A team of researchers at Technische Universität Dresden has developed a biocompatible implantable AI platform that is capable of classifying in real time healthy and pathological patterns in biological signals like heartbeats. The platform doesn’t need medical supervision to detect medical changes.
The research was published in the journal Science Advances.
The Challenge of Implantable AI
While diagnostic data, such as ECG, EEG, and X-ray images can be analyzed with machine learning to detect diseases early on, it is still extremely difficult to implant AI into the human body. This is why the new advancement from the TU Dresden scientists at the Chair of Optoelectronics is such a big deal, as it is the first time such a system has demonstrated success.
The research team was led by Prof. Karl Leo, Dr. Hans Kleemann and Matteo Cucchi.
They presented a new approach for real-time classification of healthy and diseased bio-signals based on a biocompatible AI chip. The team relied on polymer-based fiber networks, which structurally resemble the human brain. These are what enable the neuromorphic AI principle of reservoir computing.
Polymer Fibers and Recurrent Networks
When the polymer fibers are formed in a random arrangement, this is referred to as a “recurrent network,” and it can process data like a human brain. Because the networks are non-linear, even extremely small signal changes can be amplified. An example of this would be a heartbeat, which doctors often struggle to evaluate. Tasks like this can be done through the polymer network easily thanks to the nonlinear transformation.
The AI demonstrated an ability to differentiate between healthy heartbeats from three common arrhythmias during trials, and it achieved an 88% accuracy rate. The polymer network also consumed less energy than a pacemaker.
According to the team, potential applications for such an implantable AI system include the monitoring of cardiac arrhythmias or complications after surgery. These can then be reported to both doctors and patients through a smartphone, which enables fast medical assistance.
Matte Cucchi is a PhD student and first author of the paper.
“The vision of combining modern electronics with biology has come a long way in recent years with the development of so-called organic mixed conductors,” said Cucchi. “So far, however, successes have been limited to simple electronic components such as individual synapses or sensors. Solving complex tasks has not been possible so far. In our research, we have now taken a crucial step toward realizing this vision. By harnessing the power of neuromorphic computing, such as reservoir computing used here, we have succeeded in not only solving complex classification tasks in real time but we will also potentially be able to do this within the human body. This approach will make it possible to develop further intelligent systems in the future that can help save human lives.”