1. ELECTRONIC STETHOSCOPE ARRAY Groupe 70 Robin GUIGNY & Fatima Zohra HASNAOUI ECE 445 Spring 2016

2. INTRODUCTION  2/5 medical misdiagnosis & underdiagnosis per year  Traditional stethoscopes only provide qualitative diagnosis  Find a way to provide a more accurate diagnosis

3. Objectives  Have a quantitative assessment of lung activity  Capturing and processing the lung signal to identify symptoms  Easy-to-read output for the doctor

4. Overall System Patient Sound capture&treatment Signal analysis and diagnosis

5. Table of contents 1.Block diagram 2.Microphone 3.Filter 4. Bluetooth module 5 Headphone amplifier 6. Power supply 7. Software

6. Block diagram

7. Choosing the microphone  Sensitivity and frequency response  Cost  Interferences  Power

8. Microphones comparison MEMSCondenserFiber optic Sensitivity & freq response >100Hz Cost$400 Interferences Power Acceptable Very bad Good Final choice!

9. Choosing the filter  High attenuation in the stopbands  Flat passband  Easy design for multiple modifications Requirements: Attenuation < 6 dB in the passband [50Hz ; 2500 Hz] Attenuation > 12 dB in [0 - 30 Hz] and 6000Hz +

10. Filter design  Butterworth filter : maximally flat in the passband.  3rd order : to obtain desired attenuation in the stopband.  Sallen-Key design : easy to implement and modify.

11. Identification  Butterworth  2 nd cutoff frequency : f=2500Hz Q=1/√2

12. Schematics

13. LTSpice Bode diagram

14. Results

15. Attenuation < 6dB passband Vs/Ve > 1/2 at 2500 Hz and at 50Hz ✔

16. Attenuation > 12dB Vs/Ve < 1/4 for 6000Hz+ and [0 – 30 Hz] ✔

17. Bluetooth module choice  Analog input  Resolution > 8 bits

18. Numato’s GPIO BT Module  10 bits resolution  3,3V ADC range  RN-42 2.4GHz Bluetooth module  Command through HyperTerminal

19. Headphone amplifier design  Able to drive low-impedance output found in headphones (9Ω to hundreds of Ω).  Provide >0,5mW to 9Ω earphones (96dBSPL/mW) for the loudest expected sound. This corresponds to 67mV output. Requirement

20. Cmoy audio amplifier

21. Results ≈70mV Pk-Pk for loudest sound ->93dBSPL : tolerated for 8 hours X100 amplification

22. Power unit  Relatively small in-use time (estimated <30 min/day)  No particular need for rechargeable battery.  Able to provide +/-9V to all the opamps Requirement  The battery must allow 10 hours of use (last about a month)

23. Power supply schematics

24. Battery life  2 x 9V alkaline batteries + 3V coin cell battery  Battery life calculation : LT082 opamps : 2,5mA Absolute max for BT module : 95mA

25. Software overview  Aim : visualize lung sounds and spot symptoms  How? According to lung sounds features (FFT, shape, length)  Two main abnormal sounds are linked to lung diseases: Wheezes Crackles Fine crackles Coarse crackles

26. Lung sounds features SoundsMax peak frequencyWaveformDuration Normal200Hz Breath in = Cycle*2/3 Breath in = breath out + 11dB 2s<cycle length<10S Wheezes400HzSinusoid>80 ms Fine Crackle650HzDampened wave deflectionAround 5ms Coarse Crackle350HzDampened wave deflectionAround 15 ms Focus of the software

27. Steps Pick the soundPress a buttonCompute fft Find max peak’s frequency (f>200Hz) Conclusion

28. Flowchart

29. User interface Scroll bar

30. Example

31. Requirements  Latency : 2 to 10 s  95% success

32. Verifications  Latency : ok ✓  95% success rate : ok on lung sounds files provided by medical research articles  -> need for a bigger sample

33. Issues  Bluetooth : not able to connect the bluetooth module  Analog identification can be possibly implemented if we take into account localization : A) Fine crackles B) Wheezes C) Coarse crackles Lung sound preferential spots

34. Conclusion  Hardware  Software  Link between the two?

35. Further work  Bluetooth transmission between hardware and software  Add microphones (multiple inputs)  Take microphone localization into account  Try with other type of filters (FIR..)  Design stethoscope head for better acoustic clarity

36. Thank You!