As of October 12, 2014, the new Open Ephys site is open to the public. Not only does it look a lot nicer and contain a lot more information than the previous iteration, but we now feature a variety of projects developed outside of MIT. We hope you like it!
Josh Siegle and Matt Wilson have authored the first paper based on data collected with the Open Ephys system. "Enhancement of encoding and retrieval functions through theta phase-specific manipulation of hippocampus" appeared in eLife in July 2014. In their experiments, Siegle & Wilson used closed-loop optogenetic feedback to alter behavior in mice performing a spatial navigation task. The Open Ephys software and hardware were designed with this type of experiment in mind. The lightweight headstage and cable did not impinge animals' movement, while the modular software made it simple to incorporate algorithms for triggering stimulation at specific phases of the theta rhythm. The experiments also incorporated a dual-site flexDrive capable of simultaneously targeting both sides of the hippocampus. We're looking forward to many more papers based on Open Ephys data in the future!
A Q&A with Jakob and Josh was just published in a blog post on SparkFun.com, as part of their "Enginursday" series. This is a real honor, since SparkFun was an essential resource for us during the formative years of Open Ephys. Our first engineering project, the Twister, consisted of parts from the SparkFun Inventor's Kit before we turned it into a polished product. SparkFun's tutorials taught us a great deal about soldering, designing circuit boards, and general electronics know-how. We're proud to be featured on their site!
Thanks to Toni Klopfenstein for making this happen!
Open Ephys recently donated equipment and a teaching assistant to the Transylvanian Experimental Neuroscience Summer School near Cluj-Napoca, Romania. For the past 3 years, TENSS has taught students from around the world how to build their own cutting-edge rigs for two-photon imaging, in vivo whole cell recordings, behavior tracking, and chronic electrophysiology. It's organized by Florin Albeanu of Cold Spring Harbor Laboratory, Adam Kampff of the Champalimaud Neuroscience Programme, and Raul Muresan of Coneural Cluj-Napoca.
For seventeen days, students participate in modules that teach the theory behind techniques in systems neuroscience and provide hands-on experience with data collection. The venue is basically a bed and breakfast when it's not being used by the course, so everything had to be built from scratch. Since the focus is on DIY rigs and open-source tools, Open Ephys was the perfect choice for the electrophysiology module.
Within a few days of arrival, we had two Open Ephys data acquisition systems up and running, which allowed us to monitor single units from tetrodes in the prefrontal cortex and striatum of freely moving mice. We also recorded data from a mouse navigating in virtual reality, thanks to the help of Georg Keller from the Friedrich Miescher Institute in Basel. During the final days of the course, the students applied their new knowledge to design and carry out experiments that combined behavior, optogenetics, and electrophysiology.
Participating in TENSS was an extremely rewarding experience. If you're looking to gain firsthand exposure to the latest techniques for recording neural data, we'd definitely recommend applying next year!
Eliza Strickland from recently wrote an article about Open Ephys for IEEE Spectrum. It includes quotations from Josh and Jakob, the Open Ephys co-founders, as well as Reid Harrison of Intan Technologies, Keith Stengel of Neuralynx, and Andy Gotshalk of Blackrock Microsystems.
Here's one choice excerpt, courtesy of Reid:
"The existing tools are like the PCs and the Macs of the neuroscience world, but now we also have this Linux."
Successful Distribution of 100 Additional Acquisition Boards
In the summer of 2013, Open Ephys kicked off its beta testing program by paying Advanced Circuits to assemble 50 of our acquisition boards. Generous donations allowed us to distribute most of these boards for free, which lowered the barrier to entry for those interested in trying out our platform. Based on the feedback we got from our beta testers, we made some improvements to the boards, then initiated a second round of manufacturing in spring 2014. Advanced Circuits produced 100 more boards for us, to meet the increased demand following our presentation at the Society for Neuroscience conference in San Diego. As of last week, all of these boards have been sent to their final destinations.
Assembling 100 boards cost approximately $32,000, which included the price of the completed circuit boards, cases, and power supplies. Thanks to a donation of key components from Texas Instruments, we saved around $14,000 on parts.
We shipped these boards to over 50 labs around the world, adding China, Korea, Belgium, Switzerland, and Finland to the list of countries using Open Ephys. In this round of distribution, there were three institutes that requested 10 or more acquisition boards each: University College London, the Champalimaud Centre for the Unknown in Lisbon, and the Donders Institute in Nijmegen. Along with MIT, where Open Ephys was launched, there are now four "hubs" in which our acquisition systems are concentrated. Having hubs like these will be important for increasing adoption, since scientists are more likely to try out new hardware that their neighbors are already familiar with.
In total, we have now delivered over 38,000 channels of ephys recording capacity to the field. The first two basic science publications that include data collected with our platform are now in submission. We're looking forward to seeing many more in the future!
Open Ephys Hires Its First Support Person
When choosing an ephys system to buy, the availability of support is a crucial factor. Having a guarantee that faulty hardware will be replaced, or that someone will be available to help troubleshoot problems, often makes it worth the price of investing in a commercial platform. Since its inception, Open Ephys has successfully served its small user base entirely through volunteer efforts. But with the number of new systems about to come online, we decided it was time to hire an official support person.
Leftover donations from the last round of manufacturing will fund a contract with Miguel Hernández University in Alicante, Spain to provide technical support for Open Ephys. The point person will be Aarón Cuevas López, a PhD student who has already contributed substantially to developing and testing our platform. Having Aarón as an official support person will make it easier for everyone to use our system. We'll continue to rely on the constantly growing community for adding new features, but it will be hugely helpful to have Aarón available for fixing bugs and responding to technical questions.
We recently established an official partnership with the Champalimaud Neuroscience Program in Lisbon to manufacture Open Ephys acquisition hardware. This is the first time anyone outside of MIT will build our designs for distribution. Investigators at the Champalimaud—including Alfonso Renart, Adam Kampff, Leopoldo Petreanu, Megan Carey, and Zach Mainen—have been some of the most enthusiastic supporters of Open Ephys. We plan to produce 100 boards in Portugal in the next few months. Once these become available, we'll send out a newsletter with detailed information about how to purchase them.
Another avenue for getting your hands on an Open Ephys acquisition board is through CircuitHub, a startup aimed at lowering the barrier to entry for obtaining custom hardware. If you order a board using this link, CircuitHub will purchase all the parts and assemble the circuit board for you. You'll still have to find a way to 3D print or machine the case; instructions for that can be found on our wiki. We haven't ordered anything from CircuitHub yet, but it could become the easiest way to order acquisition boards in the future. If you're interested in testing this out, please get in touch with us—we may be able to coordinate a group order.
Stay tuned for more information about updates to the wiki and website, as well as the launch of the official Open Ephys store!
Wired.com just posted an article on Open Ephys, as part of their "Out in the Open" column. It does a great job of capturing the motivations behind our initiative:
We've been using the flexDrive (wiki) for over a year now in the Moore lab, recording almost 100 sessions in 5 mice. I'm just now starting to analyze neural ensemble statistics that require simultaneously recorded neurons.
Here's the real-world distribution of how many simultaneous neurons in primary somatosensory cortex (with some thalamic electrodes) I could sort over a total of 75 sessions in awake mice with 16 nichrome tetrodes.
The mean unit yield was 25.8, with a minimum of 8 and a maximum of 46 units. These numbers include some not so great recordings, and bad tetrodes hat got damaged etc., but only very few sessions were outright discarded, mostly in the beginning of the drive lowering process where it looked like some electrodes were not in cortex yet.
All in all, these numbers should be good enough to do some interesting assembly-analysis, though the relatively low density of the tetrode array (250 micron pitch) results in a relatively low occurrence of strong fast-timescale correlations between spike trains.
Open Ephys recently registered as a nonprofit corporation in the state of Massachusetts. This makes us an official entity, separate from any individual lab. Having nonprofit status will allow us to accept donations and sell supplies through our website. We plan to redistribute custom components, such as electrode interface boards, to cut down on wasted time and manufacturing costs. We'll send out an update when that's up and running.
Our founding board of directors is Josh Siegle, Jakob Voigts, Christopher Moore, Matt Wilson, and Caleb Kemere. Our mission statement (as written in our Articles of Organization) is:
"To promote tool-sharing among members of the worldwide systems neuroscience community. Open Ephys will support the development, distribution, and maintenance of open-source hardware and software for collecting and analyzing neuroscientific data. Special focus will be given to tools with expensive or inflexible commercial alternatives, and which serve the needs of a broad user base. Open Ephys strives to make it easier for investigators to share the tools they develop by establishing a centralized tool repository and by coordinating distributed support networks."
We think this represents an important unmet need in our field, and we hope Open Ephys can grow to fill this niche. Our ultimate goal is to not to create an open-source electrophysiology platform, but to change the way tools for neuroscience are developed and shared.
Open Ephys in new species
The initial testing of our data acquisition system was carried out in mice. Now we're happy to report that Open Ephys has been used to collect data from a number of other species. We've received reports of successful recordings from rat, zebra finch, and primate subjects. We also have some fresh data from human EEG. Our design for an EEG adapter board makes it possible to connect our headstages to a standard electrode cap.
Software, hardware, and firmware updates
The latest release of our GUI (version 0.2.5) includes some interface upgrades that make Open Ephys more convenient to use. We now have a "Graph Viewer" component that allows you to visualize your entire signal chain at once, making it easier to navigate between modules. Users now have the option to automatically load the last-used configuration upon launch, so it takes less time to start experiments. And minor tweaks, like ensuring the buttons inside the control panel collapse gracefully (instead of overlapping as they did previously), make the overall user experience more enjoyable.
You can download pre-compiled binaries for the GUI from our website. If you're a new user, we recommend starting with this tutorial.
Our acquisition board has been updated for the most recent round of manufacturing. It includes two useful new features: (1) a port that makes it possible to synchronize timestamps across boards connected to different computers and (2) protective circuitry in case the wrong power supply is used.
To view the design files, browse through our repository on GitHub.
Finally, the firmware for our acquisition boards now allows digital input channels to trigger an amplifier reset. This makes it possible to minimize electrical artifacts, for example when doing antidromic stimulation. Thanks to Reid Harrison at Intan Technologies and Shay Ohayon at Caltech for their help with implementing and testing this feature.
The FPGA firmware is also available on GitHub.
Shay Ohayon at Caltech just conducted an experiment that makes use of the numerous new modules that he developed for the Open Ephys system over just the last two months. He recorded neurons from the Middle Face Patch and verified the recording by analyzing the data in real time. Here's what he had to say:
When I first heard about Open Ephys, I got very excited. The system is extremely cheap and everything is open source. However, after I got my hands on the hardware and software, I was initially disappointed. The initial software release lacked many basic features one would need to run a full-blown acute monkey electrophysiology experiment. There wasn't an option to do real-time spike sorting, or to display real-time firing rates. Furthermore, it lacked the ability to connect to external sources of information, like events arriving from a machine which presents stimuli. Nevertheless, I saw a great potential in the design and decided it would be worthwhile to program all the missing components.
Two months have passed since. With a lot of help from Josh Siegle and the rest of the Open Ephys community, we are now close to releasing a new stable version with many new features that make the system much more useful for acute experiments. In many ways, it has surpassed the capabilities of my old recording system (MAP by Plexon).
Last week I finally found the time to test Open Ephys on my monkey. Below are some notes on my configuration and the new features that I have added:
Summary of the experiment and configuration
My first goal was to try and recreate the standard way signals are processed in Plexon. Josh was extremely helpful in debugging bugs related to split & merge modules, and recently Josh added this great feature that permits the visualization of the entire signal chain:
The advancers module is used to keep a record on where each electrode was placed in the recording chamber, and also records information about the depth of each probe that changes during the experiment. This makes post-processing analysis much easier!
The new spike detector module gives the user the ability to isolate units in real time, either by the box method, or polygons in true PCA space:
Finally, the PSTH module can display firing rates, averaged relative to trial onset, and aggregated across similar trial types (junk data, just for demonstration purposes):
Both trial and category information is sent over TCP/IP. High accuracy trial alignment can also be achieved by sending a single TTL pulse. However, software timestamps are quite accurate as well (~3-4 ms jitter, when sent from a different machine). Software timestamp taken in the machine running HUI scan be easily converted into hardware timestamps with very good accuracy using robust linear regression:
For my actual experiment, I recorded with a single electrode (1 MOhm), targeting the so called "Middle Face Patch" (image generated with Planner)
At depth 52mm, I was able to isolate a noisy unit (data shown below is from the real time spike sorting):
Trials were sent from our behavioral machine, which displays images to the monkey. The behavioral machine sent information about which image was displayed. This information was acquired in GUI using the Network Events source (see Spike sorting & PSTH). I could determine in real time that unit 2 was face selective by looking at the PSTH curves (unfortunately, I didn't take a snapshot). In post-processing, it is quite easy to read out the trial information that was sent and build an average raster plot (smoothed with a 3 ms gaussian kernel):
Here, you can see the average responses relative to image onset (approximated onset, photodiode information still not taken into account). First 16 images are face images, and the rest are non-face images.
The PSTH, averaged across the six image categories is:
which looks very very similar to what the PSTH module showed in real time.
The new spike sorting branch seems to be ready for action! It can be used in acute experiments in which real time characterization of isolated units is required.
Minor issues still remain, but all will be addressed in the upcoming weeks.
To view the code Shay used for these experiments, check out the "spikesorting" branch of the GUI on GitHub.
The latest version of the Open Ephys GUI includes several important upgrades. We've now made it possible to construct more complex signal chains. For example, you can split the signal chain, do separate analyses on each branch, then combine the branches with a merger. This is useful for analyzing spikes and LFP simultaneously, then integrating the results for visualization or closed-loop feedback.
To make it easier to navigate through your signal chain once it's constructed, we've added a Graph Viewer that allows you to visualize the connections between all your processors. Clicking on any one of the nodes will take you directly to its editor interface, so you no longer have to search through splitters and mergers to find it. Everything can be seen at a glance:
We've also made it possible to collapse and expand editors by double-clicking on their name. This will make it easier to construct long signal chains without taking up more screen real estate:
There are a variety of other useful features and bug fixes in this release, so we recommend upgrading as soon as possible. You can either download precompiled binaries from our GUI page, or download the source code from GitHub.
Thanks to Shay Ohayon for his input on these features! The next major release will incorporate some awesome new modules he developed, such as an Advancer Node to track electrode position, a PSTH node to display stimulus-locked firing rates, and a Spike Sorter (incorporated into the Spike Detector) to identify units in real time. Stay tuned for more info...
Jenni Siegel from the Cellular Mechanisms of Working Memory group at UT Austin recently sent us some screenshots from her latest experiments. She's been using the Open Ephys acquisition board and GUI to record from M2 and anterior cingulate in awake, head-fixed mice. The tetrode projections (right side of the screen) show some beautiful units! So far she's been really happy with the data quality—the noise floor is greatly reduced compared to her previous recording system.
There are a variety of opportunities to check out the latest from Open Ephys at the Society for Neuroscience conference in San Diego next week. Whether you're already using our tools, or just want to find out more about what we've been up to, we'd love to chat with you!
Open Ephys acquisition boards are now being evaluated by over 30 labs as part of our beta testing round. So far, we've received great feedback from these users. Everyone that's tried it has reported that the system is easy to set up and yields high-quality neural signals. They've also pointed out software bugs and feature requests that have made our system more user-friendly and robust.
Now it's time to spread the word about our tools, and hopefully recruit some new developers. Over the past two months, we've visited labs at the Karolinska Institute, UCL, NYU, Princeton, Harvard, and Cold Spring Harbor to discuss potential collaborations. But we expect to gain even more exposure at this year's Society for Neuroscience conference, which takes place in San Diego from November 9th through 13th.
At last year's SfN, we presented a working prototype system that was very well-received. Many of you signed up for this newsletter after seeing our tools in New Orleans. This year, we have an acquisition system that we've already mass-produced, and that's being used by labs around the world. We're excited to show it off at our poster presentation, which takes place on Wednesday morning. It's poster number NNN33—we hope to see you there!
This is the first year that Intan Technologies will have an exhibition booth at SfN. The Open Ephys system wouldn't be possible without the chips that Intan manufactures, and we're especially grateful for all the technical advice provided by Reid Harrison, the president of Intan. We encourage everyone to stop by booth #920, where our acquisition board will be on display.
We're also organizing an informal meetup, which will take place on Monday from 5:00-6:30 pm at Neighborhood in San Diego. Anyone is welcome to join. We'll be discussing ways to grow our initiative in the short term, and how to sustain it over the long term.
Finally, we're currently working out the details of the next round of manufacturing, which will take place in December. We'll make some small tweaks to the hardware (such as adding a port for synchronizing multiple acquisition systems), then have 50-100 boards assembled by Advanced Circuits. Stay tuned for more information...we'll send out a request form after the conference.
Open Ephys recently reached an important milestone: our hardware has spread beyond the labs that developed it. Over the next few months, we'll get lots of feedback from our new users. We also hope they'll help us improve the software by fixing bugs and adding new modules. A list of the labs that volunteered as beta testers can be found on our people page.
If you didn't receive one of our acquisition boards but are interested in testing one out, we now have a wiki page with instructions on how to build one from scratch. We've built a number of systems by hand with great success.
We'd love to kick off another round of manufacturing, but we don't know exactly when that will happen. Keep an eye on upcoming newsletters for more details. The timing will depend on both how quickly things progress with our beta testing phase, and how long it takes to secure funding for manufacturing. The first 50 boards cost around $20,000 to produce, and we expect the next 50 will be the same. We still haven't figured out the best financing model, but donations have worked well so far. If you might be interested in funding our efforts, definitely get in touch. Eight labs contributed to the first round of manufacturing, in addition to the generous donation of parts by Texas Instruments. We don't feel comfortable selling our hardware without some mechanism for providing support, but perhaps this will become a possibility in the future.
We're also looking for people to help with software development. Our platform is already at the point where it has all the functionality needed to carry out basic electrophysiology experiments and observe data in real time. We've been using the Open Ephys system on a daily basis for the past few months, and we're very happy with its usability. But there's plenty of room for improvement, especially when it comes to making the software easy to modify. The fact that it's open source already represents an advantage over the commercial alternatives, but we'd like to make it simpler to add new processing modules, even for those with limited development experience. If you work with a programmer that might be able to contribute some of his or her time, or have access to funds that could be used to hire one, please get in touch. We've been amazed by how liberating it feels to collect all of our data with open-source tools, and we'd like others to experience the same thing.
Now that our first round of manufacturing is complete, it's time to take a look at our finances. We funded the construction of our first 50 acquisition boards solely through donations. A handful of labs chipped in to get the process up and running. We received generous gifts from the Meletis and Carlén Labs at the Karolinska Institute, the Jazayeri Lab at MIT, the Yizhar Lab at the Weizmann Institute, the Kemere Lab at Rice University, and the Goldberg Lab at Cornell. The Wilson and Moore Labs have continued to support us through miscellaneous small purchases.
A major contribution also came from Texas Instruments. Through their partnership with Rice University, we received over $17,000 worth of free components (around $7200 of which went toward our current manufacturing run). This included the auxiliary analog-to-digital and digital-to-analog converters, the most expensive parts on our boards.
The final product was 50 acquisition boards with custom cases, not including the FPGA used for USB communication ($400 extra). The grand total for the parts, assembly, and cases was $19,294, or approximately $386 per board. This would have been $26,500 if we hadn't gotten the donation from Texas Instruments.
Here's the breakdown of what we spent it on:
- Printed circuit boards: $3064
- Assembly and DigiKey parts for first 5 boards: $1600
- Assembly and DigiKey parts for last 45 boards: $5319
- Omnetics connectors: $2728
- Samtec connectors: $388
- LEDs: $248
- Cast urethane cases: $4420
- Acrylic tops: $349
- Screws, rubber feet, and hex keys: $66
- Spray paint and glue: $15
- I/O boards: $1010
- Boxes for shipping: $69
- "5V DC ONLY" stickers: $18
Obviously these costs don't account for the time we spent orchestrating all of the orders, but this wasn't that much in the end. We only lost a few days of work, in return for spreading our tools to over 30 labs around the world.
To put this in more relatable terms, we just built fifty 256-channel data acquisition systems for a fraction of the cost of one commercial system. If you add $20k to account for the FPGAs purchased by individual labs, the cost remains lower. Of course, labs still need to purchase headstages and cables, but at around $1000 for 32 channels, this is still a bargain.
For the average lab, will the incredibly low price justify the extra time investment needed to get an open-source system up and running? Perhaps not at the moment, but potentially it will in the future. Over the next few months, our beta testers will help us spot problems with our hardware and software, and hopefully add some useful features. If all goes well, we should reach a point at which the extra effort is negligible, but the price difference is substantial.
Last week we received 45 assembled acquisition boards from Advanced Circuits and placed each one in a cast urethane case. Needless to say, they're looking great!
After some bugfixing in the spike display, we are now using the software for our actual long recordings in mice implanted with the 64 channel flexDrive (see GitHub for the EIB design). The dual-screen interface works perfectly and the runtime-extracted spikes look as good as the one we extracted from the full voltage traces.
This week we received a shipment of five fully assembled Open Ephys acquisition boards from Advanced Circuits. Previously, all of our boards had been assembled by hand, a process that takes 2-3 hours per system. In order to spread our tools beyond our labs, we needed to outsource the manufacturing.
After running the boards through a battery of tests, we're happy to report that everything is working exactly as expected. We're now ready to order an additional 45 units, which we'll distribute to labs around the world for beta testing.
In our first blog post, we described the fantastic signal quality of the Open Ephys acquisition system. Today, we made our first direct comparison between signals recorded with Open Ephys and those recorded with a Neuralynx Digital Lynx system. The conclusion? The recordings are virtually indistinguishable. There's still more rigorous testing to be done, but to the trained eye of a seasoned electrophysiologist, the signal quality of the two systems is qualitatively identical. Read on to see for yourself.
We performed the test by recording from a mouse implanted with an eight tetrode flexDrive. The Omnetics connector on the electrode interface board was compatible with both the Open Ephys headstage and a special-ordered Neuralynx HS-36.
The mouse was placed in a 1' x 1' arena, and attached to either recording system via a tether that was counter-balanced via a system of pulleys.
We made sure the settings on each system were the same (1 to 7500 Hz bandpass, ~30 kHz sampling rate), then recorded for 3 minutes with Open Ephys and 3 minutes with Neuralynx
The examples below all come from a single electrode, but the signals across all 32 channels are identical between the two systems, as far as we can tell. We hope to carry out a more detailed comparison in the near future, but we couldn't be happier with the initial results!