Design and Performance of a Localized Fiber Optic, Spectroscopic Prototype Device for the Detection of the Metabolic Status of “Vulnerable Plaque”: in-vitro Investigation of Human Carotid Plaque
Design and Performance of a Localized Fiber Optic, Spectroscopic Prototype Device for the Detection of the
Metabolic Status of “Vulnerable Plaque”:
in-vitro Investigation of Human Carotid Plaque
2
OUTLINE• INTRODUCTION
– Problem identification, objectives, specific aims, hypotheses, and background review
• OPTICAL DESIGN– ~ 1 mm3 tissue volume interrogation achieved with optical probe
• METHODS– Laboratory setup– Data collection – Calibration model development
• RESULTS• DISCUSSION/CONCLUSIONS
– Limitations– Future work
3
PROBLEM IDENTIFICATION• Atherosclerotic cardiovascular disease 6.3 M deaths / yr
worldwide • Cardiovascular disease #1 killer in the U.S.• 1.5 M myocardial infarctions (MI) / yr in the U.S. • 250,000 / yr sudden cardiac deaths • $111.8 billion / yr health care costs (direct/indirect)• Major risk factors
– Smoking– High blood cholesterol (LDL/HDL ratio)– Physical inactivity– Overweight/Obesity – Diabetes mellitus
Source: American Heart Association. 2002 Heart and Stroke Statistical Update. 2001. http://www.americanheart.org
4
Everybody has atherosclerosis, the question is who has vulnerable plaque
Sudden Cardiac DeathAcute MI
VulnerablePlaque(s)
5
Unknown Diagnosis – Vulnerable Plaque
• The “precursor” that ultimately ends in acute thrombi (clots) of sudden death MI
• Inflammatory cells found preferentially in vulnerable plaque
• Activity sustained through anaerobic metabolism and lactate production
6
Morphology vs. Activity Imaging
Inactive and non-inflamed plaque
Active and
inflamed plaque Appear Similar in
IVUS OCT MRI w/o CM
Morphology
Show Different
Activity
Thermography, Spectroscopy, immunoscintigraphy, MRI with
targeted contrast media…
7
HISTOLOGY
LIPID CORETHROMBUS
FIBRO-CALCIFIC
J Am Coll Cardiol. 2001 Sep;38(3):718-23.Am J Pathol. 2000 Oct;157(4):1259-68.
Courtesy of Texas Heart Institute
8
LONG-TERM OBJECTIVES
• Develop an optical spectroscopy catheter system to determine the metabolic status of atherosclerotic vessels – No exogenous dyes– No ionizing radiation– Low cost addition to existing cardiac catheterization laboratory
• Locate and identify vulnerable plaque based on metabolic status with optical spectroscopy
9
SHORT-TERM GOALS• Demonstrate feasibility in-vitro of optical
spectroscopy to accurately determine metabolic status– Tissue lactate concentration– Tissue pH
10
Specification Tissue pH Tissue Lactate Concentration
[moles lactate / gram tissue]
in-vitro Temperature (C)
in-vitro Experimental Stability
< 0.03 change/hr
- <0.4C change/hr
Optical Calibration Range
6.80 – 7.60 2 – 20 32.0 – 38.0
Optical Calibration Accuracy/Prediction Error(R2 / SECV)
0.75 / 0.08pH units
0.75 / 1.0 umoles/ gram
tissue
-
Maximum # of Factors in Optical Calibration Model
5 to 6 independent samples per
factor
5 to 6 independent samples per
factor
-
11
SPECIFIC AIMS1) Design a reflectance-based fiber optic probe that uses
visible to near-infrared light optimally to interrogate a small volume of tissue.
2) Estimate the depth penetration of fiber optic probe, based on theory and experiments.
3) Identify major interferents to the optical spectra and tissue reference measurements collection.
4) Collect and analyze fresh tissue from human carotid endarterectomies to create large optical calibration training set while maintaining tissue in a viable physiological state in-vitro.
12
HYPOTHESES1) A small fiber optic prototype can make optical
measurements in-vitro for the assessment of metabolically active plaque in a defined region of tissue (< 1 mm3 volume).
2) in-vitro experimental factors can be assessed to their importance in affecting the optical calibration accuracy. The tissue temperature, experiment time, and gross pathology are identified a priori.
3) Mathematical models can be developed which relate the corresponding optical spectra to the individually measured metabolic parameters (tissue pH and lactate concentration) in the presence of inherent pathological variability.
13
SPECTROSCOPY BASICSIn general, spectroscopy is the use of the electromagnetic spectrum to perform physical or chemical analysis
14
PREVIOUS WORK
• Optical spectroscopy proposed by Lodder (UKY), Feld (MIT) and Jaross (Germany) to characterize morphological properties of atherosclerotic plaques such as thin fibrous cap, large lipid core
15
LACTATE AND PLAQUE
• Metabolite produced by activated macrophages• Studies show lactate is present in plaque (Kirk, Zemplenyi)
Anaerobic glycolysis: LDHPyruvate + NADH Lactate + NAD+
Overall anaerobic process: Glucose + 2ADP + 2Pi 2 Lactate + 2ATP +2H20 + 2H+
16
NIR Spectrum
0.15
0.2
0.25
0.3
0.35
0.4
0.45
1700 1800 1900 2000 2100 2200 2300 2400 2500
wavelength (nm)
arbi
trary
uni
ts (a
bsor
banc
e)
Near infrared absorbance of lactic acid
17
PLAQUE pHLarge scale, ex-vivo study on carotid plaques demonstrated metabolic heterogeneity in grossly pathological areas (Grascu, 1999)
Inflamed regions of plaque are lower in pH in the atherosclerotic Watanabe rabbit and human carotid plaques; plaque pH heterogeneity demonstrated (Naghavi, 2002)
18
DR. SOLLER’S LAB• Tissue pH can be measured by NIR spectroscopy in heart muscle
(Soller, Zhang 1998)• Lessons learned: volume of optical measurement >> volume of
reference measurement• Heterogeneity in a “large” tissue volume may be solved with
smaller optical probe
OPTICAL DESIGN
20
DESIGN PROCESS
Define of optical probe requirements Theoretical considerations of tissue optical
properties Monte Carlo simulations interpretation Building and testing several optical probes Depth penetration assessment
21
Optical Catheter System Diagram
• Optical fibers carry light to tissue• Light is reflected and/or backscattered
toward fibers that return light to spectrometer and tissue absorbance calculated
• Catheter geometry and optical coupling important
• Small source-receiver separations: light penetrates tissue while restricting volume interrogated
~2 mm
Light inTo spectrometer
wavelength
Abs
orba
nce
tissueinterface
22
THEORY• Monte Carlo Simulations
– Estimate light paths in complex absorbing and scattering medium
– Random events: reflection, absorption, scattering, or transmission
– Define grid geometry, specify tissue optical properties
Probe – tissue interfacen1=n2
dr= 25 m
dz=5 m
SourceDiffuse Reflectance(r)
23
0.00E+00
5.00E+01
1.00E+02
1.50E+02
2.00E+02
2.50E+02
3.00E+02
3.50E+02
0.000 0.005 0.010 0.015 0.020 0.025
radius (cm)
diff
use
refle
ctan
ce [1
/cm
2]
514 nm 633 nm 1064 nm
0.00E+00
1.00E+02
2.00E+02
3.00E+02
4.00E+02
5.00E+02
6.00E+02
7.00E+02
0.000 0.005 0.010 0.015 0.020 0.025
radius (cm)
diff
use
refle
ctan
ce (1
/cm
2)
470 nm 633 nm 514 nm 1050 nm 1064 nm
a) b)
normal atheroscleroticDiffuse Reflectance(radius)
Theoretical Depth PenetrationWavelength
(nm)Normal Aorta
(microns)Atherosclerotic Aorta (microns)
470 -- 417.5
514 1063 722
633 1338 1197
1050 1333 1187
1064 -- 927
24
OPTICAL EXPERIMENTS
• Compared signal-to-noise ratios (SNR) for several fiber types / configurations– Different core sizes / number of fibers– With or without optical windows– Source-receiver separations
25
PROBE GEOMETRY
Large OD
Probe 360 degree illumination w/ optical window
Forward illuminationNo optical window
Final probe W/ optical window
26
TISSUE PENETRATION STUDY1) Reference spectra collected for each optical configuration (50, 500 micron
separations).
2) Absorbance spectra collected with n-th slice of 50 micron tissue.
3) Second absorbance spectrum collected with n-th slice plus diffuse reflector. Both absorption and scattering attenuate tissue signal.
Optical Probe
1) 2) 3)
To spectrometer
Diffuse Reflector
~50 um slicesAortic tissue
Light source (source fibers)
Diffuse Reflector
27
0
200
400
600
800
1000
420 550 760 970 1450 1750 2250
wavelength (nm)
pene
tratio
n de
pth
(um
)
S-D separation = 0.05 mm S-D separation = 0.5 mm
28
FINAL PROBE DESIGN
• Using a source-receiver separation of 50 microns, adequate depth resolution could be achieved in plaque in both the visible and near-infrared
• Increasing the collection fiber core diameter size to 200 microns with improved transmission out to 2400 nm, higher signal-to-noise ratio is achieved by improving the fiber collection area by 4 times and collection efficiency
• Using a 0.5 mm thick quartz optical window fused on the common end, with forward-viewing optics, the signal-to-noise ratio would be further improved across all wavelengths
29
Fiber optic probe used for all optical determinations in this study.
1 cm
METHODS
31
Heating pad
Balance
Fiber optic probe
Gas lineMicro pH and reference junction electrodes
Thermistor
37C, >85% RH
Spectralon 50% reflectance standard
Tissue bath
Laboratory setup for all studies.
32
Humidified Incubator maintained at 37°C.
33
in-vitro Plaque Validation Study• Minimum Eagle’s Medium (MEM), pH 7.4, 5.6 mM glucose, 26.2 mM
NaHCO3, with non-essential amino acids was used (Invitrogen).
• Media equilibrated with 75% O2 / 5% CO2 gas mixture prior to plaque addition.
• Seven human carotid plaques (UMass Memorial IRB Approval #10041) were collected and placed immediately in 37°C media enclosed in a humidified incubator at 37°C.
• Two plaques that were not placed in the liquid media, only in the humidified air of the incubator, served as controls.
• Measurements were taken with a 0.5 mm OD multi-parameter sensor placed in the tissue (Diametrics, MN).
• Changes in tissue pH, temperature, PO2 and PCO2 over time were analyzed.
34
Control plaque with multi-parameter sensor in place.
35
Box-whisker plots for ∆pH / hour and ∆temperature / hour (top and bottom, respectively) for the control and test plaques. The change in pH and temperature over time is greater in the controls than the test plaques.
36
STABILITY REVIEW• Experiment time fixed max. 4 hrs • in-vitro experimental stability criteria met
– <0.03 pH units/hr and < 0.4°C /hr
• Tissue temperature in media > 32°C to ensure tissue viability
• Tissue values stable and different from media, controls• Plaques in oxygenated media had higher pO2 readings
versus control plaques• Thickness of plaque affected magnitude of pO2 readings• Unable to measure calcified areas over time
37
OPTICAL CALIBRATION • ~24 additional human carotid plaques were collected and placed in-vitro.
• Absorbance spectra (667 – 2500 nm) of each area were taken using Nicolet FTIR 670 spectrometer with fiber optic probe (Remspec, MA) for optical lactate determination.
• Tissue biopsies of the same area were taken using a 4-mm punch biopsy and immediately frozen in liquid nitrogen.
• Reference tissue lactate (LA) assayed using micro-enzymatic methods. Values reported as micromole LA per gram wet tissue.
• Matching spectra and reference values modeled by multivariate calibration techniques. R2 and the standard error of cross-validation (SECV) used to assess model accuracy.
38
OPTICAL CALIBRATION • Absorbance spectra (400-1100 nm) were collected for a smaller subset
of 14 plaques using a Control Development spectrometer and same optical probe for optical tissue pH determination.
• Reference tissue pH was measured using 750 um diameter micro-pH electrodes.
• Matching spectra and reference values modeled by multivariate calibration techniques. R2 and the standard error of cross-validation (SECV) used to assess model accuracy.
39
MODEL DEVELOPMENT
• Partial least-squares, factor analysis– Create calibration with as many points as possible
• Cluster analysis– Investigate (in)homogeneity of spectra
RESULTS
41
SPECTRA COLLECTION: Lactate• 82 raw absorbance spectra shown below (667-2400 nm)• Key features are water (~970, 1450 and 2000 nm), cholesterol and its
esters (~1750 nm).
“Thrombus/Red”
n=22
“Fatty/Yellow”
n=41
“Calcified/White”
n=19
42
SPECTRA COLLECTION: Tissue pH• 48 raw absorbance spectra shown below (400-1100 nm)• Key features are hemoglobin (550 – 575 nm) and water (~970 nm)
absorption
400 500 600 700 800 900 1000
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
wavelength (nm)
Abs
orba
nce
(arb
itrar
y un
its)
400 500 600 700 800 900 10000
0.5
1
1.5
2
2.5
3
wavelength (nm)
Abs
orba
nce
(arb
itrar
y un
its)
400 500 600 700 800 900 10000
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2
wavelength (nm)
Abs
orba
nce
(arb
itrar
y un
its)
“Thrombus/Red”
n=11
“Fatty/Yellow”
n=23
“Calcified/White”
n=14
43
Histogram of lactate reference measurements3.2 2.7 (mean SD)
n=82
0
5
10
15
20
25
30
35
0 0.2 0.5 1 2 5 8 10 More
lactate concentration (micromoles/gram tissue)
frequ
ency
44
Histogram of tissue pH reference measurements7.33 0.21 (mean SD)
n=48
0
2
4
6
8
10
12
14
6.70 6.80 6.90 7.00 7.10 7.20 7.30 7.40 7.50 7.60 7.70 More
pH units
frequ
ency
45
REFERENCE MEASUREMENTS
• No spurious correlations between measured variables• Tissue temperature, experiment times within validated
experiment parameters• Pathology subjective
Variables Correlation Coefficient between data
Tissue Lactate – Tissue Temperature
0.03 (n=82)
Tissue pH – Tissue Temperature
0.00004 (n=48)
Tissue Lactate – Tissue pH 0.0008 (n=48)
46
RESULTS: Tissue Lactate• 6-Factor model from 17 points• Wavelength regions
contributing to model: – 2030 – 2330 nm
• The R2 of the determination for optical lactate (LA) calibration = 0.83.
• Estimated accuracy ~ 1.4 micromoles LA/gram tissue.
47
2000 2050 2100 2150 2200 2250 2300 23501
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
2000 2050 2100 2150 2200 2250 2300 23501.35
1.4
1.45
1.5
1.55
1.6
1.65
1.7
1.75
1.8
1.85
2000 2050 2100 2150 2200 2250 2300 23500.8
0.9
1
1.1
1.2
1.3
1.4
1.5
A
B
C
Clustering solution for 82 spectra for the optical determination of lactate. Cluster A – 45 spectra, B – 31 spectra, and C – 6 spectra. Cluster A contained the first 21 calibration spectra collected.
48
RESULTS: Tissue pH• 3-Factor model from 17 points• Wavelength regions contributing
to model: – 1: 400 – 615 nm– 2: 925 – 1890 nm– 3: 2044 – 2342 nm
• The R2 of the determination for optical tissue pH calibration = 0.75.
• Estimated accuracy ~ 0.09 pH units.
Region 1
Region 2 Region 3
6.80
7.00
7.20
7.40
7.60
7.80
6.80 7.00 7.20 7.40 7.60 7.80
electrode pH
NIR
pH
49
400 500 600 700 800 900 1000 11000
0.5
1
1.5
2
2.5
400 500 600 700 800 900 1000 11000
0.2
0.4
0.6
0.8
1
A
B
Clustering solution for optical determination of tissue pH. Two clusters: A contains 39 spectra, B contains 9 spectra. The underlying pathology in cluster B was identified as all thrombotic points.
50
DISCUSSION
• Lactate model on portion of entire data set– Further factor analysis showed spectra weakly
associated with theoretical lactate peaks – Number of factors in model too high for # of samples
used; need more samples• Tissue pH model on portion of entire data set
– Further factor analysis showed spectra associated with Hb and water peaks, evidence of pH-induced shift
– Number of factors acceptable for # of samples used
51
CONCLUSIONS1) A small fiber optic prototype can make optical measurements
in-vitro for the assessment of metabolically active plaque in a defined region of tissue (< 1 mm3 volume).
• Hypothesis accepted
2) in-vitro experimental factors can be assessed to their importance in affecting the optical calibration accuracy. The tissue temperature, experiment time, and gross pathology are identified a priori.
• Hypothesis accepted
52
CONCLUSIONS3) Mathematical models can be developed which relate the
corresponding optical spectra to the individually measured metabolic parameters (tissue pH and lactate concentration) in the presence of inherent pathological variability.
• Hypothesis rejected for large n; pending work
– Limited feasibility of models generated
– Pathological variability large
– Unmodeled tissue variability
– Lactate reference method precision
– Optical tissue volume >> real tissue pH heterogeneity
– Long-term spectrometer drift could not be ruled out
53
FUTURE WORK
• Considerable in-vitro work needs to continue• Other clustering algorithms, pre-processing methods• Reference lactate measurement precision• Reduce unmodeled variability, better tissue model• Larger data sets• Spectrometer stability • Rigorous acceptance criteria must be met before use
in-vivo animals or humans
54
ACKNOWLEDGEMENTS
Texas Heart Institute, Center for Vulnerable Plaque Research/UT HoustonDr. S.W. Casscells Dr. Silvio LitovskyDr. Morteza Naghavi
Department of Surgery, University of Massachusetts Medical SchoolVascular Surgeons: Dr. P. Nelson, Dr. B. Cutler, Dr. A. Fox and Dr. M. RohrerDr. Babs R. Soller Dr. Patrick Idwasi
This work was supported by US Army DREASM Grant