Overview
ISO 18374:2025 - Dentistry - Artificial intelligence (AI) and augmented intelligence (AuI) based 2D radiograph analysis - Data generation, data annotation and data processing defines requirements for preparing and documenting data used to develop, validate and test AI/AuI software for 2D dental radiograph analysis. The standard focuses on static (non-dynamic) AI/AuI systems and covers defining goals, limitations, target end users and the target patient population, plus the generation, annotation, handling and protection of training, validation and test data.
Key topics and technical requirements
- Data scope and quantity: Requirements for the amount and representativeness of input data (training, validation, test) to support intended clinical use.
- Quality control: Processes to ensure image quality, metadata completeness and consistency across datasets.
- Bias analysis factors: Identification and mitigation of bias sources such as patient demographics, image acquisition variability and clustered pathologies.
- Data annotation (labelling):
- Annotation strategy: Guidelines for choosing annotation granularity (image, tooth, site, pixel) that aligns with intended outputs.
- Annotation procedure and competence: Requirements for annotator qualifications, procedures and documentation of inter- and intra-annotator agreement.
- Pre-processing and post-processing: Controls for normalization, augmentation, image calibration and the risk assessment of processing steps that can affect model performance.
- Data protection and privacy: Alignment with health data security practices and controls for patient confidentiality.
- Test method and reporting: Defined test methods, inclusion/exclusion criteria, validation procedures and test report content (including electronic instructions for use).
Practical applications and who uses it
ISO 18374:2025 is intended for:
- AI/AuI software developers creating diagnostic or decision-support tools for panoramic, bitewing, periapical and cephalometric radiographs.
- Clinical researchers and dataset curators building representative, annotated dental imaging datasets.
- Regulatory and quality assurance teams assessing performance, safety and documentation for medical device submissions.
- Dental clinics and hospitals implementing AI tools and seeking guidance on dataset provenance and limitations.
- Annotation service providers who must demonstrate competence and traceability in labelling.
Practical benefits include improved dataset quality, clearer documentation of intended use and limitations, better bias mitigation and stronger foundation for regulatory compliance and clinical acceptance.
Related standards
- ISO 1942 (Dentistry - Vocabulary) - terminology used in dental standards.
- ISO 27799 (Health informatics - Information security management in health) - guidance on protecting health data used in AI workflows.
Keywords: ISO 18374, AI in dentistry, 2D radiograph analysis, data annotation, dataset generation, data processing, training data, validation data, test data, bias analysis, data protection.