With this research, eyetec will pursue new methods to test the efficiency gains that can be made by introducing Machine Learning for Visual Inspection of ATMPs.
Innovation as a driving force
The research project aims to test the feasibility for developing a new system for visual inspection of ATMPs (Advanced Therapy Medicinal Products). The goal is to develop a system that can (self-)adjust based on machine-based learning. In turn, eyetec expects to improve efficiency, decrease rejection rates, compared to the manual process, reduce waste, and curtail costs by minimising the spillage of ATMP medication. With this study, eyetec will also check to what extent this process can take place under the GMP (Good Manufacturing Practice) quality standard that prevails for drug manufacturing.
The eyetec team has a strong background in process engineering and project management, in combination with hands-on expertise in compliance, process improvements and qualification. Therefore, the necessary expertise in pharmaceuticals and visual inspection is already available within the company, in addition to optimum capacity to properly execute the project.
The nature of the study
Over a period of twelve months, the team will investigate some concrete objectives that will determine the feasibility of implementing a system for Visual Inspection of ATMPs with Machine Learning.
The first objective is to investigate the current systems used for visual inspection of ATMP. For this eyetec will use a questionnaire (which you can find here) and evaluate the answers through an analysis. The study should cover at least 70% of the production volume and include different types of ATMPs. This to ensure that a generic solution can be developed. The chosen method should then be tested to check what efficiency gains can be made by introducing Machine Learning. A cost-benefit analysis will be an important and determining factor in selecting a possible solution. In addition, eyetec will also examine whether a Machine Learning-based solution could potentially be approved under Good Manufacturing Practice (GMP), a regulatory quality standard for producing drugs.
Challenges
While the study sounds promising, eyetec is not immune to potential challenges. It might be difficult to make a generic solution due to the big variety of ATMPs. In addition, the lack of a clear regulatory framework for the Visual Inspection of ATMPs makes it harder to determine real-world methods. Furthermore, because of the broad range of substances, the chosen visual inspection method should be able to recognise and deal with both diversity as well as small, often patient-specific, sample sizes.
Researchers may also face time constraints for visual inspection, due to rapidly increasing cell death from the moment of contact with the freezing medium, in addition to the occurrence of shear when spinning containers, which could also result in a die-off of material. Lastly, the limited use of AI in Flanders within the visual inspection niche, partly due to the restrictive framework of GMP, could potentially complicate the research.
Impact and follow-up process
However, the team is hopeful about the outcome. If the feasibility study has a positive outcome, eyetec will continue to develop the project, where they will explore further development of the solution. Assuming positive study results, eyetec estimates a five-year implementation term after the market launch.
