Clinical Practice Guidelines

1. Transparent Algorithmic Logic for Pathology Detection Models

This section details the methodology employed by the Software to analyse data inputs to produce Software output recommendations.

1. Data Input and Preprocessing
   Dental images, such as X-rays or 3D scans, chosen by the health practitioner are loaded into the system. The Software preprocesses these input images to enhance quality and clarity, which might include steps like normalisation, contrast adjustment, and noise reduction to ensure that the input images are in a consistent and appropriate format for AI analysis.
2. Feature Extraction
   The Software then identifies and extracts key features from the input images relevant for pathology detection. This involves identifying specific landmarks in the tooth anatomy, such as the enamel, dentin, pulp, and periodontal ligament, and analysing these features for signs of disease or abnormality. The Software uses image recognition algorithms to distinguish between different structures within the tooth and surrounding tissues.
3. Anomaly Detection
   Based on the extracted features, the Software applies algorithmic analysis to detect anomalies that may indicate potential pathology. This process involves a comparison of the observed features against normative contrast patterns. The model has been trained on a dataset of dental images annotated with expert diagnoses, allowing it to learn the patterns and characteristics that typically indicate potential pathology.
4. Recommendation Logic
   When the Software detects an anomaly, it references a database (sourced from peer reviewed published research of dental pathologies), and their characteristics to classify the anomaly and suggest potential diagnoses. For example, if the Software detects an irregularity in the enamel’s thickness, it might suggest enamel erosion or dental caries as potential issues, based on the patterns it has learned during training.
5. Result Interpretation and Guidance
   The Software output to the health professional includes not only the potential diagnosis but also an explanation of the findings that led to this conclusion. This might include highlighting areas of concern on the dental image and providing a summary of the features that the Software identified as indicative of potential pathology. The Software output also references the clinical guidelines or best practices that inform its recommendations, ensuring that health professionals can understand the basis of any Software outputs.
6. User Interface and Interaction
   Health professionals can use the Software interface to review outputs, examine any highlighted areas of concern, and read the Software’s interpretations and recommendations. Health professionals have the option to accept, reject, or modify the Software’s recommendations based on their own clinical judgment and additional information.

2. Clinical Guidelines referenced in the Pathology Detection Model

The Software’s logic is informed by specific clinical guidelines, procedural standards, and practices that are foundational to its recommendations. For instance, when the software calculates the acceptable range for dental measurements, it references the clinical guideline defining this range, thereby ensuring transparency and reliability.

3. Accessibility of Sources

Each potential pathology detected by the Software is associated with a specific reference or set of references that be accompanied by the relevant clinical guidelines or radiographic characteristics that the Software’s logic has been informed by to produce the relevant outputs. These references are:

1. Integrated into the Software
   The Software will, upon detection of a possible pathology, display the relevant clinical guideline or study reference or otherwise make it accessible to the health professional in the form of a hyperlink embedded within the Software's user interface.
2. Included in Documentation
   The Software's accompanying documentation include a section dedicated to explaining the clinical basis for each type of potential pathology detected, citating the relevant guiding literature and guidelines.

These references allow healthcare professionals to use the Software output to make informed clinical decisions by facilitating their independent verification of the underlying clinical guidelines and evidence used to inform the Software’s logic.

4. Independence in Professional Judgment

The Software is designed to support health professionals in their decision-making process without replacing their judgment. The responsibility for any clinical decision rests with the health professional, who may consider the Software's output in their decision-making process:

1. Design for Decision Support Rather Than Replacement
   The Software aids health professionals in their decision-making process. It is a tool for decision support, not a substitute for a health professional’s judgement.
2. Implement Recommendations Based on Clinical Guidelines
   The Software provides recommendations to health professionals based on independently verifiable clinical guidelines, ensuring they are evidence-based up to the date the Software was made available.
3. Facilitate Verification of Software Recommendations
   The Software facilitates health professionals to verify the Software’s recommendations according to identified clinical guidelines.

These references allow healthcare professionals to use the Software output to make informed clinical decisions by facilitating their independent verification of the underlying clinical guidelines and evidence used to inform the Software’s logic.

5. Transparent Updates and Revisions

Recognising the dynamic nature of clinical practice and guidelines, the Software incorporates a transparent change log made available to healthcare professionals to track updates and understand their implications on the Software's recommendations against the latest research and clinical guidelines.

1. Implement and Document Updates
   Any updates to the Software is managed within our Change Management processes, which include providing detailed documentation and transparent change logs for any changes to reflect new clinical research and updated guidelines.
2. Communication Plan
   Users will be informed about any Software updates, including changes made and their implications, in accordance with our communication plan.

6. Adherence to Life Cycle Requirements for Medical Device Software

The Software adheres to the framework for the life cycle processes of medical device software (IEC 62304). This includes safety, performance and verification criteria including software planning, requirements analysis, architecture design, testing, release and change management, and maintenance.

1. Software Development Planning
   Development tasks are planned and executed according to defined releases. This includes incorporating Quality Management and Risk Management systems and processes (ISO 13485, ISO 14971).
2. Software Requirements Analysis
   This encompasses the detailed definition of software requirements, ensuring traceability, and establishing acceptance criteria for testability.
3. Software Architecture Design
   The architecture for the Software is designed, developed, and documented in adherence to best practice frameworks such as the AWS Well-Architected Framework.
4. Software Testing
   A built-in workflow covers end-to-end quality management, including integration, system, and user acceptance testing, with documented results.
5. Software Release
   Updates and changes are comprehensively documented with each release.
6. Software Configuration/Change Management
   Configuration items are meticulously documented, and changes are controlled through a robust change management process.
7. Software Maintenance
   Software maintenance activities are regularly planned and executed, including the frequent re-assessment of risks.