Dataset Overview
| Dataset Name | Clinical indicator datasets for obstetrics and gynecology |
| Total Data Volume | 1,000,000+ clinical treatment samples |
| Data Type | Structured clinical records, multi-patient dialogues, ultrasound reports, laboratory results |
| Data Format | JSON / CSV structured storage to support relational database import |
| Annotation Dimensions | Thematic classification, emergency level classification, disease diagnostic codes (ICD-10), evaluation of the reasonableness of clinical recommendations |
| Coverage of pregnancy | Preconception counseling - early pregnancy - mid-pregnancy - late pregnancy - childbirth - full puerperium cycle |
| Diversity of populations | Coverage of pregnant women of different ages (18-45 years), parity, and underlying disease backgrounds |
| Quality Requirements | Records with incomplete or contradictory information are excluded by medical practitioners and above. |
Clinical scope and coverage
Antenatal emergencies
Massive vaginal bleeding Severe abdominal pain Regular contractions before term A significant reduction in the number of fetal movements Rupture of membranes
Regular maternity check-ups
NT / Tong-Sie / Big-Back-Aberration Nutrition and weight management during pregnancy Consultation on safe medication use Guidance on pregnancy lifestyle
Childbirth and postnatal
Birth signs are recognized. Assessment of the method of delivery Post-natal care Breastfeeding guidance Blanket bottom muscle/breather abs
Management of high-risk pregnancies
High blood pressure during pregnancy Pre-eclampsia period Diabetes during pregnancy Pre-tender Placental stripping. Multiple pregnancies
Pregnancy complications
The thyroid function is abnormal. Cardiovascular diseases Diseases of the respiratory system Immunological system diseases
Perinatal psychology
Antenatal anxiety Postpartum depression Fear of childbirth Emotional and stress management
Genetic and pre-conception counselling
Risk assessment of genetic diseases Carrier screening Causes of repeated abortion Pre-conception health check-ups
Detailed Description of Data Dimensions
| Field Name | Data Type | Value Range / Description |
|---|---|---|
| record_id | String(24) | The only mark on de-identified clinical records. |
| patient_anon_id | String(16) | De-identified patient identification, linked to multiple visits by the same patient |
| visit_date | Date | Date of visit 2019-01-01 to 2025-06-01 |
| age_group | Enum | 18-25 | 25-30 | 30-35 | 35-40 | 40-45 | 45-50 | 50+ |
| clinical_department | Enum | GYNECOLOGY | OBSTETRICS | FAMILY_PLANNING | ONCOLOGY | REPRODUCTIVE |
| diagnosis_category | Enum | Cervical change, cervix, cervix, ovarian cyst, ovary, uterine, membrane, pregnancy management, infertility, pelvic functional impairment, cytology, gynaecology tumours. |
| diagnosis_detail | String(200) | Standardized diagnostic description, ICD-10 code map |
| clinical_text | Text | De-identified clinical text (principal complaint, current medical history, medical examination, etc.), removed PII |
| lab_results | JSON | Key laboratory results (HPV testing, TCT, sex hormones, tumor markers, etc.) |
| imaging_type | Enum | PVS (Transvaginal Ultrasound) · TAS (Transabdominal Ultrasound) · MRI · CT · HSG (Hysterosalpingography) |
| imaging_findings | Text | The key findings of the video report, structured summary text |
| pathology_result | Text | Diagnostic findings of pathology reports, if any, including tissue type and grade |
| treatment_plan | Text | Summary of treatment programs (drugs/surgery/observation follow-up), de-identified |
| follow_up_days | Integer | Number of follow-ups and follow-up days to support longitudinal pre- and post-analysis |
| outcome_status | Enum | RECOVERED | IMPROVED | STABLE | PROGRESSED | COMPLICATION | UNKNOWN |
De-identified JSON Data Sample
{ "record_id": "OBG-20240822-00347", "patient_anon_id": "POB-0047821", "visit_date": "2024-08-22", "age_group": "35-40", "clinical_department": "GYNECOLOGY", "diagnosis_category": "[uterine fibroids]", "diagnosis_detail": "，，5.2cm", "clinical_text": "：6。Present Illness：RegLaw，7-10，1.5，。。G2P1。Physical Examination：10，。", "lab_results": { "hemoglobin": 98, "ca125": 25.3, "hpv_typing": "[negative]", "tct_result": "NILM", "hormone_panel": { "fsh": 6.8, "lh": 5.2, "e2": 185, "amh": 1.85 } }, "imaging_type": "TVS", "imaging_findings": "，，5.2x4.8cm，，，CDFI。2.1x1.8cm。0.9cm，。。", "pathology_result": null, "treatment_plan": "uterine fibroids，GnRH-aProcessing3", "follow_up_days": 90, "outcome_status": "STABLE"
}
{ "record_id": "OBG-20250314-01289", "patient_anon_id": "POB-0093214", "visit_date": "2025-03-14", "age_group": "30-35", "clinical_department": "OBSTETRICS", "diagnosis_category": "[]", "diagnosis_detail": "Diabetes during pregnancy (GDM)，[]28[]OGTT[]", "clinical_text": "：28Reg。Present Illness：G1P0，，。OGTT：5.8mmol/L，1h 11.2mmol/L，2h 9.6mmol/L，。120/78mmHg，BMI 26.4。", "lab_results": { "ogtt_fasting": 5.8, "ogtt_1h": 11.2, "ogtt_2h": 9.6, "hba1c": 5.9, "urine_glucose": "++", "urine_protein": "-" }, "imaging_type": "TAS", "imaging_findings": "，，BPD 72mm，HC 258mm，AC 249mm，FL 52mm，EFW 1350g。AFI 14.2cm。，I。。", "pathology_result": null, "treatment_plan": "+，<5.3mmol/L、2h<6.7mmol/L，2", "follow_up_days": 14, "outcome_status": "IMPROVED"
}
Other Organizer
"Record id": "OBG-20241105-00892",
"Patient anon id": "POB-0065412",
"visit date": "2024-11-05",
"age group": "25-30",
"Clinical department": "GYNECOLOGY",
"Diagnosis category": "CIN",
"diagnosis detail": "CIN2, HPV high-risk positive"
"Clinical text": "Chief complaint: HPV positive found at medical examination 2 weeks ago. Present illness: Regular medical examination TCT showed ASC-H, HPV16/18 positive. Colposcopy: 3 o'clock, 9 o'clock on the cervix. G0P0. Physical examination: slight erosion of the cervix, blood contact (-).",
"lab results":
"hpv typing": "HPV16 positive, HPV18 positive",
"tct result": "ASC-H",
"Colposcopy impression": "CIN2",
"Biopsy pathology": "CIN2 on the cervix, 3 o'clock on the cervix, 9 o'clock on the cervix."
{\cHFFFFFF}{\cH00FFFF}
"imaging type": "TVS",
"imaging findings": "Anteverted uterus, normal size, even echoes. The endometrium is 0.7 cm thick. The cervix is unknown. No anomaly in the bilateral adnexa area."
"Pathology result": "Cervical 3 o'clock: CIN2 (middle-grade squamous intraepithelial lesion), p16 is strongly positive. Cervical 9 o'clock: CIN1 (low-grade squamous intraepithelial lesion).",
"Treatment plan": "Recommend LEEP cervical conization, followed by regular TCT+HPV",
"follow up days": 180,
"outcome status": "STABLE"
♪ I'm sorry ♪
Annotation Specifications and Quality Control
Annotation Standards
- Standardization of diagnostics: use of the ICD-10-CM coding system, mapping of gynecological and obstetrical diagnoses to N70-N98 and O00-O99
- Clinical text sub-sections: chief complaint, present illness, past history, past marriage and childbearing history, physical examination, auxiliary examination, diagnosis, treatment plan — an eight-part structure
- Named entity categories: 6 major subcategories, 113 types (symptoms/signs, disease/diagnosis, examination/testing, drugs/treatment, anatomy, time/frequency)
- Relationship types: symptom–disease, examination–diagnosis, drug–treatment, disease–disease, disease–pregnancy — 5 relationships
Quality Control Process (Quality Control)
- Standard training:Training for annotators on 7 broad subject definitions, core scope, and illustrative examples
- Pre-annotation exercise:Completion of pre-annotation of thematic dimensions and presentation of theme-by-theme narratives
- Official annotation:Dual independent annotation + cross-validation, diagnostic category Kappa ≥ 0.85
- Expert Review：The quality control reviewer checks the results, and any non-qualified batch is returned for re-annotation.
Quality Metrics
≥0.90
Entity label F1
≥0.85
Inter-annotator Kappa
≥0.82
Relationship annotation F1
3 []
Quality Control System
AI Training Use Cases
AI-assisted diagnosis
Applicable models: XGBoost / TabNet / FT-Transformer. Cervical disease classification, endometrial tissue classification, and ovarian tumor differential diagnosis based on structured and textual features, supporting SHAP interpretability analysis.
Structured clinical text
Applicable models: BioBERT / BioClinicalBERT / GatorTron. NER entity identification, relationship extraction, ICD-10 automatic coding, supporting automatic extraction of structured information from unstructured clinical text.
Image-Text Multimodal
Applicable models: CLIP Medical Fit / Multimodal LLM. Ultrasound image + report joint modeling, automatic report generation, disease description generation, support for cross-modal retrieval and contrastive learning.
Prediction and follow-up
Applicable models: Cox proportional hazards model / DeepSurv. Disease progression prediction, relapse risk assessment, survival analysis, supporting longitudinal data modeling and temporal extrapolation validation.
Pregnancy management AI
Applied models: LSTM / Transformer time series models. Pregnancy outcome prediction, recommendation of delivery mode, evaluation of fetal growth, and end-to-end prediction based on examination time series data.
Questions and Answers on the Large Gynecology Model
Applied models: ChatGM-Med / HuatuoGPT / RAG-enhanced search.
Data Distribution Statistics
Distribution of Diseases
28%
Cervical change
22%
A uterus tumor.
16%
The ovarian cyst.
34%
Other (endometrial / pregnancy / tumor, etc.)
Supplementary Statistics
| Age Distribution | 18-25 years 8% · 25-35 years 35% · 35-45 years 38% · 45-50 years 14% · 50 years and over 5% |
| Distribution of Data Sources | Outpatient records 52% Hospital records 31% Medical examination records 12% Emergency records 5% |
| Image Data Coverage | 68% with ultrasound reports 15% with MRI reports 8% with CT reports 8% with pure clinical text 22% |
| Pathology Recognition Rate | Pathology confirmed in surgical cases 42% · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · |
| Follow-up Completeness | ≥ 2 follow-ups 55% 1 follow-up 28% no follow-up 17% |
Delivery and licensing
| StructuredJSON | Each clinical record is a standalone JSON document with full field dimensions to support bulk import into MongoDB / Elasticsearch |
| CSV/Excel Summary | Structured fields exported in CSV / Excel format for rapid modeling with traditional machine learning (XGBoost / TabNet) |
| NLP Package | Clinical text exported in CoNLL / BIO word-sequencing format to support fine-tuning of the BERT series |
| API Online Access | RESTful API supports filtering by disease category / age / data source to obtain de-identified data in real time |
| Licence | The commercial license includes authorization to deploy derivative models. |
Frequently asked questions
Does the dataset contain rare gynecological tumors?
Yes. A few types, such as ovarian granulosa cell tumors, uterine carcinosarcoma, and Paget's disease, are covered, but sample numbers are limited (50-200 cases of each type) and are suitable for data augmentation and small-sample studies.
Are video data and clinical text from the same patients?
Yes. The same patient_anon_id can link clinical text, laboratory tests, video reports, and pathological results, supporting multi-modal joint modeling.
Do the text data support NLP pre-training?
Yes. Clinical text is pre-processed at the sentence and word levels, and a format for BERT / RoBERTa pre-training is provided, which includes a medical-domain vocabulary.
How is consistency ensured?
Diagnosis categories are independently labeled by two gynecologists; a Kappa coefficient > 0.85 is considered qualified, and cases below 0.85 are arbitrated by the chief physician.
Have the datasets undergone ethical review?
The approval has been reviewed by the Ethics Committee, and the batch number is available.
Is longitudinal data analysis supported?
Yes. Multiple visits by the same patient are linked through a series of visits to support modeling and prognosis of disease progression.
Get This Dataset
The Obstetrics and Gynecology Clinical Dataset provides authorized business use, including complete documentation and structured clinical data, and supports AI model training and clinical decision research.
137-5502-0164 Online ConsultationQuick Navigation
Commercial Application Value
AI-Assisted Clinical System
Training in the smart diagnosis and pre-consultation system for obstetrics and gynecology, covering the detection of symptoms during pregnancy, interpretation of prenatal examination reports, early warning of high-risk pregnancies, and improving the effectiveness of treatment at the primary care level.
Perinatal Health Management Platform
The smart module embedded in maternal health management applications achieves early detection of postpartum depression and risk prediction for pregnancy complications, with a market size of over $3 billion.
Pharmaceutical Clinical Research
Supporting the generation of real-world data for clinical trials of obstetric and gynecologic drugs and medical device registration, and accelerating product development and post-market surveillance.