Bio-optical Sensing Dissertation

UT Biomedical Informatics Lab
Depth Resolved Diffuse
Reflectance Spectroscopy
Ricky Hennessy
The Biomedical Informatics Lab (BMIL)
The Biophotonics Laboratory

Committee
1/60
Mia K. Markey, Ph.D. - Advisor
The Biomedical Informatics Lab
James W. Tunnell, Ph.D. - Advisor
The Biophotonics Lab
Stanislav Emelianov, Ph.D.
Ultrasound Imaging and Therapeutics
Andrew K. Dunn, Ph.D.
Functional Optical Imaging Lab
Ammar M. Ahmed, M.D.
Dermatology at Seton

Diffuse Reflectance Spectroscopy (DRS)
2/60

Applications of DRS
3
Cancer
Detection
Soil
Characterization
Wearable Tech
Food Quality Endoscopic Surgery
Cosmetic
Applications

Cancer Detection with DRS
4/60
Extract Features from Data Use Features to Create Classifier
Rajaram et al. Lasers Surg Med 42:876-887 (2010)

DRS Instrumentation
5
Spectrometer
LightSource
SourceFiber
DetectorFiber
Computer
FiberBundle

Light Absorption in Tissue
6/60
400 450 500 550 600
0
0.2
0.4
0.6
0.8
1
WAVELENGTH (nm)
ABSORPTION(a.u.)
Melanin
HbO
2
Hb
Other Absorbers
• Water
• Bilirubin
• Lipid
• Protein
• Collagen

Light Scattering in Tissue
7/60

Biological Origins of Scattering
8/60
Scattering is caused by index
of refraction mismatches
• Cells = ~10 μm
• Nuclei = ~1 μm
• Collagen = 0.1 μm
• Membranes = 0.01 μm

Scattering Coefficient (μs)
9/60
Scattering coefficient (μs) is proportional to
concentration of scatterers in a medium
μs
-1 is the average
distance a photon
travels between
scattering events
μs

Direction of Scattering
10/60
Isotropic Scattering Anisotropic Scattering
Tissue scattering is in forward direction
(g = ~0.9)
Henyey-Greenstein Phase Function
g = 0

Radiative Transport Equation (RTE)
11/60
ΔEnergy Scattering in Scattering out Absorption

Reduced Scattering Coefficient
12/60
1
2
10
9
8
7
6
5
43
Using reduced
scattering with isotropic
scattering is equivalent
to larger scattering with
anisotropic scattering

Diffusion Approximation to the RTE
13/60
Kienle et al., JOSA A, 1997, 14(1), 246-254
Assumes isotropic scattering
• Scattering >> absorption
• Source Detector Separation > ~ 1 mm
Blood is highly absorbing
Epidermis is ~100 μm thick

Monte Carlo Simulation
14/60

UT Biomedical Informatics Lab 15

UT Biomedical Informatics Lab 16

The Monte Carlo Lookup Table
(MCLUT) Method
Hennessy et al., JBO, 2013, 18(3), 037003

Tissue Properties  Optical Properties
SCATTERED
ABSORBED
TISSUE PROPERTIES  OPTICAL PROPERTIES  SIGNAL
FORWARD MODEL
SCATTERING PROPERTIES
ABSORBER CONCENTRATIONS
• HbO2
• Hb
• Melanin
18/60

Forward Model Flowchart
TISSUE PROPERTIES  OPTICAL PROPERTIES  SIGNAL
FORWARD MODEL
Light Transport
Model
Scattering at λ0, Concentration of chromophores
Calculate absorption and scattering coefficients at
each wavelength
Scattering
Absorption
Use light transport model to
calculate diffuse reflectance
Generate diffuse reflectance spectrum
19/60

Monte Carlo on GPU
Modern processor w/ 4
cores = 4 times speedup
Modern GPU w/ 500 cores =
 500 times speedup!
 < $300
Alerstam et al., Biomed. Opt. Express., 2010, 1, 658-675
20/60

Monte Carlo Lookup Table (1 Layer)
400 450 500 550 600 650 700
0
1
2
3
4
5
6
7
8
9
WAVELENGTH (nm)
REFLECTANCE
21/60

MCLUT Inverse Model
REFLECTANCE OPTICAL PROPERTIES  TISSUE PROPERTIES
Optimization
Routine
22/60

Calibration
MCLUT Modeled Spectra
RMC = photons counted
Measured Spectra
Rmeas = Iraw/Istandard
Modeled spectra and measured
spectra have same optical
properties
23/60

Validation of One-Layer Model
RMSPE = 2.42% RMSPE = 1.74%
Validation with 3 X 6 matrix of phantoms containing hemoglobin
and polystyrene beads.
Decreased percent error of 3.16% and 10.86% for μs' and μa,
respectively, when compared to experimental LUT method
24/60

Errors Caused by One-Layer
Assumption for Skin

Fit Two-Layer Data with One Layer Model
26/60
Hb + HbO2
melanin
melanin +
Hb + HbO2
1. Create two-layer
spectra
2. Fit with one-
layer model
• [mel]
• [Hb]
• SO2
• Scattering
• Epidermal thickness
• [mel]
• [Hb]
• SO2
• Scattering
• Vessel radius
Notice that the fit
is very good

Melanin
27/60
• One-Layer model
underestimates [mel]
• Magnitude of error is
dependent on epidermal
thickness (Z0)
• Z0 Error

Hemoglobin
28/60
• One-Layer model
underestimates [Hb]
thickness (Z0)
• Z0 Error

Oxygen Saturation
29/60
• One-Layer model
overestimates SO2 when
SO2 < 50%
• One-Layer model
underestimates SO2
when SO2 > 50%
thickness (Z0)
• Z0 Error

Pigment Packaging
30/60
• In tissue, blood is
confined to vessels
• This significantly
reduces the optical
path length where
absorption is high
(Soret Band)
• Causes a flattening of
the absorption
spectrum

Vessel Radius vs. Epidermal Thickness
31/60
0 50 100 150 200 250 300
0
50
100
150
200
250
300
350
400
450
500
Top Layer Thickness (mm)
VesselRadius(mm)
• Vessel radius (pigment
packaging) factor is
highly correlated with
top-layer thickness
• Pigment packaging
factor is likely a
combination of vessel
packaging and
epidermal thickness

Correlation Between [Hb] and [mel]
32/60
R = 0.04 R = 0.80
One-layer assumption causes artificial correlation
between [Hb] and [mel]

Conclusions about One-Layer Errors
Causes underestimation of [Hb] and [mel]
– Magnitude of error is function of epidermal
thickness
Causes error in SO2 that is a function of
epidermal thickness as well as SO2
Vessel Packaging factor and epidermal
thickness are highly correlated
Causes an artificial correlation in [Hb] and
[mel]
33/60

The Two-Layer Monte Carlo
Lookup Table Method
Sharma, Hennessy et al., Biomed Optics Express, 2014, 5(1), 40-53

Motivation of Two-Layer Model
Epidermis
Dermis
Stratum Corneum
Melanin
Hemoglobin
35/60

Motivation of Two-Layer Model
• Pigmentary disorder studies
• Disease (rosacea, lupus, scleroderma, morphea, lymphederma) monitoring
• Treatment outcome measures for many cosmetic procedures
• Topical medical absorption studies
• Measuring thickness of psoriatic plaque
• Determination of epidermal thickness
Melasma
Method of melasma
treatment depends on
depth of melanin
Wood’s lamp is current
method to determine
location of melanin
Qualitative – More contrast for
epidermal melasma. Doesn’t work
for patients with dark skin.
Asawansa et al., Int. J. Derm, 1999, 38, 801-807
36/60

Two Layer MCLUT
We can create a 5D LUT
1. Top layer absorption
2. Bottom layer absorption
3. Scattering
4. Top layer thickness
5. SDS
37/60
Segment of 5D lookup table
• Z0 = 200 μm
• SDS = 200 μm
• R1 = R2 = 100 μm
Sharma, Hennessy et al., Biomed. Opt. Express, 2014, 5(1), 40-53

Two-Layer Forward Model
38/60

Two-Layer MCLUT Inverse Model
39/60

Two-Layer Phantom Construction
40/60

Phantom Validation Study
PHANTOM μs’ (mm-1) μa,t (mm-1) μa,b (mm-1)
1 1.5 0.25 1.275
2 1.5 0.25 1.275
3 1.5 2.3 0.25
4 1.5 2.3 0.25
5 1.5 0.25 2.3
6 1.5 0.25 2.3
7 2.85 0.25 2.3
8 2.85 0.25 2.3
9 2.85 2.3 0.25
10 2.85 1.275 0.25
11 2.85 1.275 2.3
12 0.75 0.25 1.275
13 0.75 1.275 0.25
14 0.75 0.25 2.3
Set of Two Layer Phantoms
Range of scattering and
absorption in skin
μs’ = 1-3 mm-1
μa = 0 – 2.3 mm-1
Nichols et al., JBO, 2012, 17(5), 057001
41/60

Phantom Data
42/60

Results: Top Layer Thickness
43/60
Error = 10% for Z0 < 500 μm
Photons are not
sampling this
deep.

Results: All Optical Properties
44/60

Results: Dependence on SDS
45/60
Main Takeaway
Accuracy of extracted
parameters is dependent on
probe geometry. This is due
to sampling depth of
probe.

Sampling Depth of Diffuse
Reflectance Spectroscopy Probes

Probe Geometry and Sampling Depth
47
Tissue
Source
Fiber
Detection
Fiber
SDS
Sampling
Depth
Z(μa,μs’,rS,rD,SDS)
rS rD
-Tissue Optical Properties
-Source/Detector Fiber Sizes
-Source/Detector Separation
* Credit to Will Goth for this slide
47/60

Defining Sampling Depth
48/60
This experiment was performed computationally (MC simulation) and
experimentally (phantoms)

Experimental Validation
49/60
E = 1.71% E = 1.27% E = 1.24%
SDS = 370 μm SDS = 740 μm SDS = 1110 μm
Look at axes to
see deeper
sampling for
larger SDSs

Analytical Model of Sampling Depth
50/60
This expression can be
used to aid in the design
of application specific
DRS probesE = 2.89%
Expression was found using
TableCurve 3D. Free parameters
[a1, a2, a3, a4] were selected using a
least-squares fitting algorithm.

Choice of g and Phase Function
51/60
The choice for g and phase function had negligible
impact on the sampling depth model

Applying the Sampling Depth Model
52/60
6-around-1 adjacent fiber
orientation with medium (series
1), high (series 2), and low
(series 3) absorption.
Sampling depth changes
with wavelength

Pilot Study: Two-Layer MCLUT
Method Applied to In Vivo Data

Study Population and Data
54/60
• 80 Subjects
• IRB approval from UT Austin - #00002030
• 51 males, 29 Females
• Average age of 25.7 years
• Ages 18-46
• Measured spectra from the following
anatomical locations
1. Back
2. Calf
3. Cheek
4. Forearm
5. Palm
• Measured the following
1. Melanin
2. Hemoglobin
3. Scattering
4. Epidermal Thickness
This study is still unpublished

Instrumentation
55/60
x 2
Source Diameter = 40 μm
Ring 1 Diameter = 40 μm
Ring 2 Diameter = 200 μm
Ring 1 SDS = 55 μm
Ring 2 SDS = 205 μm
Unfortunately, data from
ring 1 was unusable

Melanin
56/60
Back Calf Cheek Forearm Palm
0
0.5
1
1.5
2
2.5
3
3.5
4
MelaninConcentration(mg/ml)
Palm has less melanin,
which agrees with the
expected result.
Average of 1.83 mg/ml
is within range of
published values for
melanin concentration
[0-5 mg/ml]

Hemoglobin
57/60
0
0.5
1
1.5
2
2.5
3
3.5
HemoglobinConcentration(mg/ml)
Higher levels of
hemoglobin in the face
and forearm agrees
with the expected
results.
Average of 1.37 mg/ml
is within range of
[Hb]
[0.5-10 mg/ml]

Scattering
58/60
No significant
difference between
anatomical locations.
Average of 22.75 cm-1
is within range of
scattering at 630 nm
[15 – 25 cm-1]

Epidermal Thickness
59/60
0
20
40
60
80
100
120
140
160
EpidermalThickness(mm)
No significant
difference between
Average of 90 μm is
within range of
epidermal thickness.
[40 – 200 μm]
We expected to see a
difference between

Conclusions about Pilot Study
A two-layer model can be used to extract
depth dependent properties from in vivo DRS
data
The results agree with previously published
values
However, we expected to see a difference
between anatomical locations for epidermal
thickness
– This could be due to the absence of data from
the inner ring of fibers.
60/60

Conclusions about Pilot Study
Data should be recollected with multiple
SDSs
Additional patient data such as race/ethnicity
and skin color should be documented
61

Overall Conclusions
 The MCLUT method is an accurate and fast way to
analyze DRS data
 A one-layer assumption for skin causes significant
errors in DRS data analysis
 The MCLUT method can be extended to two-layers,
allowing the extraction of depth dependent
properties
 Depth sampling of DRS probes can be tuned by
changing the probe geometry
 DRS can be used to measure the depth dependent
properties of skin in vivo
62/60

Contributions to Field
63
The MCLUT Method Two-Layer MCLUT Method
One-Layer Errors Analysis
DRS Sampling Depth Analysis

Funding
The NIH (R21EB0115892)
The NSF (DGE-1110007)
CPRIT (RP130702)
64

Acknowledgements
65
The Biophotonics Lab
• James W. Tunnell
• Will Goth
• Bin Yang
• Manu Sharma
• Sam Lim
• Sheldon Bish
• Xu Feng
• Varun Patani
The Biomedical Informatics Lab
• Mia K. Markey
• Nishant Verma
• Gezheng Wen
• Clement Sun
• Nisha Kumaraswamy
• Hans Huang
• Juhun Lee
• Gautam Muralidhar
• Daifeng Wang
UT BME Staff
• Margo Cousins
• Brittain Sobey
• Michael Don

Bio-optical Sensing Dissertation

More Related Content

What's hot (20)

Similar to Bio-optical Sensing Dissertation (20)

Recently uploaded (20)

Bio-optical Sensing Dissertation