April 26, 2023

Artificial Intelligence has become an integral part of our everyday lives. It is therefore not surprising that AI is also playing a growing role in medical technology and drug manufacturing. However, companies from highly regulated industries such as pharmaceutical companies and medical device manufacturers face particular challenges in this context, as computerized systems are required to operate transparently, traceably and reproducibly in such an environment (GxP). The review of such requirement is done through the process of validation. Validation provides a documented proof that a process or a system reproducibly fulfills the specified requirements under field conditions. AI-based systems are currently not able to fulfill these requirements readily, and the regulatory requirements for validation of AI/ML-based systems are still in their infancy. 

However, not only does validation of AI-powered systems promise to unleash untapped potential, but so does AI-assisted validation. In the current work of validators, AI-based tools play an inferior role. Specifying requirements, writing scenarios, and determining user stories is a time-consuming process and involves the development, control, and maintenance of extensive documents and/or long, bulky lists. At the same time, working with system requirements also puts many demands on the use of language. For example, requirements must not be ambiguous, should include roles and actors, and must also be measurable (SMART). With the current advances in Natural Language Processing (NLP), a subset of AI/ML, it is possible to significantly ease the work of validators and substantially increase the quality of specifications. Going beyond thesauri and ontologies, NLP techniques are able to understand language and examine descriptions of requirements for problematic language usage and ambiguities. This helps to ensure optimal results in the requirements engineering process, the writing of User Requirements Specifications (URS), and to keep requirements SMART and redundancy-free.

Inconsult has developed a series of AI/ML methods to improve validation, especially in the medical/pharmaceutical environment, in order to automatically check the specification of requirements for compatibility with the SMART criteria as well as to keep them non-redundant. The improvement of quality in the planning processes for the introduction and modification of computerized systems pays off in particular because this way problems, which would normally only become apparent in the implementation phase, can be avoided from the outset. The tools developed by Inconsult to facilitate data exchange via URS/FS also greatly simplify requirements engineering work, especially on the customer side, by avoiding collaboration via data formats used by popular spreadsheet applications. In addition, the tool facilitates the traceability of requirements.

Feel free to contact us to learn more about our AI-powered validation approach, our experts Till Jostes (Head of Validation) and Dr. Matthias Rüdiger (Head of AI) will be pleased to answer your questions.