1. Measuring Ocular Microtremor Thomas Rutkowski Vivian Phinney Salman Al-Saif Mark Hasemeyer Steve Carroll

2. Ocular Microtremors (OMT) -Low-amplitude, high-frequency tremors of the eye. (20-100Hz) -Caused by random, ground-state signals on oculo-motor nerves -Oculo-motor nerves originate in brain stem. OMT indicates brainstem health. -OMT correlated with patient’s level of consciousness. -Gives strong indication of the chances of recovery for coma patients.

3. System Overview Waveform Generator OMT Simulator Sensor (Provided) Power Supply Analog Input/ Preprocessing Circuitry - Amp Test MUX Microprocessor with Integrated A/D, UART, and DSP Bedside Monitor (Philips MP-60/70 or Agilent V24/26) Philips VueLink Module – M1032A For MCU, currently looking at Atmel AVR32 with 10-bit A/D and integrated DSP functions,50 MHz clock. System Module Overview LCDSD Card

4. 16-bit Timer 10-bit A/D UART Controller 256kB Flash AVR32UC3B1256 Microcontroller 16Mhz Crystal PLL – 50 Mhz CPU Clock Signal Conditioning Sensor RealTerm/ Bedside Monitor 3.3 V Power Supply MAX3380E UART Level Shifter SPI Controller 3.5” Graphic LCD 16-bit Parallel Interface SD Card Basic CPU Architecture and Connectivity GPIO Controller

5. Start Sampling Buffer 0 Wait until Buffer 0 Full (1 sec) Start Sampling Buffer 1, Start FFT on Buffer 0 Wait until Buffer 1 Full (1 sec) – FFT on Buffer 0 finishes ~1 ms Start Sampling Buffer 0, Start FFT on Buffer 1 Find OMT frequency from FFT output just completed Currently using 1024 sample buffer and 1024 kHz sampling frequency, giving 1 second of data in buffer. Send frequency and waveform samples to UART(monitor), LCD, and SD card. Perform inverse FFT on frequencies within valid range. Wait until Buffer 0 Full (1 sec) – FFT on Buffer 1 finishes ~1 ms Processing Algorithm

6. Signal Processing – General Overview Using FFTs, much faster than DFT algorithms Atmel studio has several built in FFT algorithms that require minimal modification  But… it’s not working as we had anticipated Further work in DSP is needed  Just recently got sampling, buffers, and FFT working  Further testing required to find optimal DSP algorithm

7. Signal Sampling Using the built in ADC in the Atmel processor to sample data The input voltage ranges from 0V to 3.3V (adjustable), the signal needs to be scaled and clamped to avoid causing damage to the processor Use of a 8:1 voltage divider, or an op-amp circuit with gain K = 0.125, assuming a maximum 12V input from the sensor IC A 3V zener diode is used to clamp the analog input, or a comparator circuit

8. Signal Sampling (contd.) -Sensor output is in the range ±12 V -A/D takes 0 - 3.3 V -First stage inverts and scales sensor output to ±1.5 V -Second stage adds 1.5V DC offset for new range of 0 – 3 V for A/D

9. Signal to be sampled Frequency components at 15Hz, 90Hz, and 165Hz

10. Signal recreated with MATLAB

11. FFT As with other things involving signal processing, we are currently experimenting with the built in FFT functions The FFT routines functioned properly with minimal modifications Going forward, two possible solutions to processing that we are considering After determining the frequency of the OMT signal, a sinusoidal wave is generated with varying amplitudes proportional to the OMT Using wavelet processing We will know more once we’ve done more careful testing, since our sensor is very sensitive, there is a lot of room for error

12. FFT (contd.) Filtering and Noise Reduction As mentioned previously in PDR, the OMT frequency range lies between 40Hz and 120Hz Use of a high pass filter at 20Hz, and a low pass filter at 150Hz We identified the main source of noise as that coming from various power lines across the room A notch filter is used (bandstop) for the frequency range 57Hz-62Hz Sensor is sensitive, picks up the slightest movement, another sensor might be used to provide a differential solution The other main artifact that affects the signal is heartbeat and voice, we are still considering solutions to both problems

13. FFT (contd.) The following slide represents an example of the same signal sampled above We used the FFT algorithms to represent the frequency content of the analog signal Simple processing was done to filter out unwanted frequency content (below 20Hz, above 150Hz, and 60Hz) The slide after that is a capture of Vivian’s OMT frequency content

14. Processed signal

15. Captured OMT Signal over 1 Second Interval

16. We believe that thorough testing and experimentation is very important to the success of our project After CDR, we will continue with OMT processing

17. Power Management Utilizing a medical grade power supply, since project is not power centric, plus it is safer to go with this option and avoid any risk or harm to the experimental subjects (basically us) ELPAC MED113TT, provides three regulated outputs, 5V, ±12V LP2951 voltage regulator from TI will be used to power the Atmel processor

18. Power Management (contd.)

19. Bedside Monitor Leasing from Medelco Lease will include:  Philips MP70 Touch Screen Monitor  M1032A Vuelink Interface Module  M8048A FMS (Flexible Module Server)  M3001A MMS (Multiple Monitoring System) Possibly:  M1032A#K66 Cable with DSUB-9 connector

20. VueLink Cable Pinout Showing UART Tx/Rx Lines VueLink ConnectorDB-9 Connector

21. Backup Plan - Analog Input Signals

23. LCD Display http://www.crystalfontz.com/product/CF AF240320K-T-TS -Parallel or SPI Interface to LCD - 4-Wire Resistive Touch Panel - Will display small waveform and OMT frequency, in addition to bedside monitor display. -Touch-screen menu may allow user to configure data-logging options, sampling rate, and buffer size.

24. SD Card Socket Board http://www.sparkfun.com/commerce/produ ct_info.php?products_id=204 Breakout Board for SD-MMC Cards Routes SD card pads to solder able through-hole connector. We will use SPI mode to read/write from card. Card requires 3.3 V Data Logging can be started using touch screen menu on LCD.

25. Risks and Contingencies Blowing the sensor piezo-transistor or integrated IC amp Eyetect shipped a backup sensor earlier this week The FFT signal processing algorithm is a concern Waveform spiky and aperiodic Now that we have the sampling, buffers and basic FFT working, can test other processing strategies: Wavelets FIR Filters on Buffer followed by Peak-Count Algorithm Statistical Analysis

26. ProductUnit CostAmountTotal CostProvider PROCESSING/MANUFACTURINGAtmel AVR32 Development Kit179.001 http://mouser.com/ 4-layer PCB66.003198.00 http://www.4pcb.com Packaging95.002190.00 http://www.firstcut.com/ INTERFACEPhilips M3 Bedside Monitor600.00/mo.2.51,500.00 http://www.medelco.com/patient_monitors.html Vuelink M1032A-1- M3001A MMS (Multiple Monitoring System)-1- M8048A FMS (Flexible Module Server)-1- Vuelink connector to DSUB9 Cable175.001 http://www.dotmed.com/listing/594385 LEMO Circular Push Pull Connectors24.92374.76 http://mouser.com/ 2GB SD Flash Card Model CT2GBSD7.991 http://www.newegg.com/ Breatkout Board for SD-MMC Card BOB-0020417.95353.85 http://www.sparkfun.com/ LCD Display - CFAF240320K-T-TS30.74392.22 http://www.crystalfontz.com/ MEDICAL GRADE POWER SUPPLYVoltage Regulator LP2951_33 (3.3V)0.9654.80 http://www.newark.com/ Surface mount bypass capacitors (25 capacitors)4.5029.00 http://www.ti.com/ Surface mount resistors0.20306.00 http://www.amazon.com/ Power Connector (5-pin)5.50211.00 JB Saunders ALPS_SKHUQBO56A Switch2.0024.00 http://www.allelectronics.com ADMINISTRATIVE COSTSPrinting Manuals40.001 Poster12.00/ft.4 feet48.00 Rider Insurance Policy for Bedside Monitor250.001 Shipping/Handling Fees100.002200.00 MARGIN7% 213.05 TOTAL COSTS 3,256.67 Outside Funds OUTSIDE FUNDINGEyeTect3,000.00 UROP800.00 Northrop Grumman200.00 TOTAL FUNDING 4,000.00 Current Budget

27. Division of Labor TomVivianMarkSalmanSteve Power Power Supply X Powering Sensor X X MCU Signal Sampling - A/DX X DSP of SignalX X X Interface LCD Interface on Dongle X Data Capture on SD Card X X Condition Sensor Output X X X Patient Bedside Monitor InterfaceX X Manufacturing PCB X X Dongle X Documentation Technical ManualX X User's ManualX X UROP X Extra (if time) Data Mining/Trend Evaluation X X

28. Deliverables Expo Bedside Monitor Interfaced LCD on Dongle Functioning Packaging Finalized Milestone 2 PCB Sized for Production SD Card reader Integrated Signal Processing Implemented CAD Design sent to Plastic Casing Company Milestone 1 Obtain Bedside Monitor and Open Interface Specification SD Card Interfaced with Microcontroller Sample the Sensor Signal (save on SD Card) Finalize which Sampling Algorithm to Use

29. Timeline

30. Questions? In Order of Appearance: Tom Steve Salman Mark Vivian