Category: Automation and High-Throughput Technologies

1069-E - Applying image-guided scene analysis for process optimization and automated documentation in regulated environments

Wednesday, February 7, 2018
11:30 AM - 12:30 PM

Laboratory workflows are highly complex efforts with multiple challenges with regard to organizational as well as technical aspects. Accordingly, constant improvement and simplification is a constant and ongoing effort in the lab. Documentation represents one of these challenges, which is a very time and labor intensive task with a high priority - although not very popular task among the workers. By examining a simple liquid handling process, a new approach for the documentation task has been derived in order to improve and simplify the current state of the art. The prototype was developed in the Innovation Center for Lab Automation nICLAS and uses a stereo vision camera and state of the art computer vision techniques to track the hands of the worker and their movement. The captured movement of the hands can then be categorized to not only capture the hands of the worker but also the separate the information into distinct process steps. Each single process step as well as the whole process itself can then be documented on the fly showing additional process information. By automatically generating the required documentation, the worker has more time for the primary process and the documentation is uniform. Additionally, with a database containing the process in its desired procedure, deviations can be detected between the model process and the captured process. These deviations can either be directly indicated to the worker or marked in the documentation to indicate production errors. With a direct feedback, the worker has the ability to discover errors at their source and verify that faulty products are eliminated from the workflow. Also, the system can provide a step-by-step guidance to support new or long-time workers in learning new processes. Based upon the seamless tracking of the workflow, errors can be detected and a direct feedback can be used to support the learning process. Interactions with the detection system are kept at a minimum because the system can automatically detect finished process steps and jumps to the next one. According to the developers the system might give valuable support in all regulated lab environments as diagnostics, sterile processes (e. g. isolator, cell culture) as well as small scale GMP production as used for cell therapy products.

Christian Jauch

Research Associate
Fraunhofer Institute for Manufacturing Engineering and Automation IPA
Stuttgart, Baden-Wurttemberg, Germany

Christian Jauch graduated as a M.Sc. in technical cybernetics at the University of Stuttgart in 2015. He currently works at the Fraunhofer Institute for Manufacturing Engineering and Automation IPA in Stuttgart at the Department Machine Vision and Signal Processing. His current research focuses on scene analysis in different environments, industrial as well as domestic.