Smartphone based Blood Oxygen Level Measurementusing Near-IR and RED Wave-guided Light
Nam Bui, Anh Nguyen, Phuc Nguyen, Hoang Truong, Ashwin Ashok, Thang Dinh, Robin Deterding, Tam Vu
1/30
PhO2
Chronic Obstructive Pulmonary Disease (COPD)
24 Million Americans
have COPD
2/30
Leading cause of death in the US, 2016
OFTENFOUND IN LATE STAGE
$34,000 treatment in ICU
Early detection of COPD is key to successful treatment
Heart rate
Blood pressure
Breathing rate
Wheezing sound
Oxygen saturation (SpO2)
Vital signs screening in COPD
3/30
Oxygen Saturation needs to be measured frequently
Oxygen Saturation (SpO2)
𝑺𝒑𝟎𝟐 =𝑵𝒐. 𝒐𝒙𝒚𝒈𝒆𝒏𝒂𝒕𝒆𝒅 𝒉𝒆𝒎𝒐𝒈𝒍𝒐𝒃𝒊𝒏
𝑻𝒐𝒕𝒂𝒍 𝑵𝒐. 𝒉𝒆𝒎𝒐𝒈𝒍𝒐𝒃𝒊𝒏𝟏𝟎𝟎%
O2
O2
O2
O2
Oxygenated hemoglobin Deoxygenated hemoglobin
4/30
Photon Detector
IR emitter Red emitter
AC
R IR
Non-invasive Measurement of Oxygen Saturation
Intensity
t
Intensity
t
ACIR
ACR
𝑺𝒑𝟎𝟐 = 𝒇𝑨𝑪𝑹𝑨𝑪𝑰𝑹
5/30
Red and IR Wavelength Selection
At Red= 660 nm and IR = 960 nm, the difference of light absorption
by oxygenated hemoglobin and deoxygenated hemoglobin is largest.
6/30
Absorbed by red
pigmentation
Harmful
Low penetration
Pros Cons
High accuracy Inconvenient
Costly
Smartphone-based blood oxygen measurement
Dedicated Hardware Solutions
A solution to capture the oxygen level and is
High accuracy.
Low-cost.
Easy to make by patients
8/30
9/30
PhO2
Smartphone based Blood Oxygen Level Measurement
10/30
IR emitter Red emitter
PULSE OXIMETER
Hardware Challenges
PHONE BASED PULSE OXIMETER
Light
Source
11/30
Challenge. Phone’s camera IR filter
Sensor
IR filter
Existing Software-based Solutions
DigiDoc iHealthCare
RGB
Color
Image
Red Light
IR Light
DISADVANTAGES
Inaccurate: Green channel is mostly absorbed by red pigmentation.
Coarse-grained: one record per 10 seconds.
Solution.
Red and NIR for Oxygen Saturation Measurement
NIR
14/30
NIR
is available in most of phone’s flashlight and is not filtered by the phone’s camera.has similar characteristics of hemoglobin absorption with the IR lights.
NIR
15/30
PhO2 NIR Extraction Technique
Absorbed by redpigmentation
Green filter
Absorbed
NIR
Red and Green filter cuttoff
And 5 More Challenges…
16/30
Smartphone’s camera picks up most of the lights
Finger movement and Contact pressure
Pressure
Micro movements
RGB Intensity
Conversion between RGB and light intensity
Flashlight heat can cause skin burn
17/30
3D printed Optical Hardware Add-on
18/30
Smartphone’s camera picks up most of the lights
RGB Intensity
Conversion between RGB and light intensity
Finger movement and contact pressure
Pressure
Micro movements
Flashlight can cause skin burn
3D printed Optical Hardware Add-on
Wavelength selection
𝝀𝑹= 678 nm
𝝀𝑵𝑰𝑹= 724 nm
𝑺𝒑𝟎𝟐 =𝜺𝑯𝒃 𝒓 − 𝜺𝑯𝒃 𝒓 𝑹
𝜺𝑯𝒃 𝒊𝒓 − 𝜺𝑯𝒃𝑶𝟐 𝒓 + 𝜺𝑯𝒃𝑶𝟐 𝒊𝒓 − 𝜺𝑯𝒃 𝒊𝒓 𝑹
20/30
X:678
Y: 724
Z:1.002
𝜶
𝜶 = 𝒂𝒓𝒄𝒕𝒂𝒏(𝒂/𝒃)
a
b
Pressure Detection and Recommendation
No
rma
lized
In
ten
sit
y
PPG signal in a 7 seconds window
Sample
No
rma
lized
In
ten
sit
y
Sample
PPG signal in a 7 seconds window
21/30
No
rma
lized
In
ten
sit
y
PPG signal in a 7 seconds window
Sample
22/30
Heat Reduction with Distributed Lighting Source
𝐸𝑝 = ℎ𝑓
𝑬 =
2.55 mm 0.6 mmModifyBandpass filter
To maintain
the high light
intensity
𝑯
Detecting pressure
ROI detection and create thePPG signals
PPG
Identify the peaksand troughs
Calculate peak-to-peak ratio
95
%
Collecting raw data
23/30
AlgorithmsPhO2
App overview
PPG display
Image frame
24/30
25/30
In-lab experiments
• 3 males, 3 females
• 5 times performs the hyperventilation (a breath
holding technique to reduce oxygen).
• Totally 560 samples.
26/30
27/30
PhO2 vs. Existing apps
Sp
O2 (
%)
Measurement Indices
DigiDoc
Sp
O2 (
%)
Measurement Indices
iHealthCareS
pO
2 (
%)
PhO2
Time (s)
Experimental results
28/30
3.5%
In clinic experiment
In hospital experiment with patient.
29/30
30/30
Fine-grained evaluation of the patient
Low Oxygen Level
Coughing
31/30
Heat Reduction Evaluation
Hand multimeter
Thermalcouple Probe
32/30
Output temperature of different add-on designs
0
10
20
30
40
50
60
Without Banpass Filter With Bandpass Filter
58.9
4538.5
36.840.3
37.6
Tem
peratu
re (
Cels
ius)
Large Tunnel Small Tunnel Multiple Small Tunnels
Conclusion
❑Accurately measuring oxygen level using low-cost opticalfilm filters.
❑Detect the appropriate contact pressure between fingerand phone’ camera.
❑3D-printed add-on to handle the problem of lightscattering and finger movement.
34/30
Application for optical light based sensing
35/30
Cholest-erol
Optical light
based sensing
Urea Lactate
Water
Alcohol Glucose
Smartphone based Blood Oxygen Level Measurementusing Near-IR and RED Wave-guided Light
Nam Bui, Anh Nguyen, Phuc Nguyen, Hoang Truong, Ashwin Ashok, Thang Dinh, Robin Deterding, Tam Vu
37/30
PhO2