raman and ir spectroscopy - university of illinois at urbana-champaign
TRANSCRIPT
6/8/2009
1
Beckman InstituteBeckman InstituteBeckman Instituteat The University of Illinois
Vibrational Spectroscopy Vibrational Spectroscopy and Imaging
Rohit Bhargava
Spectroscopy generalization
Known radiation goes in Collect exiting radiation
Compare the light that went in with the light that comes out
A general definition of spectroscopy is the study of the interaction between radiation and an analyte as a function of wavelength.
Interacts with sample
Spectroscopy (Radiation)
Radio
1010-108 cm-1: Re-arrangement of elementary particles (gamma ray)108-106 cm-1: Transitions of inner electrons (X-ray)106-104 cm-1: Transitions of valence electrons (UV-vis spectroscopy)104-102 cm-1: Transitions between vibrational levels (IR and Raman)102-100 cm-1: Transitions between rotational levels (Microwave, IR, Raman, THz)100-10-2 cm-1: Transitions between electron spin levels (magnetic – ESR)10-2-10-4 cm-1: Transitions between electron nuclear spin levels (magnetic – NMR)
Cosmic Gamma X
UV IR Micro UHF
Short
Medium
Long
Ultra violet Infrared
Near Mid Far
1 400 750 2,500 16,000 1,000,000 nm
Vis
6/8/2009
2
Spectroscopy of liquid water
200 250 300 350 4000
0.01
0.02
0.03
0.04
0.05
0.06
0.07
400 500 600 7000
0.01
0.02
0.03
0.04
0.05
0.06
0.07
1000 1500 2000 2500
20
40
60
80
100
120
800 1000 1200
0
0.5
1
1.5
2
2.5
4000 6000 8000 10000 12000 14000 160000
2000
4000
6000
8000
10000
12000
14000UV Spectrum
Wavelength (nm)
Abs
orpt
ion
(1/c
m)
Visible Spectrum NIR Spectrum IR Spectrum
Ultra violet Infrared
Near Mid Far
1 400 750 2,500 16,000 1,000,000 nm
Vis
Wavelength (nm)
Why is water blue?
Gas Assignment liquid
3651 v1 3400 3755 v3
5332 v2+v3 51507251 v1+v3 69008807 v1+v2+v3 840010613 2v1+v3 10300
•Gas and liquid have different positions
•Combinations and overtones are weaker
• Significant example of absorption determining color10613 2v1+v3 10300
13831 3v1+v3 13160 (760 nm)
14319 v1+3v3 13510 (740 nm)
15348 3v1+v2+v3 15150 (660 nm)
15832 v1+v2+3v3 15150 (660 nm)
16822 3v3+2v2+v1 weak
determining color
•What is the effect of cold water? Ice?
So, what are these vibrational modes
Symmetricalstretching
Antisymmetricalstretching
Scissoring Rocking Wagging Twisting
Simplest Model of a Diatomic Molecule: beads and spring harmonic oscillatorSimplest Model of a Diatomic Molecule: beads and spring harmonic oscillator
- Hooke’s Law Restoring force is proportional to displacement (K~105 dynes- Newton’s law: force if mass x acceleration
012
~ 1012-1014 Hz
6/8/2009
3
Spectroscopy : Molecular Basis
: 0 2 0
: 0 2 0
: 0 2
: 00
0 …….
0 0 2 012 0 0
0 0 2 0 2 0
Raman Scattering
:
Note: - change in polarizability, excitation frequency and shift by vibrational frequency- Two types of shifts to lower (Stokes) and higher (anti-Stokes): which population is favored?
Raman Spectroscopy: Basic Concept
785
Detector
First observed by Sir C. Venkata Raman in 1928 using sunlight and photographic filters (won Nobel price in physics in 1930). Nobelprize.org
Ultra violet Infrared
Near Mid Far
1 400 750 2,500 16,000 1,000,000 nm
Vis
785nm Sample
Raman Spectroscopy: Basic Concept
Electronic States
UV-VISAbsorption
http://en.wikipedia.org/wiki/File:Raman_energy_levels.jpg
Chance of occurring is Wavelength DependantResonance Raman Scattering?
6/8/2009
4
Raman Spectroscopy: Basic Concept
Electronic States
UV-VISAbsorption
http://en.wikipedia.org/wiki/File:Raman_energy_levels.jpg
If light is not absorbed… the majority of the photons pass through the sample by means of Rayleigh Scattering
Raman Spectroscopy: Basic Concept
Electronic States
UV-VISAbsorption
http://en.wikipedia.org/wiki/File:Raman_energy_levels.jpg
Some of the photons (~1/10,000,000) that are not absorbed will pass through the sample and undergo Stokes Raman Scattering
Raman Spectroscopy: Basic Concept
Electronic States
UV-VISAbsorption
http://en.wikipedia.org/wiki/File:Raman_energy_levels.jpg
Even fewer photons (temperature dependant) that are not absorbed will pass through the sample and undergo Anti-Stokes Raman Scattering
6/8/2009
5
Raman Spectroscopy: Basic Concept
Molecules in the sample
785 nm
Photon 786nm787nm788nm789nm790nm791nm792nm
780nm781nm782nm783nm784nm
… …
Ultra violet Infrared
Near Mid Far
1 400 750 2,500 16,000 1,000,000 nm
Vis
785nm Sample
Detector
Detector Array
793nm794nm795nm796nm797nm798nm799nm800nm801nm802nm803nm804nm805nm806nm… …
Raman Spectroscopy: Basic Concept
Molecule in the sample
786nm787nm788nm789nm790nm791nm792nm
93
780nm781nm782nm783nm784nm
… …
Ultra violet Infrared
Near Mid Far
1 400 750 2,500 16,000 1,000,000 nm
Vis
785nm Sample
Detector
Detector Array
793nm794nm795nm796nm797nm798nm799nm800nm801nm802nm803nm804nm805nm806nm… …
786nm787nm788nm789nm790nm791nm792nm
93
Raman Spectroscopy: Basic Concept
Molecules in the sample
785 nm LightGets rejected
780nm781nm782nm783nm784nm
… …
Rayleigh Scatter
Detector Array
793nm794nm795nm796nm797nm798nm799nm800nm801nm802nm803nm804nm805nm806nm… …
Ultra violet Infrared
Near Mid Far
1 400 750 2,500 16,000 1,000,000 nm
Vis
785nm Sample
Detector
6/8/2009
6
Raman Spectroscopy: Basic Concept
Molecules in the sample
785 nm
Photon 786nm787nm788nm789nm790nm791nm792nm
780nm781nm782nm783nm784nm
… …
Ultra violet Infrared
Near Mid Far
1 400 750 2,500 16,000 1,000,000 nm
Vis
785nm Sample
Detector
Detector Array
793nm794nm795nm796nm797nm798nm799nm800nm801nm802nm803nm804nm805nm806nm… …
Raman Spectroscopy: Basic Concept
Molecule in the sample
786nm787nm788nm789nm790nm791nm792nm
93
780nm781nm782nm783nm784nm
… …
Ultra violet Infrared
Near Mid Far
1 400 750 2,500 16,000 1,000,000 nm
Vis
785nm Sample
Detector
Detector Array
793nm794nm795nm796nm797nm798nm799nm800nm801nm802nm803nm804nm805nm806nm… …
786nm787nm788nm789nm790nm791nm792nm
93
Raman Spectroscopy: Basic Concept
Molecules in the sample
781 nm LightPassed through to the detector
780nm781nm782nm783nm784nm
… …
Anti-stokes Raman
Detector Array
793nm794nm795nm796nm797nm798nm799nm800nm801nm802nm803nm804nm805nm806nm… …
Ultra violet Infrared
Near Mid Far
1 400 750 2,500 16,000 1,000,000 nm
Vis
785nm Sample
Detector
6/8/2009
7
Raman Spectroscopy: Basic Concept
Molecules in the sample
785 nm
Photon 786nm787nm788nm789nm790nm791nm792nm
780nm781nm782nm783nm784nm
… …
Ultra violet Infrared
Near Mid Far
1 400 750 2,500 16,000 1,000,000 nm
Vis
785nm Sample
Detector
Detector Array
793nm794nm795nm796nm797nm798nm799nm800nm801nm802nm803nm804nm805nm806nm… …
Raman Spectroscopy: Basic Concept
Molecule in the sample
786nm787nm788nm789nm790nm791nm792nm
93
780nm781nm782nm783nm784nm
… …
Ultra violet Infrared
Near Mid Far
1 400 750 2,500 16,000 1,000,000 nm
Vis
785nm Sample
Detector
Detector Array
793nm794nm795nm796nm797nm798nm799nm800nm801nm802nm803nm804nm805nm806nm… …
786nm787nm788nm789nm790nm791nm792nm
93
Raman Spectroscopy: Basic Concept
Molecules in the sample
790 nm LightPassed through to the detector
780nm781nm782nm783nm784nm
… …
Stokes Raman
Detector Array
793nm794nm795nm796nm797nm798nm799nm800nm801nm802nm803nm804nm805nm806nm… …
Ultra violet Infrared
Near Mid Far
1 400 750 2,500 16,000 1,000,000 nm
Vis
785nm Sample
Detector
6/8/2009
8
786nm787nm788nm789nm790nm791nm792nm793nm794nm795nm796nm797nm798nm799nm800nm801nm
Billions of 785nm photonsMolecules in a soybean
Raman Spectroscopy: Basic Concept
Ultra violet Infrared
Near Mid Far
1 400 750 2,500 16,000 1,000,000 nm
Vis
785nm
Detector
Detector Array
801nm802nm803nm804nm805nm806nm… …
Soybean
5
6
7
8
9
nten
sity
(A.U
.)
786nm787nm788nm789nm790nm791nm792nm793nm794nm795nm796nm797nm798nm799nm800nm801
Raman Spectroscopy: Basic Concept
Detector A
rray
786nm787nm788nm789nm790nm791nm792nm793nm794nm795nm796nm797nm798nm799nm800nm801nm802nm803nm804nm805nm806nm
…
…
400 600 800 1000 1200 1400 1600 18000
1
2
3
4
Raman Shift (cm-1)
Ram
an I
n
810nm 915nm
Detector Array
801nm802nm803nm804nm805nm806nm… …
5
6
7
8
9
Raman Spectroscopy: Raman Shift
785nm = 0.0000785cm
1
0.0000785cm= 12738.9 cm-1
12738.9 cm-1 – 400 cm-1 = 12338.9 cm-1
Excitation Frequency
Wavelength of 400 cm-1
Raman shift in nm
Soybean
400 600 800 1000 1200 1400 1600 18000
1
2
3
4
Raman Shift (cm-1)810nm 915nm
1
12338.9 cm-1= 0.0000810cm
810nm
***The Raman shift is independent of excitation frequency… the spectral band positions will be the same whether you excite with a 785nm laser or a 532 nm laser
6/8/2009
9
Raman Spectroscopy: Raman Shift
Important advantage of Raman… wavelength select ability to avoid light absorption and fluorescence… however
Wavelength Dependence of Raman Cross-Section
the higher the wavelength the lower the Raman efficiency (Raman Cross Section).
3400 cm-1 O-H stretch of liquid water
Appl. Opt. 36, 2686-2688 (1997)
Raman Spectroscopy: Selection Rules
Raman Bands result from the Vibrations of specific chemical Bonds.
Soybean band assignments:
800-900 C-C Backbone
1003 Phe Ring breathing
Wave number Molecule Assignment
1010-1200 C-C Backbone
1265 RC=ONH2 Amide III
1450 CH2 Stretch
1660 RCH=CHR Cis
1670 RCH=CHR Trans
1750 RC=OOR Ester
NIR Spectra
Raman Spectroscopy: Examples
4
5
6
7
8
9
n In
tens
ity
(A.U
.)
Raman Spectrum
400 600 800 1000 1200 1400 1600 18000
1
2
3
4
Raman Shift (cm-1)
Ram
an
810nm 915nm
IR Spectrum
2500nm 11,000nm
6/8/2009
10
Raman Spectroscopy: Instrumentation
Las
CCD
Kaiser Optical
ser
Raman Spectroscopy: Instrumentation
grap
h
-This is a axial dispersive spectrograph
-Reflective spectrographs are also common
-Interferometers are used for FT-Raman
Kaiser Optical
Spec
tro
Raman Spectroscopy: Instrumentation
ctor
-Charged-coupled devices (CCDs) are the most common detectors(256x1024 pixel array)
-FT Raman is used for frequencies where CCDs have low quantum efficiency
Det
ec
Spectral Dimension
Hei
ght
6/8/2009
11
Raman Spectroscopy: Advantages and Disadvantages
Advantages:
-Little sample preparation (Polishing and fixing to a slide is common)
-Not sensitive to water (Good for biological samples)
-High chemical specificity (Narrow spectral bands)
-Qualitative and Quantitative information
-Non-destructive (A measurement does not chemically or physically change the sample)
-Can take measurements on solids, liquids, or gases
-Measurements are taken without touching the sample (Remote sensing)
-Easily coupled with fiber-optics
Raman Spectroscopy: Advantages and Disadvantages
Disadvantages
-Acquisition times tend to be longer than other techniques (real time measurements have been demonstrated… but something like video rate imaging is not yet a reality)
-Raman signal tends to be weak
-Raman signal is often mixed with a fluorescent background signal, which can make signal processing difficult.
-High laser powers and burn delicate samples
Raman Microscopy
Spectrograph & CCD
785 nm Laser
Dichroic mirror
Microscope
1.2
)
0
0.2
0.4
0.6
0.8
1
307
342
376
411
446
480
515
549
584
619
653
688
722
757
792
826
861
895
930
965
999
1034
1068
1103
1138
1172
1207
Wavelengh (nm)
Tran
smitt
ance
(%)
Wavelength (nm)
6/8/2009
12
Confocal Raman Spectroscopy
Confocal Raman Microscopy
Tabaskblat et al., Appl. Spec 1992
Hyperspectral Raman Mapping
Spectrograph & CCD
785 nm Laser
Fram
e G
rabb
er
Microscope
PhAT Probe
~~~~Fiber Launcher
Illumination Collection
~~~~ 1
50
Collection fibers are transposed into a line when to enter the spectrograph
White Light Illuminated Region
Bone
PMMA50μm
Hyperspectral Raman Mapping
1
2
3
4
5
6
78
9
10
11 1213
14
15
16 17
18
1920
21
22
23
24
2526
27
2829
30
31
32
3334
35
36
3738
39
40
4142
43
44
45
4647
48
49
50
Raman map
Raman Shift (cm-1)
Ram
an In
tens
ity (A
.U.)
Raman Shift (cm-1)
Ram
an In
tens
ity (A
.U.)
6/8/2009
13
Spatially Offset Raman Spectroscopy
1
2
3
4
5
6
7 8
9
10
11
12
13
14
15
16
17
18
19
20
2122
2324
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
4041
42
43
44
45
46 47
48
49
50
ID
OD
Illumination ring
Spacing
Spectrograph& CCD
785 nm Laser
Fiber
Probe
Converging
Collection Fibers
Collection Fibers
g
launcher
Sample/Specimen
Fiber collim
ator
Axicon
Converging lens
Diverging
lens(es)
Dichroicmirror
Converging lens
Illumination
Collection
Spatially Offset Raman Spectroscopy
ensi
ty (A
.U.) Transcutaneous
Exposed boneBone factor
Anesthesia
Wat
er c
oolin
g
Schulmerich et. al., Appl. Spectrosc., 63(3) 286-295
Ram
an In
te
Raman Shift (cm-1)
Other Raman Techniques
-Surfaced enhanced Raman spectroscopy (SERS)
-Tip enhanced Raman spectroscopy (TERS)
-Resonance Raman spectroscopy
-Raman tomography
-Raman imaging
-Coherent anti-stokes Raman spectroscopy (CARS)
-Stimulated Raman spectroscopy
-Others I can’t think of at the moment and much more to come!
6/8/2009
14
Infrared Spectroscopy
Theory of Infrared – Chemical bonds
• Infrared spectroscopy : Principle: chemical bonds rotate or vibrate at specific frequencies (Group theory basis of IR Spectroscopy)
• Practical AspectsThe frequency at which resonance occurs depends on the properties of the chemical bondFactors include shape of molecular bond, energy levels and mass of atoms.Chemical bonds can be divided into those which are IR active and those which are IR inactive
Molecular Basis of IR Absorption
• Direct absorption of light
• Beer’s law relates absorption to concentration0 log 0
Δ
C‐Hstretching C‐H
bending
Electronic States
UV-VISAbsorption
• Selection rule– Dipole moment (symmetrical mode: diatom)
Increasing wavenumber (energy, frequency)
Increasing absorption of IR radiation
Increasing wavelength
C‐Cbending
6/8/2009
15
• To be an IR active mode, the motion must have a change in the electric dipole moment of the bond
• ImplicationsThe intensity of absorbance peaks is related to size of dipole moment.Bonds with higher dipole moments tend to be covalent bonds with highly different electronegativities e.g C=O
Theory of Infrared – Chemical bonds
highly different electronegativities e.g C O Symmetrical bonds are typically IR inactive e.g. C-C, N-NThe same chemical bond can also have multiple modes of vibration e.g. phosphate has a symmetrical and antisymmetrical modes
Infrared Spectra• Vibrational frequencies are directly resonant with optical frequencies – vibrational transition is energy matching
Infrared Spectroscopy
• IR was discovered in 1801 by William Herschel who split the EM spectrum using a prism – noted increase in temperature beyond the red part of visible spectrum
• 1930’s began to be exploited for spectroscopy studies
• Two mains types of spectrometer – Dispersive and Fourier TransformDi i h i f li h d h f• Dispersive uses monochromatic source of light and changes frequency over time by moving grating, mirror or detector. Originally used prisms and later grating.
• Largely replaced by Fourier Transform – Allows for all frequencies to be measured at once: Fellgetts advantage – No slits: higher throughput
6/8/2009
16
IR Spectroscopy: Theory
1 13
2 1
0.5 1 2∞
∞
2 and 2’
3
∞
0 0.5 1 2∞
∞
or
Detector
Sample1 2
3
4Focusing mirror
4000 3200 2400 1600 800 0
1.2
1.4
1.6
1.8
2.0
2.2
Retardation (cm)
I0
0.0
0.5
1.0
1.5
2.0
2.5
BD,0
-0.05 0.05 0.1 0.15 0.2 0.25
1
Cb
W
Computation (FT)
0.5 1 2∞
4or
,0 0.5 2∞
∞
, 0.5 2∞
∞
, 2 , 2∞
0
Moving mirror
Beamsplitter
Fixed mirror
BroadbandSource δ= 0
δ= δmax
2’
Wavenumber (cm-1)Cb:Centerburst, W:Wings
4000 3500 3000 2500 2000 1500 100010
100
1000
10000
Abs
orpt
ion
Coe
ff(c
m-1
)
Wavenumber (cm-1)
WaterTissue
A popular commercial instrument
FTIR Microscope
combining the FTIR spectrometer with a microscope
Point spectra or Mapping/Imaging?
A) Single Point Detector(mapping)
B) Linear Array Detector(imaging)
C) Focal Plane Array(imaging)
Example improvements in speed1 16 16,384
Single point 16 detector linear array 128x128 Focal Plane Array
In fact, Imaging is even quicker and allows for diffraction limit measurements
6/8/2009
17
FT-IR spectroscopy Imaging
0.4
1.6
100 μm
0.6
a.u.
)
y
• Typical characteristics– Wavelengths (2048 elements over 2.5 – 12.5 μm), x, y typically ~1024
• Computation is essential to recover data– Manual examination is prohibitive
• Trade-offs: spatial coverage vs. resolution, spectral resolution vs. signal-to-noise ratio, time vs. data quality/size vs. information
3200 2800 2400 2000 16000.0
0.2
0.4
Abs
orba
nce
(a
Wavenumber (cm-1)
x
z
FTIR Chemical Imaging
Sample
Aperture
Aperture
CCD Visible Detector
Mirror
Stage
10 -310 -2
10 -1
11010 2
10 3
10 4
10 5
MAPPING 3
1
2
Tim
e (
hrs
)
A
4
5 7a
6
8910
11
7b
8 cm-1 , SNR of 1000:1, 1 Mpix, 6 x6 micron res., 1 cm x 1 cmBlue – hardware, red – interferometry/software, green – emerging technology
Microscope
Mirror
Visible LightSource
Rapid-ScanInterferometer
1. E.N. Lewis et al. Anal. Chem. 67, 3377-3386 (1995) ; 2. Snively et al. Appl. Spectrosc. (1998); 3. Bhargava et al. Appl. Spectrosc. 53, 1313-1322 (1999); 4. Snively et al. Opt. Lett., 24; 1841-1843 (1999); 5. Bhargava et al. Appl. Spectrosc. 54, 486-495 (2000); 6. Perkin-Elmer Inc.; 7a. Varian Inc. ; 8. NIH-FBI camera; 9. Reddy et al (To be submitted); 10 & 11 – under development.
1992 1996 2000 2004 2008
10 -4
Year
Settings and Parameters
• Sample preparation is critical!• Signal to Noise – Number of co-additions
A number of important considerations must be taken into account prior to experiments;
• Mirror speed• Spectral resolution• Spatial resolution• Point spectra vs. Point Mapping vs. Imaging
6/8/2009
18
Sample preparation• Transmission or reflection
• IR transparent substrates – BaF2, CaF2• IR reflective substrates – Gold, MirrIR Slides• Reflection is low-cost, however reduction in spectral quality and
increase in scattering artifacts
Signal to noise• It is important to get high quality noise-free data –
This can involve multiple strategies;– Better sample preparation (more or less sample)– More scans and averaging– Reduce Michelson mirror speed
• Spectral resolution is important this is the number of• Spectral resolution is important, this is the number of points within a spectra, the fewer the faster the scans but potential for loss of information
• Spatial resolution is also important, the higher the spatial resolution, the lower the signal.Take home message; Many factors must be adapted depending on what
you are interested in
Application 1: Forensic Sciences
0.005
Absorbance at 2920 cm-1
Body oils
1 mm
0.000
Absorbance at 1568 cm-1
Skin Flakes
6/8/2009
19
An example of imaging…..
• Polymer crystallization….
900 1000 1100 1200 1300 1400 1500
-0.50
-0.25
0.00
0.25
0.50
0.75
1.00
1.25
Abso
rban
ce (a
.u.)
Wavenumber (cm-1)
Amorphous
Crystalline
Difference
9 s 21 s 24 s 27 s
0.75
0.45
0.60
0.90
Wavenumber (cm )
30 s 36 s 45 s 75 s
0.15
0.30
10 100
0.0
0.2
0.4
0.6
0.8
1.0
Abs
orba
nce
(a.u
.)
Time (s)
A B C D Average
A
B
C
D
Maximum Rate Maximum Rate Time (s)
0.075
0.015
0.045
0.030
0.060
0.090
60
20
40
30
50
70
Motivation
• Cancer pathology– Prostate cancer as a paradigm – 1
in 6 men– Biopsies: >1 million annually with
disease ~20%– Diagnoses: >200, 000 annually
with lethal ~20%G di i bj ti i bl
Screening
Biopsy
Is it suspicious?
Is it cancer?
D I A G N O S T I C
– Grading is subjective, variable, leads to conflicting therapy routes
– Prognosis tools are not perfect –97% undergo therapy
– “Holy Grail” of oncologic pathology– Primary evaluative standard for
research
• Manual recognition in stained tissue
Therapy
Follow-up
Diagnosis
What is the grade? Stage?Will it metastasize?
Is adjuvant therapy needed?
Has cancer recurred?
R E S E A R C H
Cancer and Diagnostic Process
Biopsy
• No stains, no manual decisions
Epithelium
Stroma
BenignGrade 3Grade 4
American Cancer Society, est. 2008
6/8/2009
20
From Data to Knowledge
Information
x=2048, y=2048, z=2048, t ~ ms to days
■ Model based design of experiments■ Hypothesis driven analysis – supervised data analysis■ Biologically inspired statistical pattern recognition of spectra■ Approach – Model, Train algorithm, Classify, Validate
x
y
z
Analytical Approaches
The types of analysis available to spectroscopists can be divided into two groups;
Un-SupervisedAll data has the same weighting and no-prior information about what the
t d t i kUn Supervised
Supervised
spectra corresponds to is known-Useful for interest of discovery
Known information about data e.g. classes is used for data analysis
-Useful for classification (more amenable to clinical setting?)
PLSPCA
Analytical Approaches
Which MVA approach?
Bayesian
SIMCA
ANNKNNLDA
FUZZYHCA
Important to determine which are best for analysis of point spectra and which for images
6/8/2009
21
Model
Potential Metric Set
Acquired Data
Epith
elium
Fib. Strom
a
Pale Strom
a
Smooth M
uscle
Stone
Lymphocytes
Endothelium
Nerve
Ganglion
Blood
Prostate Histology
Integrative classification process
Gold Standard
Algorithm(Modified Bayesian)
Optimized PredictionAlgorithm
Optimization
Optimal Metric Set
Calibration/ValidationStatistics
Sensitivity Analysis
Visualizing Classification Origin and Variance
met
rics
patients
3
6
9
12
1
4
7
10
2
5
8
11
15
18
13
16
14
17
96 3 104 5 6 1171 8 12 15 181316 14 17 19 20 22 262321 24 25 2927 30 343128 32 33 35 36 37 38
Determining outliers – quality controlVariance and classification
Spectral Diversity and Clustering Control and significant Variables
Examples of FTIR for Biomedical applications
Fundamental biological processes
ApoptosisNecrosisStem cell differentiationToxicologyDisease initiation and progression
Clinical Applications
p g
Disease diagnosisDisease prognosis
6/8/2009
22
FTIR-Biomedical Applications
Biochemical-cell fingerprint region
Integrative Process and Key Technologies
Hi-fi, fast FTIR Imaging
Samples
Custom Tissue Microarrays
* Kononen et al. Nat. Med. 4, 844-847 (1998).Levin and Bhargava Annu. Rev. Phys. Chem. 56, 429-474 (2005).Kong and Bhargava, Submitted (2009): Living tissue arrays
* Lewis et al. Anal. Chem. 67, 3377-3381 (1995).
Biopsy
Pattern Recognition Algorithms
Imaging Data
Diagnosis
Statistical Validation
Llora , Priya, Bhargava Nat. Computing In Press (2009).Bhargava et al. Biochim Biophys Acta. 1758, 830-845 (2006).Reddy and Bhargava (2009): Statistical Model
Fernandez et al. Nat. Biotechnol. 23, 469-474 (2005).Bhargava et al. Nat. Biotechnol. 25, 31-33 (2007).Bhargava Anal. Bioanal. Chem. 389, 1155-1169 (2007).
Lewis et al. Anal. Chem. 67, 3377 3381 (1995).Bhargava and Levin Anal. Chem. 73, 5157-5167 (2001)Bhargava and Reddy Anal. Chem., Submitted (2008): Noise rejection
Prognosis
6/8/2009
23
Application 1: Prostate HistologyValidation of Prostate Histology: 970 samples, 10 million spectra
EPITHELIUM
FIBROUS STROMA
MIXED STROMA
SMOOTH MUSCLE
NERVE
GANGLION CELLS
BLOOD
EPITHELIUM
FIBROUS STROMA
MIXED STROMA
SMOOTH MUSCLE
NERVE
GANGLION CELLS
BLOOD
EPITHELIUM
FIBROUS STROMA
MIXED STROMA
SMOOTH MUSCLE
NERVE
GANGLION CELLS
BLOODBLOOD
LYMPHOCYTES
CORPORA AMYLACEA
ENDOTHELIUM
BLOOD
LYMPHOCYTES
CORPORA AMYLACEA
ENDOTHELIUM
BLOOD
LYMPHOCYTES
CORPORA AMYLACEA
ENDOTHELIUM
1.3 mm
Fernandez, Bhargava, Hewitt and Levin Nat. Biotechnol., 23, 469-474 (2005)
Prostate Cancer Diagnosis
• Overall pixel accuracy ~ 88.5% ; Heterogeneity in samples?• 1 cancer sample classified as benign (71)• 1 benign sample classified as cancerous (69)• Sensitivity and specificity exceeding human capabilities• Large validation studies underway
Array – 80 Patients Array – Histology Pathology Design Pathology Result
Whats in store?
• Label-free methods– Talks on Monday focus on applications
• Lab toursTheory• More information Theory
Instrumentation
SamplingData Analysis
Applications