Researchers developed energy-efficient AI chips for diagnosing a common cardiac condition called atrial fibrillation.
Artificial Intelligence is promising technology for biomedical applications such as assisting physicians in diagnosing. But the challenge is that AI based algorithms and systems require high computational power, and as a result, consumes enormous amounts of power. Researchers from the Fraunhofer Institutes for Integrated Circuits IIS and for Industrial Mathematics ITWM have developed energy-saving AI chips that can help with early stage detection of atrial fibrillation, which is a heart rhythm disorder.
Atrial fibrillation is a common cardiac arrhythmia. This condition can lead to a stroke if it is not detected in time. A way to detect this conduction is by recording electrocardiograms (ECGs) over a long period of time, reducing the chance of detecting irregular heart rhythm. But the algorithms for evaluating patient data can be very computationally intensive, which results in high energy consumption.
The highest priority of any mobile diagnosis device is energy efficiency. Researchers rely on deep learning algorithms that use neural networks with processing in multiple different layers. The ECG signal is used as an input into the neural network. The sections of the signal are filtered, and the individual signal components are weighted and summed up in several layers.
“In the first layer of the neural network, a certain signal behavior is detected. In the second layer the characteristics are placed in relation to each other. A total of six layers are used. A complex image of the ECG signal that indicates an existing disease does not emerge until the last, sixth layer,” explains Marco Breiling, a scientist at Fraunhofer IIS.
A research team led by Dr. Marco Breiling developed a way to process these digital ECG time-series signals in an efficient way. They put the signal processing as a part of the AI chip to sleep as long as it is not needed, saving 95% of the energy.
“The chip collects the ECG signal for 12.7 seconds and then processes it in just 24 milliseconds, or 0.2 percent of the time. So the processing sleeps for over 99.8 percent of the time and uses a neglectable amount of energy. Thanks to non-volatile RRAM memories that are part of the system, signal processing can resume immediately after wake-up, after almost 12.7 seconds, without consuming any energy,” explains Breiling.
Moreover, researchers implemented systolic arrays, which is a special chip architecture for high energy efficiency. “For permanent operation, our chip requires so little power that a solar cell with an area of 6 x 6 mm2 operated in moonlight would suffice. Alternatively, the chip could evaluate ECGs for 330 consecutive days using the very smallest coin cell available on the market,” says the researcher.
According to the researchers, the developed circuit is not only suitable for medical use, but also for other applications where time-series signals are processed, such as condition monitoring and predictive maintenance.