Research Projects

Intelligente Diagnostik 2

Artificial intelligence for skin tumor diagnostics

Start: 06/2021

End: 12/2022

The second phase of the Intelligente Dianostik (Intelligent Diagnostics) project focuses on developing an improved. miniaturized and cost-effective diagnostic system that can be used to automatically expertly diagnose malignant skin lesions in a robust, explainable and safe manner. To this end, the project will first expand data acquisition to build a larger pool of training data and then improve the existing AI model. Cooperation with several clinics is planned that will allow higher case numbers and hence better training data. In addition, case numbers are to be increased by including additional skin diseases.

Within the scope of the project, the FZI is focusing on challenges that arise from the cross-site operation of AI systems. This also includes the exchange of the comprehensive training data between different locations. The FZI is modifying existing federated learning approaches in order to implement, on the one hand, an efficient exchange of information, and on the other a cross-site orchestrated training process to facilitate the creation and distribution of new improved AI models.

Further work will focus on miniaturizing the existing demonstrator. In this context, the FZI is developing a mobile client for the compact, hand-held diagnostic system (KOMPASS). However, in this regard the FZI is also studying possibilities to balance the functionality between the backend and the mobile device to enable lightweight and performant application despite fewer mobile client resources.

Contact

Christoph Becker

Vice Division Manager
Division: Software Engineering

Research focus

Applied Artificial Intelligence

In this research focus, the FZI prioritizes the topics of Artificial Intelligence (AI) as well as human and AI engineering. In addition, the FZI deals with questions on dedicated AI hardware and predictive AI.

More projects