Wildstash/dental-treatment-planning-2.5k

Dental Treatment Planning Dataset (2.5k Synthetic Cases)

Synthetic dataset of 2,494 dental clinical cases for dental diagnosis, triage, and treatment-planning research.

Dataset Details

• Size: 2,494 synthetic dental cases
• Format: JSONL with structured conversations
• Synthetic: Artificially generated cases (no real patient data)
• Purpose: Training dental diagnostic AI models
• Language: English
• License: Apache 2.0

Keywords: dental treatment planning, dental diagnosis, dental triage, dentistry, oral health, clinical NLP, instruction tuning, synthetic medical data.

Release Timeline

• September 2025: original seed dataset released as Wildstash/dental-2.5k-instruct
• April 13, 2026: simplified public dataset repo published as Wildstash/dental-treatment-planning-2.5k

Dataset Processing Pipeline

Initial Data Audit

• Found 2,494 total cases in the dataset
• Placeholder patterns dominated content in earlier drafts
• Identified need for clinical enhancement to create realistic, diagnostically useful cases

Clinical Enhancement

• Demographics enrichment: age ranges, gender balance, and diverse occupations
• Condition-specific insertions:
  • caries, periodontal disease, cysts, and mucosal lesions
  • matching radiographic findings
  • urgency levels (0 = elective, 1 = moderate, 2 = urgent)
• Structured clinical narratives with realistic patient presentations

Quality Control Gates

• JSON schema validation for required fields
• Internal consistency checks for missing values, duplicate profiles, urgency assignment, and plausibility

Expert Validation Process

• Practicing dentists graded sample cases for clinical plausibility and completeness
• Structured evaluation of diagnostic accuracy and treatment appropriateness
• Cross-comparison against strong model outputs to flag inconsistencies

Data Structure

Each example is a conversation between a dentist and AI assistant:

{
  "messages": [
    {
      "role": "system",
      "content": "You are an expert dental clinician..."
    },
    {
      "role": "user",
      "content": "PATIENT: [synthetic patient description]"
    },
    {
      "role": "assistant",
      "content": "[structured dental assessment]"
    }
  ]
}

The assistant response is structured to include:

• diagnosis
• etiology
• urgency
• management plan
• antibiotic recommendation
• follow-up
• patient counseling
• guideline source

Patient Case Format

Each case includes:

• Demographics: age range, gender
• Chief Complaint: primary symptom
• History: relevant medical and dental history
• Examination: clinical findings
• Radiographic Findings: x-ray descriptions
• Medical History: relevant conditions and medications

Intended Uses

• Model Training: fine-tuning language models for dental diagnosis
• Research: comparative studies of dental AI systems
• Education: training materials for dental students
• Development: testing dental AI applications

Files

• Dental_Cases_TrainingData.jsonl: main dataset file
• README.md: dataset card and usage notes

Limitations

• Synthetic Nature: not based on real patient cases
• Coverage: limited to general dentistry
• Validation: expert-reviewed but not clinically tested
• Bias: may reflect generation biases

Ethical Considerations

• No PHI: contains no protected health information
• Synthetic: artificial scenarios only
• Educational: intended for learning and research
• Transparency: clearly labeled as synthetic

Citation

@dataset{dental-training-dataset,
  title={Dental Treatment Planning Dataset (2.5k Synthetic Cases)},
  author={Arnav Salkade},
  year={2026},
  url={https://huggingface.co/datasets/Wildstash/dental-treatment-planning-2.5k}
}

Contact

For questions or issues: itsarnavsalkade@gmail.com

License

Apache License 2.0