Development of a Smartphone-based Pulse Oximeter with Adaptive SNR/Power Balancing Tom Phelps, Haowei Jiang, and Drew A. Hall University of California, San Diego http://www.BioEE.ucsd.edu
Motivation $2,060/person/yr Life Expectancy (Yrs) $9,403/person/yr $930/person/yr $48/person/yr Millions of people worldwide suffer from preventable diseases, but lack access to adequate healthcare equipment. 2 WHO Global health expenditure database, http://www.who.int/nha/expenditure_database/en/
Pulse Oximetry • Non-invasive measurement of peripheral oxygen saturation (SpO 2 ) and heart rate (HR) • Commonly used to monitor: • Pregnancies (i.e., preeclampsia) • Chronic respiratory illnesses (i.e., COPD, asthma, CF) and pneumonia • Cardiovascular diseases • Sleep apnea 3
Existing Pulse Oximeters Clinical-grade Portable Lower accuracy at a modest High accuracy due to advanced cost ($40-$200) signal processing and (mostly) stationary patient Problem: Limited Problem: High cost (~$1k) and computational power, motion high power (~10W) → not artifacts → not sensitive portable Challenge: Achieving high accuracy at low cost 4
Mobile Phones Processor Memory Battery Storage Display User Input Inertial Sensors Biometric Sensors Wireless Radios How can one tap into the mobile phone for mHealth devices? 5
Smartphone-based Pulse Oximeter P av = 5-20 mW (depending on phone) Use the infrastructure in a mobile phone to realize a low cost (but high accuracy) portable pulse oximeter Chengyang Yao, Alexander Sun, and Drew A. Hall, “Efficient Power Harvesting from the Mobile Phone Audio Jack for mHealth 6 6 Peripherals”, Global Humanitarian Technology Conference (GHTC) , Seattle, WA, October 8-11, 2015
Circuit Implementation Current-mode LED Driver: Photoreceiver: • Control V LED → I LED → Light • Zero-bias photodiode → low dark intensity current → save voltage headroom • AC-coupled to right audio channel • AC-coupled to mic. Channel • C 2 filters out interference • Clamp diodes to protect mic input 7
Signal Processing Advanced Signal Processing • Quality index assessment • Motion artifact removal • Powerline interference removal • etc. Signal processing entirely done on the phone! Easily updated, adaptive, and more computationally intensive algorithms possible. 8
� Power and SNR Optimization (𝐼𝑆 %)** 𝐼𝑆 + ) - +(𝑇𝑞𝑃 -,%)** 𝑇𝑞𝑃 -,+ ) - 𝐹𝑠𝑠𝑝𝑠 = 4.5 1.2 4 1 Error (unitless) 3.5 Power (mW) 3 0.8 2.5 0.6 2 Dynamic Power 1.5 0.4 1 0.2 Static Power 0.5 0 0 0.1 0.16 0.22 0.28 0.34 0.4 0.46 0.52 0.58 Amplitude (a.u.) 9 9
� Power and SNR Optimization (𝐼𝑆 %)** 𝐼𝑆 + ) - +(𝑇𝑞𝑃 -,%)** 𝑇𝑞𝑃 -,+ ) - 𝐹𝑠𝑠𝑝𝑠 = 4.5 1.2 4 1 Error (unitless) 3.5 Power (mW) 3 0.8 2.5 0.6 2 Dynamic Power 1.5 0.4 1 0.2 Static Power 0.5 0 0 0 5 10 15 20 25 30 35 40 45 50 Pulse Width (%) 10 10
Measurement Results 200 1.84% %Error SpO 2 %Error Heart Rate 180 Subject 1 2 3 µ 1 2 3 µ Measured Heart Rate 160 Gal. S6 -6 1 1.8 3.2 5.3 1.4 1.8 2.8 -0.56% This Work 140 Note 4 4.2 0 -6 3.4 4.9 0 -6 3.7 120 0.04% Note 3 3.2 0 -2 1.6 -3 0 -2 1.6 100 -0.002% LG V10 0 4 -3 2.5 -2 4.3 -3 3.1 80 Gal. S6 5.3 4 0 3.1 -1 -1 0 0.7 60 0.37% iOximeter Note 4 4.9 0 3.4 2.8 1 1 -2 1.3 Expected 40 Note 3 -3 6 3.5 4.1 -2 -2 -2 2.0 Measured 20 LG V10 -2 3 0 1.6 2.1 -5 -3 3.4 0 0 50 100 150 200 Expected Heart Rate BE Biomedical PS- Masimo RAD 87 used to collect 2110 Patient Simulator true HR and SpO 2 Despite low-cost and simplicity, HR accuracy < 1.8% and SpO 2 < 3.7% 11
Conclusion • Developed a low-cost (BOM < $20) smartphone- based pulse oximeter • Adaptive SNR and advanced signal processing techniques enabled by using the smartphone for all computation 12
Thanks! Gabby Kang
Recommend
More recommend
