1. Functional Imaging with Diffuse Optical Tomography
Mark A. Elliott, PhD
Department of Radiology
University of Pennsylvania

2. Overview Mechanisms of functional imaging with NIR light
Methodology of fNIR
Comparison with and without Difuse Optical Tomography (DOT)

3. Methods for Imaging Neural Activity
metabolic response
FDG PET
- ATP tightly regulated
- glucose consumption
electrical activity
- excitatory
- inhibitory
- soma action potential
- oxygen consumption
H215O PET
hemodynamic response
- blood flow
fNIR
- blood volume
electrophysiology
- blood oxygenation
fMRI
EEG
MEG
Perfusion MRI

4. Vascular Sensitivity of fMRI and fNIR
Venous
Arterial II I
fNIR
Intravascular
Perfusion MRI II IV
fMRI
III I
Extravascular
III IV
Vessel Size

5. Mix of blood volume, blood flow, and O2 metabolism
Vascular Response
fMRI vs fNIR
fMRI
fNIR
Spatial Resolution
8-27 mm3
“Blobs”
1-10 cm3
Temporal Resolution
Slow (1-2 sec)
Fast (50 Hz)
important?
Measurement parameter
Mix of blood volume, blood flow, and O2 metabolism
[Hb] and [HbO]

6. Mechanisms of fNIR: Overview
fNIR = functional Near InfraRed
Measure changes in infrared light absorption and scattering
Primary source of signal contrast  [Hb] and [Hb0]
Biological tissue is highly scattering in NIR window
Primarily used in vivo as a spectroscopic modality
Not used to produce true images
DOT = Diffuse Optical Tomography
Methods for accurate image reconstruction

7. Mechanisms of fNIR: Absortion of [Hb] and [Hb0]
Water Absorption
Near infrared “window” ~ nm
Water absorption is mimized
Hemoglobin species are dominant absorbers
[Hb] & [HbO] Absorption

8. Mechanisms of fNIR: Beer-Lambert Law
Beer-Lambert law models ballistic photon propagation in absorbing media
Transmittance, T = I/Io
Absorbance, A = -log(I/Io)
Beer-Lambert Law:
A =  [X] d
where:
d = distance between I0 and I
 = absorptivity (M-1 cm-1)
[X] = concentration of absorber (M) d Io I
solution [X]

9. Mechanisms of fNIR: Modified Beer-Lambert Law
Photons travelling through biological tissue are highly scattered (not ballistic)
Scattering adds to “pathlength” travelled by photons
Detector
Source
Fat
photon path
Muscle d
shallow
deeper
Modified Beer-Lambert Law: (
A = -log(I/Io) =  [X] d  DPF + G
where:
DPF = differential pathlength factor
G = Scattering loss factor (generally unknown)
I am going to tell brief information of NIR spectroscopy. NIR is one of detecting and charactering a tissue non invasively and it is safe.
It use four wavelength, ,830,977 nm optical source. Light absorption and scattering concentration depends on level of oxy and deoxy hemoglobin level. This graph shows absorption coefficient of oxyhemoglobin and deoxyhemoglobin is varying in different wavelength. Absorption at 760nm is predominately from deoxigenated hemoglobin and abortion at 850 nm is predominately from oxygenated hemoglobin.
Because human tiisue is multi tissue we need muti channel of source.
Source-detector spacing influences depth penetration

10. Mechanisms of fNIR: Measures Changes in [Absorber]
Scattering factor, G, is unknown
Absolute concentrations are not derivable
Can measure changes in [Hb] & [HbO]
Need baseline assumption or independent measure of [Hb]
Measure [Absorber]
A2–A1 = -log(I2/I1) =  [X] d  DPF
where:
A2,A1 = absorption measured at two time points

11. fNIR Methodology: Tissue Penetration
NIR light penetration into biological tissue allows for surface imaging
Penetration increases with source light intensity
Limits on safe levels of source light intensity (~1mW/mm2)
SNR  sqrt(Io)
Highly sensitive detectors (PMTs) allow 2-6 cm deep probing

12. fNIR Methodology: Quantitation of Multiple Chromophores
Multiple absorbers ([Hb], [Hb0])  multiple wavelengths
Extension of MBLL to multiple absorbers:
(MBLL):
A1 = (Hb 1[Hb] + HbO1[HbO])  d  DPF
A2 = (Hb 2[Hb] + HbO2[HbO])  d  DPF
1 2 3
Source illumination is time or frequency
multiplexed at several wavelengths.

13. fNIR Methodology: Temporal Resolution
Extremely high temporal resolution possible
Practical systems ~ 10 – 100 Hz
fMRI ~ 1-2 Hz
Hemodynamic changes are slow ~ 2-5 sec
Fiber-optic systems for simultaneous fMRI
Fast signal – cell conformation and swelling
Scattering changes > 10 Hz
Extremely low signal
Ellusive to date
from Strangeman Biol Psych 2002

14. fNIR Methodology: Spatial Localization
Discrete arrays of sources and detectors
# voxels = # sources  # detectors
Typical systems  10 – 100 voxels
Poorly localized “blobograms”
Resolution  1-8 cm3
Surface FOV
Compare to low-res fMRI: 64x64x30  217 voxels!
Whole brain coverage
from Franceshini, NeuroImage, 2004

15. fNIR Methodology: Spatial Localization with DOT
True tomographic methods ~ 10,000 S-D pairs
Flying spot illumation
CCD detection
Low temporal resolution ~10 – 100 sec / image
ill suited for functional assessment
“Hitting Density”,  – poor basis set
undetermined inversion problem
(r)
A = (r) (r) dr
 = Hb[Hb] + HbO[HbO])
from Strangeman Biol Psych 2002

16. fNIR Methodology: MBLL vs DOT
Many fNIR implementations report [Hb] changes from individual S-D pairs w/o attempt at DOT
DPF in MBLL calculated from uniform background absorption and scattering. Focal changes not properly modelled.
“MBLL and DOT results did not agree in terms of absolute magnitudes, relative magnitudes, or even the relative sign for changes in [HbO] and [Hb].”
Boas, NeuroImage, 2001

17. Spatial Maps of HRF Metrics: TTP Maps

18. fMRI: Mental Chronometry
ADC compartmentalization resolves events separated by 125ms.
TTP Map
1 second right fovea & auditory delay