Training Data Governance for AI-Based AIS



Training Data Governance for AI-Based AIS

Published on 02/12/2025

Training Data Governance for AI-Based Automated Inspection Systems

The implementation and validation of Automated Inspection Systems (AIS) in the pharmaceutical industry present unique challenges, particularly regarding data governance and qualification methodologies. This comprehensive guide aims to establish best practices and step-by-step procedures for visual inspection qualification, focusing on the integration of AI technologies. We will discuss essential aspects like User Requirement Specifications (URS), Installation Qualification (IQ), Operational Qualification (OQ), Performance Qualification (PQ), and key considerations for false-reject control.

Understanding the Framework for Visual Inspection Qualification

The foundation of visual inspection qualification for AIS lies in the regulatory expectations that govern the pharmaceutical manufacturing landscape. In the US, the Food and Drug Administration (FDA) underscores the importance of ensuring that all manufacturing processes comply with Good Manufacturing Practices (cGMP). In Europe, the European Medicines Agency (EMA) and the Medicines and Healthcare products Regulatory Agency (MHRA) impose similar regulations that must be adhered to during the validation of automated inspection systems.

Before implementing an AIS, it is crucial to define a robust User Requirement Specification (URS). The URS acts as a cornerstone document that outlines functional and non-functional requirements, ensuring the system meets operational needs while maintaining compliance with industry standards. Additionally, consult regulatory guidelines such as 21 CFR Part 11 to assure compliance with electronic records and electronic signatures.

Step 1: Establishing User Requirements Specification (URS)

  • Collect Stakeholder Input: Engage with various stakeholders including production, quality control, regulatory affairs, and IT personnel to gather comprehensive input regarding requirements.
  • Define Functional Requirements: Detail features such as defect detection capabilities, types of products to inspect, and required throughput metrics.
  • Non-Functional Requirements: Include system performance criteria, environmental conditions, and compliance with specified regulations.
  • Review and Approval: Finalize the URS document and obtain sign-offs from all stakeholders to ensure complete agreement.

Installation Qualification (IQ)

Installation Qualification entails verifying that the system is installed correctly and is capable of operating according to the URS. It should document the physical and operational characteristics of the AIS. This phase typically includes activities such as hardware installation checks, software configuration validation, and verification of the environmental controls.

Step 2: Conducting Installation Qualification (IQ)

  • Documentation Review: Ensure all installation documents are complete and align with the manufacturer’s specifications.
  • System Installation: Proceed with the installation following manufacturer guidelines, ensuring proper integration with existing systems.
  • Configuration Management: Verify that the configuration settings comply with predefined requirements and specifications.
  • Initial Testing: Conduct preliminary tests to verify that the system operates as intended and captures initial data correctly.
  • Approval Process: Compile all documentation and secure approvals from the relevant stakeholders.

Operational Qualification (OQ)

Operational Qualification represents the next critical step, where the focus shifts toward ensuring that the system operates as specified under expected conditions through a series of rigorous testing protocols.

Step 3: Executing Operational Qualification (OQ)

  • Defined Test Protocols: Develop test protocols that include verification of all operational aspects, including machine settings, speed, and accuracy.
  • Execute Tests: Conduct tests in accordance with the predefined OQ protocols to ensure the system performs accurately during actual operational conditions.
  • Data Collection: Capture test results meticulously to substantiate successful completion of the qualification.
  • Evaluation and Correction: Analyse any discrepancies found during testing and implement corrective actions as necessary.
  • Final Report: Generate an Operational Qualification report detailing outcomes, analyses, and approvals.

Performance Qualification (PQ)

Performance Qualification is crucial for assessing the functionality of the AIS under varying operational scenarios, emphasizing reliability and efficiency in detecting defects as outlined in the initial requirements. One of the goals during the PQ phase is to validly assess the false reject rate which can impact productivity and operational efficiency.

Step 4: Achieving Performance Qualification (PQ)

  • Challenge Sets Development: Develop challenge sets utilizing the defect library, creating a variety of known defects that the system must identify during testing.
  • Conducting PQ Tests: Execute performance tests using the challenge sets under various operational conditions to validate that the AIS performs reliably.
  • Analysis of Results: Assess the sensitivity and specificity of the inspection systems through defect detection rates and false reject rates to determine the efficacy of the system.
  • Documentation: Document all results and prepare the Performance Qualification report, highlighting the system’s ability to meet the URS.

Continuous Monitoring and Routine Checks

After successful qualification of the AIS through IQ, OQ, and PQ procedures, continuous monitoring becomes essential to maintain compliance and efficacy in operations. This phase includes regular testing, trending of results, and managing corrective and preventive actions (CAPA).

Step 5: Implementing Routine Checks

  • Scheduled Maintenance: Create schedules for routine checks of the AIS, maintaining optimal performance and reliability.
  • Trending Analysis: Conduct periodic analysis of inspection data to identify variance trends, which may signal a need for additional training or system adjustments.
  • CAPA Management: Establish a CAPA process to address any deviations or anomalies detected during routine inspections.
  • Training Sessions: Train staff regularly on system updates and modifications to ensure compliance with evolving regulatory expectations.

Regulatory Compliance and Data Governance in AIS

The incorporation of AI into AIS constructs an additional layer of complexity regarding data governance. Ensuring regulatory compliance necessitates adherence to guidelines provided by authorities like the EMA and the FDA. Key considerations include ensuring the accuracy of data captured, the repeatability of system functions, and documentation of all operational parameters.

Step 6: Ensuring Regulatory Compliance and Data Governance

  • Data Integrity Practices: Implement practices that ensure data integrity throughout the AIS, particularly in adherence to Annex 1 and Annex 15 standards.
  • Electronic Records Compliance: Systems must comply with 21 CFR Part 11 for electronic records to ensure data security and audit trail requirements are met.
  • Training on Governance: Conduct training sessions focusing on data governance frameworks pertinent to AI in the pharmaceutical sector.
  • Annual Review of Processes: Establish a formal review process annually to ensure ongoing compliance with regulatory guidelines and updates in AIS technology.

Conclusion

The effective implementation of Automated Inspection Systems necessitates a thorough understanding of the regulatory frameworks, rigorous testing protocols, and ongoing compliance checks to ensure that all aspects of functionality meet industry standards. By adhering to a structured approach of URS, IQ, OQ, and PQ, pharmaceutical professionals can successfully navigate the complexities involved in visual inspection qualification. Furthermore, establishing robust data governance ensures adherence to the highest standards expected by regulatory authorities across the US, UK, and EU.

Through these processes, organizations not only enhance their operational efficiencies but also foster trust in their products, which is essential in today’s highly regulated pharmaceutical landscape.