Dataset Overview
Definition：Chest X-Ray Plain Films and Reports Datasetis provided byChangsha Langhui Information Technology Co., Ltd.Large-scale chest image data collections. Cover 5 million cases of DICOM-formated chest X-rays and structured diagnostic reports, using a double-review mechanism (with independent review by two senior radiologists) to ensure the standard quality of the diagnosis.
| Dataset Name | Chest X-Ray Plain Films and Reports Dataset |
| Total Data Volume | Five million chest X-rays. |
| Data Type | DICOM Image + Text Diagnostic Report |
| Data Source | Resale (partnerships with more than 50 hospitals) |
| License Type | Sublicensing Including direct authorization from three Grade-A Tertiary Hospitals |
| Quality Assurance | Double-trial mechanism (independent examination of two senior radiologists) |
| Covered Disease Categories | Pneumonia, pulmonary arthritis, aerobics, chest fluids, tuberculosis, pulmonary emphysema, normal chest tablets |
| Organisation | Positive (PA/AP), side |
Core Data Fields
Each piece of data contains complete structured fields that cover everything from raw data to specialized diagnostics.
| Field Category | Field Count | Typical Fields |
|---|---|---|
| DICOM Image | 8 | DICOM file, image matrix size, pixel spacing, window window slot, input position, device type, exposure parameters |
| Imaging Findings | 10 | Lung texture, nodal position/size/density, real variant, plethora, gastrob, insulation abnormality, bone abnormality |
| Diagnostic Conclusion | 6 | Main, secondary, normal/unusual marking, ICD-10 code, emergency level |
| Patient Information | 4 | Age, gender, examination date, clinical department (DICOM-compliant de-identification) |
| Audit information | 3 | D/level of first instance physician, level of second instance physician ID/level of second instance physician, verification consistency marker |
Data Sample Showcase
The following is a de-identified structured JSON sample illustrating the typical data format.
{
"study_id": "CXR-2023-00567890",
"meta": {"age_group": "30-40","gender": "male","view": "PA"},
"findings": {"lung_fields": "sample value","heart": "sample value","mediastinum": "sample value","bones": "sample value"},
"diagnosis": {"primary": "CNXCN","abnormal": false},
"review": {"reviewer_1": "sample value","reviewer_2": "sample value","consensus": true}
}AI Application Scenarios
Breast disease AI screening
Five million double-curricular standard data training in-depth learning models for automatic detection and classification of common chest anomalies such as pneumonia, pulmonary knots, aerobic chests, chest cavity fluids, etc.
Pulmonary Flysy Smart Test
Large-scale chest-chip data training in pulmonary knot detection models, through nodal positioning, size measurement, density classification and benign risk assessment, to support early screening for lung cancer.
Emergency chest tablets are rapidly distributed.
The development of an automated system for the distribution of emergency chests based on the abnormality detection model, which quickly identifies acute conditions such as aerobic chests and large amounts of chest fluids, improves the efficiency of emergency care.
Normal/Abnormal Automatic Classification
Training of a normal/abnormal 2 classification model for large-scale primary screening of chests and automatic filtering of normal films at the medical examination centre to reduce the workload of radiologists.
Auto-generated chest report
Image-report pairing data are used to train auto-generated models for chest report, to achieve the structured extraction and auto-writing of diagnostic reports as seen in the images.
Radiomass Control AI
The consistency analysis of double-reviewed data is used to train the ACMI model, automatically tag images of inconsistent diagnosis by two physicians, and to support quality control management.
Frequently asked questions
Need for customised medical data programs?
Changsha Langhui Information Technology Co., Ltd. The DataAssetsAPI platform is dedicated to providing high-quality, compliant medical data assets for AI companies and scientific research institutions. Cooperate with 50 + hospitals, including Fudan 100 type hospitals.
Data Sample Preview · Customized Data Solutions · Professional Technical Support