Optogenetics—the control of neurons using light—is a promising technology, but optogenetics research generates a lot of data. Edward S. Boyden, an MIT neuroscientist, found that his neural probing of a mouse brain was generating more data—a terabit per second—than his computers could handle, according to Murray Carpenter in the Boston Globe.
One approach to Boyden’s problem would be to apply racks and racks of $3,000 servers, but Andrew Meyer, chief executive of a computer company called LeafLabs, said such servers were overkill for Boyden, who needed storage but not functions like graphics processing. Consequently, LeafLabs designed a computer with few extras.
“We just started ripping stuff out until all we were left with was the storage, and just a tiny, tiny amount of a parallel processor that could suck up the data and put it on the storage, and that was it. It was as minimalist as we could do,” Meyer said, as quoted by Carpenter.
LeafLabs now offers a data-acquisition system called Willow for neuroscience applications. A Willow module communicates concurrently with as many as 32 industry-standard neural amplifier chips, and it contains an FPGA that processes 1000+ channels of electrophysiological data. The FPGA writes data directly to a storage drive, while simultaneously forwarding the data to a computer for real-time monitoring and feedback. The modular base unit is designed to be scalable with the goal of ultimately supporting 1 million channels.
Willow’s starting price is $17,000, according to Carpenter in the Globe.
The application of the computer is described in the paper “A direct-to-drive neural data acquisition system” by Meyer, Boyden, and several coauthors and published in frontiers in Neural Circuits. They write, “Here we present a novel architecture in which a digital processor receives data from an analog-to-digital converter, and writes that data directly to hard drives, without the need for a personal computer to serve as an intermediary in the DAQ process. This minimalist architecture may support exceptionally high data throughput, without incurring costs to support unnecessary hardware and overhead associated with personal computers, thus facilitating scaling of electrophysiological recording in the future.”
LeafLabs says its goal is to solve moonshot computing challenges: “We design powerful physical computing devices for control and communication among smart machines (including humans!). We work with scientists, entrepreneurs, artists, and enthusiasts to implement beautiful ideas as quickly and elegantly as possible.”
Carpenter in the Globe quotes Meyer as saying, “Independent of neuroscience, the scale at which data is useful in understanding your problem just keeps moving up and up and up.”
