A Reusable Biological Observatory (ARBO)
ARBO is a reusable/reclaimable self-contained biologging device fit for small to mid-size animals (~500 grams). The device has dedicated hardware to record electrocorticography/electroencephalography (ECoG/EEG), electrocardiography (ECG), motion, and body temperature (thermal). As an implantable, ARBO interfaces with the external world using a dynamic RFID/NFC chip and an onboard microcoil antenna. Data is logged to a microSD card using the Arduino M0 Feather as a motherboard.
Low-frequency brain signals obtained from ECoG and EEG are, at the very least, useful correlates of behavior. Additionally, they are critical to distinguishing sleep from wakefulness, whereas general measures of motion or movement are incapable of making such distinctions. These data can also be used to identify sleep states and stages that are known to influence memory and brain plasticity.
I chose to use the ADS1299 biopotential measurement chip to accomplish these recordings over an SPI interface. The ADS1299 has been successfully implemented in a number of applications (see OpenBCI Cyton) and has active community support. It has been compared against the more specialized RHD2000 series chips from Intan Technologies. While I like Intan products, and have used their development platform to perform in-vivo, single unit electrophysiology, they are simply part of a more expensive development/production ecosystem.
ARBO utilizes the 4-channel version of the ADS1299 to reduce power consumption. These inputs are configured in the bipolar “referential montage” using a single reference for all recording channels with an optional BIAS drive to reduce the potential DC offset between the chip and animal. Digital and analog ground planes separate the ARBO circuitry using a dedicated PCB layer (4-layers total: signal, signal, ground, signal) which is standard practice.
While the ADS1299 is capable of recording multiple biopotentials at once, I chose to use a dedicated ECG chip (MAX30003) for two reasons. Firstly, this allowed me to use the single reference mode on the ADS1299 (instead of each input being bipolar). Secondly, the MAX30003 performs R-to-R detection on-chip (Pan-Tompkins). Heart rate variability (gathered from R-to-R) is a surrogate marker of stress, although it represents multiple physiological factors. Nonetheless, dedicating hardware to ECG offloads computational and memory demands necessary to do online or offline R-to-R detection.
Deciphering how motion maps onto behavior without video has been tackled using a number of classification algorithms. Certainly, the more data, the better. I used the MPU-9250 which is a “System in Package” (SiP) combining the MPU-6500 (a 3-axis gyroscope and 3-axis accelerometer) and the AK8963 (a 3-axis magnetometer). The MPU-6500 has 16-bit precision and is accessible over the I2C bus. The MPU-9250 not only has community support, but includes advanced features such as low-power modes and hardware interrupts based on user selected movement thresholds.
Body temperature regulation is an important component in maintaining homeostasis, especially during cold weather or hibernation. I used the MAX31856 thermocouple converter that communicates over the SPI bus. A thermocouple is the gold standard in temperature measurement, especially in harsh environments. The thermocouple lead also allows temperature to be recorded from any accessible anatomical location, and importantly removes the temperature measurement from the circuit board.
At the very least, a passive RFID tag has some utility in identifying devices (and animals) wirelessly. Dynamic RFID/NFC tags, such as the M24LR04E, go one step further by enabling two-way wireless communication through a chip which also interfaces over the I2C bus. I have referred to this as the 4-Kbit EEPROM “ARBO mailbox” because it allows device states and a unique identifier to be read, while syncing timestamps or recording modes can accomplished by writing to the chip (ultimately read by the microcontroller). The addition of RFID/NFC also makes “smart traps” possible, where only animals of interest will trigger the trapping mechanism (in development).
After trying to configure a PIC-based system, I turned back towards Arduino and a microcontroller platform/motherboard that was already established. The Adafruit Adalogger leverages the powerful, 32-bit Cortex M0 processor which has plenty of peripheral support and several low-power modes. Additionally, the Adalogger implements power regulation, a USB programming interface, I/O expansion, and a microSD card slot for rapid (v1.x) prototyping.
The ARBO enclosure (in development) was designed using SolidWorks and aims to wrap the circuitry and battery in a biocompatible, lightweight 3D-printed resin. The current size is just slightly larger than a standard USB key, and all components included, weighs roughly 23 grams (or 4.64% of a 500 gram animal). Electrodes are soldered onto the surface pads of the ARBO circuit board and covered with a form-fitted cap (not shown). All enclosure openings will be insulated with a biocompatible silicon. ARBO can be implanted subcutaneously or intraperitonealy.
Schematic & Design Files
All the resources for ARBO will be made open source once the device undergoes testing and validation. The schematic and PCB were designed using KiCad and will be prototyped using OSH Park. Nearly the entire BOM was sourced from Digikey. Even at low quantities, each device costs less than $200, all components and 3D printing included.