A central technical focus was the research and development of neural networks for the analysis of medical time-series data, in particular ECG signals. Two model types were systematically compared: a deep convolutional recurrent artificial neural network (ANN) and an energy-efficient spiking neural network (SNN) with biologically plausible leaky integrate-and-fire neurons. Both models were fully implemented in PyTorch, Norse, and snnTorch and trained using a robust data preprocessing pipeline.
The evaluation demonstrated that the developed models outperformed the state of the art in sleep–wake detection. In particular, the SNN architecture proved superior in multi-class classification (sleep/wake and REM phases) as well as in apnea detection, while also offering advantages for deployment on energy-efficient, battery-powered hardware. This work resulted in the publication “Sleep Stage and Apnea Classification from Single-Lead ECG Using Artificial and Spiking Neural Networks,” which received the Best Paper Award at the IEEE-EMBS Conference on Biomedical Engineering & Sciences in 2024 [1]. In addition, the publication “A Formal Treatment of Homomorphic Encryption Based Outsourced Computation in the Universal Composability Framework” [2] was produced.
PriviLEG thus made a significant contribution to the protection of sensitive health data in the corporate context, combining state-of-the-art privacy technologies with innovative AI approaches and opening new perspectives at the intersection of IT security, medical technology, and artificial intelligence. The developed technologies strengthen the foundation for trustworthy digital health solutions and particularly support small and medium-sized enterprises in implementing sustainable, future-oriented health services.
- [1] G. I. Biri / A. Vasilache et al., “Sleep Stage and Apnea Classification from Single-Lead ECG Using Artificial and Spiking Neural Networks,” in 2024 IEEE-EMBS Conference on Biomedical Engineering and Sciences (IECBES), Dec. 2024, pp. 79–84. doi: 10.110
- [2] Beskorovajnov, W., Eilebrecht, S., Jiang, Y., & Mueller-Quade, J. (2025). A Formal Treatment of Homomorphic Encryption Based Outsourced Computation in the Universal Composability Framework. Cryptology ePrint Archive. Available at: https://ia.cr/2025/1
In this research focus, the FZI concentrates on practical research into the key technology of Artificial Intelligence (AI). Innovative AI solutions are developed and transferred to application areas such as mobility, robotics, healthcare technology, logistics, production, and supply and disposal on behalf of our partners and customers.
Um die sichere Digitalisierung zu ermöglichen, erforscht und vermittelt das FZI in diesem Forschungsschwerpunkt anwendungsnah innovative Konzepte, Methoden zur Absicherung von IT-Systemen sowie rechtliche Rahmenbedingungen.
