., Scientific Data 2019

annotation_task_name: "MIMIC-CXR Chest Radiograph Classification"
task_dir: "."

data_files:
  - "sample-data.json"

item_properties:
  id_key: "id"
  text_key: "image_url"
  context_key: "report_excerpt"

user_config:
  allow_all_users: true

annotation_schemes:
  - annotation_type: "multiselect"
    name: "findings"
    description: "Select all findings present in the image"
    labels:
      - name: "atelectasis"
        tooltip: "Partial or complete lung collapse"
      - name: "cardiomegaly"
        tooltip: "Enlarged heart"
      - name: "consolidation"
        tooltip: "Lung tissue filled with fluid"
      - name: "edema"
        tooltip: "Pulmonary edema"
      - name: "enlarged_cardiomediastinum"
        tooltip: "Enlarged heart and mediastinum"
      - name: "fracture"
        tooltip: "Bone fracture visible"
      - name: "lung_lesion"
        tooltip: "Mass or nodule"
      - name: "lung_opacity"
        tooltip: "Abnormal opacity"
      - name: "pleural_effusion"
        tooltip: "Fluid in pleural space"
      - name: "pleural_other"
        tooltip: "Other pleural abnormality"
      - name: "pneumonia"
        tooltip: "Pneumonia"
      - name: "pneumothorax"
        tooltip: "Air in pleural space"
      - name: "support_devices"
        tooltip: "Medical devices present"

  - annotation_type: "radio"
    name: "no_finding"
    description: "Is this a normal study?"
    labels:
      - name: "normal"
        tooltip: "No significant findings"
      - name: "abnormal"
        tooltip: "Abnormalities present"

  - annotation_type: "radio"
    name: "view_position"
    description: "What is the view position?"
    labels:
      - name: "PA"
        tooltip: "Posteroanterior"
      - name: "AP"
        tooltip: "Anteroposterior"
      - name: "lateral"
        tooltip: "Lateral view"

  - annotation_type: "text"
    name: "additional_notes"
    description: "Additional clinical notes or observations"

interface_config:
  item_display_format: "<img src='{{text}}' style='max-width:100%; max-height:500px; background:#000;'/><br/><small>Report: {{context}}</small>"

output_annotation_format: "json"
output_annotation_dir: "annotations"

Sample Datasample-data.json

[
  {
    "id": "mimic_001",
    "image_url": "https://upload.wikimedia.org/wikipedia/commons/c/c5/Normal_posteroanterior_%28PA%29_chest_radiograph_%28X-ray%29.jpg",
    "report_excerpt": "PA and lateral views of the chest. Heart size normal. Lungs are clear."
  },
  {
    "id": "mimic_002",
    "image_url": "https://upload.wikimedia.org/wikipedia/commons/2/29/Chest_X-ray_of_RTI.jpg",
    "report_excerpt": "Frontal chest radiograph. Evaluate for cardiopulmonary disease."
  }
]

// ... and 1 more items

Get This Design

View on GitHub

Clone or download from the repository

Quick start:

git clone https://github.com/davidjurgens/potato-showcase.git
cd potato-showcase/image/medical/mimic-cxr
potato start config.yaml

Details

Annotation Types

multiselectradiotext

Domain

Medical ImagingRadiology

Use Cases

Chest X-Ray ClassificationRadiology AIClinical Documentation

Related Designs

DocBank Document Layout Detection

Document layout analysis benchmark (Li et al., COLING 2020). Detect and classify document elements including titles, abstracts, paragraphs, figures, tables, and captions.

multiselectradio

DocLayNet Document Layout Analysis

Document layout analysis with bounding box annotations. Annotators draw bounding boxes around layout elements (text blocks, tables, figures, headers, footers, lists) in document page images.