Implantable AI system developed for early detection and treatment of disease

Artificial intelligence (AI) will radically change medicine and healthcare. For example, patient diagnostic data such as ECG, EEG, and X-ray images can be analyzed using machine learning, resulting in subtle changes. However, the transplantation of AI into the human body remains a major technical challenge. Scientists at the Dresden University of Technology, chaired by Optoelectronics, have for the first time succeeded in developing a biocompatible implantable AI platform that classifies healthy and pathological patterns of biological signals such as heartbeats in real time . low. Detects pathological changes without medical supervision. The research results were published in the journal Scientists progress..

In this work, a research team led by Professor Karl Leo, Dr Hans Kleemann and Matteo Cucchi will show an approach to classify healthy and diseased biological signals in real time based on biocompatible AI chips. They used a polymer-based fiber network that structurally resembles the human brain and enables neuromorphic AI principles of reservoir computation. The random placement of polymer fibers forms a “recurring network” that allows data to be processed in the same way as the human brain. The non-linearity of these networks makes it possible to amplify even the smallest changes in the signal, which are often difficult for physicians to assess, for example in the case of heartbeats. However, nonlinear transformations using polymer networks make this possible without problems.

In the trial, the AI ​​was able to distinguish between a healthy heart rhythm and three common arrhythmias by 88%. Precision report. In the process, polymer networks consumed less energy than pacemakers. The potential uses of embedded AI systems are diverse. For example, it can be used to monitor post-operative arrhythmias and cardiac complications, report both doctors and patients via smartphones, and provide prompt medical assistance.

“The vision of combining modern electronics with biology has made great strides in recent years with the development of so-called organic mixed conductors,” explains Matteo Cucchi, doctoral student and lead author of the treatise. “But so far, success has been limited to simple electronic components such as synapses and individual sensors. It has never been possible to solve complex tasks. In our research, this We have taken an important step towards realizing our vision. By leveraging the power of neuromorphic computation, such as reservoir computation, used here, we can not only solve complex classification tasks in real time, but also. It may also be possible to do it in the human body. This approach will allow us to develop smarter systems that will help save lives in the future. “

Reference: Biocompatible organic for brain-inspired biosignal classification by Matteo Cucchi, Christopher Gruener, Lautaro Petrauskas, Peter Steiner, Hsin Tseng, Axel Fischer, Bogdan Penkovsky, Christian Matthus, Peter Birkholz, Hans Kleemann, Karl Leo Reservoir Computing Using Electrochemical Networks, August 18, 2021 Scientists progress..
DOI: 10.1126 / sciadv.abh0693