Latest Neurochip published in Nature Communications

Today, a new generation of high-density microelectrode array (HD-MEA) technology has been published in Nature Communications (link). This publication not only demonstrates the excellent performance of the chip, but also describes the large-scale extraction of neuronal metrics, such as axonal conduction velocity, and includes impressive showcases of applications with various neuronal preparations. The developed versatile platform for label-free, live-cell electrical imaging will further accelerate drug discovery and basic research in the field of neuroscience.

Neuron on a high-density microelectrode array. Image courtesy of Xinyue Yuan, BEL, ETH Zurich.

The work is the culmination of Xinyue Yuan’s PhD thesis project conducted at ETH Zurich in the Bio Engineering Laboratory (BEL) of Prof. Andreas Hierlemann, which was jointly supervised by our CEO, Dr. Urs Frey, and Prof. Hierlemann. Xinyue has successfully defended her thesis in May this year.

After more than 15 years of developing various generations of microelectrode arrays (MEAs), Prof. Hierlemann’s lab keeps pushing the limits of this research field. Prof. Hierlemann explains: “The new chip will significantly enhance the capabilities to study neurons and neuronal networks across scales, from subcellular compartments through individual neurons to entire networks in very short time. This applies to preparations of primary neurons, neurons that have been derived from hiPSCs or pieces or slices of neuronal tissue in healthy and diseased states and is particularly relevant for drug screening.”

The new device features a novel “dual-mode” circuitry architecture combining the two most commonly used readout schemes, the “switch-matrix” concept, pioneered by Urs at ETH Zurich, and a “full-frame” readout also commonly found in HD-MEA systems (these concepts, including their pros and cons are explained in a review article by members of the former Frey Initiative Research Unit at RIKEN in Kobe, Japan: Obien et. al). Briefly, the switch-matrix scheme allows for recording neuronal signals at lowest possible noise levels with a reduced channel count. On the other hand, a “full-frame” architecture features lower signal quality and spatial resolution, but one can simultaneously use higher channel counts, which shortens experimental time. With the novel dual-mode concept, the pros of both worlds can be leveraged within one device, thus making experiments more efficient. The dual-mode approach renders the system highly suitable for a large number of biological applications.


  • 19,584 electrodes and a spatial resolution of 3050 electrodes/mm2
  • Full-frame readout of all electrode simultaneously at 11.6 kFrames/s with 10.4 μVrms noise
  • Switch-matrix readout of a configurable subset of 246 channels at 24.4 kSamples/s with 3.0 μVrms noise
  • New algorithm to extract and track axonal signals providing morpho-electrical features, such as axonal branch length and conduction velocity from over 1000 individual neurons
  • Application showcase featuring 2D cultured primary neurons and human iPSC-derived neurons; acute brain slices from the hippocampus and cerebellum; 3D iPSC-derived neuronal spheroids; acute retinal preparations; and iPSC-derived neurons cultured on scaffolds

Project dual-mode: from chip design to applications

Dual-mode chip ready to be used for experiments. Image courtesy of Xinyue Yuan, BEL, ETH Zurich.

We asked Xinyue about her experience on developing the dual-mode chip. “I had lots of fun in every part of the project,” says Xinyue with a smile. “I enjoyed a lot the IC (integrated circuit) design part. There are many aspects to be considered during the design, and it was fun to tune the parameters in order to get the performance I wanted.”

Aside from designing the chip, Xinyue also performed experiments with it. “Doing biological experiments was also interesting for me. Since I am an electrical engineer, many things were new to me when handling biological samples. It was definitely very satisfying and exciting when I turned on the chip and saw nice signals on the screen during an experiment,” notes Xinyue. “Different activity patterns from different samples made me realize how important it is to have high-quality signals to derive reliable conclusions from the neuronal recordings.” She explains, “it might be common for more matured systems to easily get results, but for such a prototype system as I had, it needed more work and patience to make sure that every step was working well to get good signals.”

Overall, she mentions, “I am proud that I was able to finish the entire project, not only the IC design and making the system work, but also trying different types of experiments to demonstrate its applicability for various neuronal preparations. It was a satisfying achievement to extract the signals of thousands of neurons and track them over a long period, which was not possible before.”

Of course, as with all projects, there were some challenges for Xinyue. “The most challenging part in this project was to find the appropriate application to make full use of the dual-mode capability of the system,” explains Xinyue. “Each mode has its own advantages for different applications. During the project, we got great support from colleagues in the lab, and the idea of using dual-mode for recording human iPSC-derived neurons popped up. These cells have very low-amplitude signals, and tracking of axons requires the sensitivity of the switch-matrix. Plus, to parallelize the extraction of thousands of neurons, full-frame recording is needed. It was a perfect match for the dual-mode operation, and I am looking forward to the future contributions of the dual-mode MEA towards advancing iPSC research.”

The story of its origin: why dual-mode?

We asked our CEO Urs to tell us a bit more about the story behind this work.

“One of the key contributions of my own PhD work to the field back around 2005 was the pioneering of the switch-matrix concept for HD-MEAs,” Urs explains. “Basically, at that time our lab already had designed a functional MEA system using the full-frame concept, but with an electrode density of only 16 electrodes/mm2, the applications were very limited. We, therefore, ventured in the switch-matrix concept, boosting the electrode density to over 3000 electrodes/mm2. This drastically increased the range of applications of such systems. Finally, it was possible to extract what we call a “electrical footprint” of a single neuron over many, many electrodes.” The switch-matrix also facilitated improvements in spike sorting and the extraction of new subcellular features (more information here: ETHZ News article).

Urs continues, “Prof. Hierlemann’s lab then pushed the limits of the switch-matrix concept to ever larger arrays, with more channels and new readout modalities, one of which is the core sensor chip, used in our MaxOne and MaxTwo product lines. I’m convinced that having a high spatial resolution with over 3000 electrodes/mm2 and a low enough noise level is central to achieve good results using MEAs.”

Evolution of HD-MEAs at Prof. Hierlemann’s lab. Image courtesy of Prof. Hierlemann’s lab, ETH Zurich.

Urs moved to Japan in 2011 to start his own laboratory at the RIKEN Quantitative Biology Center. It was there, where the first discussions about dual-mode started. “It was 2012 in my former bioelectronics laboratory in Kobe, Japan, when we started discussing to build a system featuring both, a full-frame and switch-matrix readout in a single system, because combining the advantages of both would produce an ultimate HD-MEA, suitable for all applications,” Urs recalls. “I have to admit that, initially, I was very skeptical, and also a bit scared to get in touch with the ‘full-frame’ world, being a pioneer of the switch-matrix concept. I imagine my postdocs must have had a hard time convincing me to enter that domain. It also proved quite difficult to combine two different readouts in one system, especially as we were not willing to compromise on the spatial resolution—requiring packing complex circuitry within the small area available in each sensor pixel. It was a great accomplishment when Xinyue, who was a master student in my lab in Japan, and Dr. Sungho Kim, a postdoc at that time, and co-authors published a fully functional dual-mode MEA system in 2016.” With that first achievement, there was still more to improve. “To really make the system applicable in various applications, a major redesign was required, which we achieved now with the HD-MEA system just published. We doubled electrode and channel count, while maintaining the high electrode density and even further improved noise performance, sampling rate and power consumption.”

And yet, that was still not the end, still measurements had to be done, showcasing the versatility of the system with various applications.

Frey Initiative Research Unit with Prof. Andreas Hierlemann in late 2015. From left to right: Alexandra Dudina, Prof. Hierlemann, Xinyue Yuan, Urs Frey, Marie Obien, Stefan Huber, Michele D’Urbino, Tomoko Kamatani, Florent Seichepine.

Showcasing the system performance in the field

Aside from the recordings that were performed at BEL, including samples, such as primary neurons, acute retina, acute cerebellar slices, and human iPSC-derived cardiomyocytes, we wanted to feature a wide range of applications to demonstrate the versatility of the system. I recall when Urs approached me with yet another of what he calls “optimistic idea”, but most would just say “unrealistic idea”. Basically, the plan was to ship a totally bulky system (imagine two 1m3 boxes), consisting of various devices, cables, computers, fragile custom-made setups and more across the globe to our collaborators in Japan for conducting proof-of-concept studies. Ok, we did not really imagine that we would need multiple trials and several trips back and forth between continents. But while deemed pretty crazy, with the help of our logistics department at MaxWell Biosystems and supported by our Japanese distributor, Physio-Tech, the plan actually worked out pretty well. The system and, also, all participants survived the long journey.

Neuronal spheroids on the dual-mode chip. Image courtesy of Dr. Tetsuhiro Kikuchi, CiRA Kyoto University and Xinyue Yuan, BEL, ETH Zurich.

The dual-mode system was dispatched to two laboratories in Japan. One was Prof. Yamanaka’s institute, the Center for iPS Cell Research and Application (CiRA) in Kyoto University. We collaborated with Dr. Tetsuhiro Kikuchi, an assistant professor in Prof. Jun Takahashi’s lab at the Department of Clinical Applications. “In our lab, we develop mature neuronal spheroids generated from human iPSC-derived neurons, such as dopaminergic neurons, for regenerative medicine,” explains Dr. Kikuchi. “These spheroids can potentially be implanted into the human brain as therapy for neurodegenerative diseases, like Parkinson’s disease. Such spheroids preferably need to be spontaneously active and produce oscillation patterns similar to those observed in healthy brains. Thus, it is essential to characterize neuronal spheroids’ activity and evaluate their function as a spheroid, without dissociating them, before implantation. We were excited when we saw electrical activities of multiple neuronal spheroids placed on the dual-mode HD-MEA. The sensitivity and large sensor area are both very desirable for us to easily assess and study our neuronal spheroids’ function.”

From Kyoto, the dual-mode system traveled to Sendai, in Prof. Ikuro Suzuki’s lab in Tohoku Institute of Technology. Prof. Suzuki is an advisor for MEA-based assay development to the Consortium for Safety Assessment using Human iPS Cells (CSAHi) in Japan. Dr. Suzuki was very pleased with the results obtained with the dual-mode system. “In our lab, we are developing pharmacological and toxicological evaluation methods based on the electrophysiological activity in human iPS neurons and acute brain slices from mice. With our experience in assay development and extensive use of microelectrode arrays (MEAs), we have come to the conclusion that sensitivity and single-cell resolution are important to effectively detect the effects of compounds and differences between various cell lines. When we used the new dual-mode high-density MEA, we were impressed by the quality of the data that we were able to capture,” said Dr. Suzuki.

Movie: Propagating local field potentials in an acute hippocampal slice. Image courtesy of Dr. Ikuro Suzuki, Tohoku Institute of Technology and Xinyue Yuan, BEL, ETH Zurich.

“For example, we are using acute brain slices to model seizures in vitro. It was the first time for us to see the initiation and propagation of chemically evoked local field potentials in acute hippocampal slices at high resolution across different areas in simultaneous recordings from 20,000 electrodes within 2×4 mm2. Also, we observed individual action potentials propagating along axons of human iPSC-derived neurons cultured in a SCAD device,” describes Dr. Suzuki. “This is very impressive, given that iPSC-derived neurons usually have considerably smaller signals compared to primary cells.”

Different applications served with the dual-mode system.

Upon the return of the one and only dual-mode system to Basel, Xinyue’s mission was not accomplished yet. A major challenge was to develop algorithms to cope with the massive amount of data. Xinyue then analyzed all the acquired data and produced an impressive set of figures, movies and statistical graphs. “It was also fun to do data analysis, especially the tracking of axons from many neurons,” says Xinyue. “It seemed like a simple algorithmic problem, but it needed quite a bit of thinking and work to make it automatic. Yet another satisfying moment for me was when I found nice-looking neurons with neuronal footprints covering the entire array.” For analyzing network connectivity, Xinyue was also supported by Manuel Schröter, a postdoc at BEL.

Movie: Action potentials propagating along the axonal arbor of a neuron in culture. Courtesy Xinyue Yuan, BEL, ETH Zurich.

Movie: Neurons communicating in culture. Courtesy Xinyue Yuan, BEL, ETH Zurich.

“I was impressed how Xinyue managed to implement algorithms to extract amplitude, firing rate, axonal conduction velocity, axonal length, axonal branching, network burst features, and network connectivity for these large datasets and to produce beautiful movies of the raw data,” Urs smiles proudly. “With the risk of sounding a bit like a nerd, I have to admit that when I saw the movies recorded with this system, I kept watching them countless times. I’m still truly astonished being able to see all individual neurons communicating over such a large area at single-action-potential resolution in raw data. Finally, the obtained data also showed, that, while a full-frame readout boosted efficiency significantly, low-amplitude signals, such as those of iPSC-derived neurons, really require very-low-noise recordings, only available through the switch-matrix mode in this system. Therefore, the dual-mode combination really renders the system very versatile for many applications”, says Urs.


Yuan, X., Schröter, M., Obien, M.E.J., Fiscella, M., Gong, W., Kikuchi, T., Odawara, A., Noji, S., Suzuki, I., Takahashi, J., Hierlemann, A., and Frey, U. Versatile live-cell activity analysis platform for characterization of neuronal dynamics at single-cell and network level. Nat Commun 11, 4854 (2020).

MaxWell Biosystems AG is currently considering to commercialize this system. At the moment we do not take any preorders, but should you be interested, you may contact Marie at