Develop Software IVD

Challenges

• Research prototypes still need to mature in various ways to meet the quality and efficiency requirements for routine use.
• Marketing AI apps as in-vitro diagnostic medical devices requires fulfilling complex regulatory requirements.
• For routine use, AI apps must be integrated into heterogeneous laboratory IT infrastructures and clincal workflows.
• To ensure reliability and gain regulatory approval, the real-world performance of AI apps must be comprehensively and meaningfully validated.

Our Approach

• We apply professional software development practices to turn prototypes into a product-ready state.
• We have many years of experience in operating a quality management system according to the EN ISO 13485 (Medical Devices) standard with a special focus on implementing a software development process in compliance with IEC 62304.
• We know how to make pathology AI apps interoperable with software from other manufacturers and support diagnostic tasks.
• We have expertise in compiling representative and unbiased test datasets to measure the real-world performance of AI apps.

Transfer prototypes into marketable products

• Despite thousands of published research papers on applications of AI in pathology, only few research implementations have matured into products for routine use.

• We conducted a comprehensive survey of the technological, business and regulatory challenges of bringing AI apps for pathology to market and of potential solutions.

• We published a guide that assists researchers, software companies, and clinicians in turning prototypes into commercially viable products [1].
   

References

[1] Artificial Intelligence in Pathology: From Prototype to Product. Homeyer et al. 2021. article

Integration into the laboratory

• Pathology IT infrastructures are highly heterogeneous and based on a multitude of proprietary software interfaces and file formats.
• We co-developed the EMPAIA App Interface, the first open standard for integrating AI apps into pathology IT infrastructures [1 , 2].
• We published a comprehensive review on the individual experiences of laboratories with digitizing pathology workflows, covering digital case management, digital slide reading, and computer-aided slide reading [3].
   

References

[1] EMPAIA App Interface: An open and vendor-neutral interface for AI applications in pathology. Romberg et al. 2022. article

[2] The vendor-agnostic EMPAIA platform for integrating AI applications into digital pathology infrastructures. Jansen et al. 2023. article

[3] Digitization of Pathology Labs: A Review of Lessons Learned. Schwen et al. 2023. article

Validating AI solutions in pathology

• Compiling suitable test datasets that cover the enormous variability of pathological images and enable realistic performance assessments is very difficult and costly.
• We brought together a committee of pathologists, representatives from commercial AI companies, and AI researchers to develop recommendations for creating test datasets for AI solutions in pathology.
• The results help AI developers to demonstrate the utility of their products and assist pathologists and regulatory authorities in verifying reported performance measures [1].
   

Reference

[1] Recommendations on compiling test datasets for evaluating artificial intelligence solutions in pathology. Homeyer et al. 2022. article