Introduction to Hyperspectral imaging technology and applications

1. BY
RAMY YEHIA
SUPERVISED BY
DR. JOE HAYWARD
Captioned from
“http://www.hightech-edge.com”

2. OUTLINES
1. HSI (What is it & How does it differ ?)
2. HSI History
3. HSI General Applications
4. HSI System Hardware
5. HSI Classification
6. Types of Dispersive elements
7. Acousto-Optic Tunable Filter (Theory)
8. Brief on HSI Image analysis
9. Medical Applications of LCTF & AOTF-based-HSI
10. Conclusion
11. References
3/3/2023
BME 707 Course 2014
2

3. WHAT IS HSI?
Hyper
Spectral
Imaging
3/3/2023
BME 707 Course 2014
3
Captioned from
“http://sametomorrow.com/blog/20
12/09/13/hyper-matrix-
installation/”
Captioned from
“http://chandra.harvard.edu/resour
ces/em_radiation.html”
Captioned from ”
http://web.khu.ac.kr/~tskim/PM_6_1
%20Introduction%20to%20Medical
%20Imaging.pdf”
Imaging

4. HSI VS SPECTRAL IMAGING
3/3/2023
BME 707 Course 2014
4
http://www.markelowitz.com/Hyperspectral.html

5. HSI HISTORY
3/3/2023
Introduction to Biophotonics 6IO3
5
Anomaly detection based on a parallel kernel RX algorithm for multicore platforms
Journal of Applied Remote Sensing
(1985)
Projects utilizing hyperspectral imaging usually have one of the
following objectives:
1-Target detection
2- Material mapping
3- Material identification
4- Mapping details of surface properties

6. HSI
DEFINITION PROBLEM
Original definition:
“The acquisition of images in hundreds of contiguous, registered, spectral
bands, such that for each pixel a radiance spectrum can be derived”1,2 by Goetz
1985.
Recent Publications considered HSI starting from approximately 20 bands especially in
the medical research field3-9.
Others still go with original definition for more than hundred bands10-14.
What are the main parameters of HSI ?
1) Number of bands (Debate between HSI & MSI)
2) FWHM of each band (HSI always narrower than MSI)
3) Continuity.
So we Propose a new updated HSI definition:
• “Hyperspectral Imaging is the process of capturing tens of consecutive, spectral,
co-registered images with narrow bandwidths, such that recorded bands establish
uninterrupted sample spectral line information”
3/3/2023
BME 707 Course 2014
6

7. HSI GENERAL APPLICATIONS
3/3/2023
BME 707 Course 2014
7
http://www.markelowitz.com/Hyperspectral.html

8. HSI SYSTEM HARDWARE
3/3/2023
BME 707 Course 2014
8

9. HSI-ILLUMINATION
 Illumination is specified
according to spectral region
of interest.
 Most of the remote sensing
HSI uses the sun light.
 Majority of the Medical
reflectance-HSI Use
Halogen sources.
3/3/2023
BME 707 Course 2014
9

10. HSI
Classification
Acquisition
mode
Spectral
range
Resolution
Measured
phenomena
Image
Detector
Dispersive element
3/3/2023
BME 707 Course 2014
10

11. 3/3/2023
BME 707 Course 2014
11
HSI
Classification
Acquisition
mode
Spectral
range
Resolution
Measured
phenomena
Image
Detector
Dispersive
element

12. HSI
Classification
Acquisition
mode
Spectral
range
Resolution
Measured
phenomena
Image
Detector
Dispersive
element
3/3/2023
BME 707 Course 2014
12

13. ACQUISITION MODE
Spatial scanning:
• Whiskbroom (point scanning)
• Pushbroom (line scanning)
• Both lack ability of image live display
• Their hypercube are calculated from spectra after spatial scanning
Spectral scanning:
• Stair looking (Area imaging)
• It has advantage of live image display for aiming and focusing
Fourier transform IR imaging (FTIR):
• Interferometer is used instead of dispersive element.
• Collected images are transformed to spectral domain using Fourier transform.
Snapshot hyperspectral imager:
• Captures the spatial and spectral information simultaneously.
• Small area investigation, very fast.
3/3/2023
BME 707 Course 2014
13

14. 3/3/2023
BME 707 Course 2014
14
Guolan Lu, Baowei Fei, “Medical Hyperspectral imaging: a review” Journal of Biomedical Optics
19(1), 010901 (January 2014).
Emission filter design to detect poultry skin tumors using fluorescence
hyperspectral imaging, Rev Colom Cienc
Pecua vol.23 no.1 Medellín Jan./Mar. 2010
L. W. Schumann and T. S. Lomheim, “Infrared hyperspectral imaging Fourier transform and
dispersive spectrometers: comparison of signalto- noise-based performance,” Proc. SPIE 4480,
1–14 (2002).
A. A. Fawzi et al., “Recovery of macular pigment spectrum in vivo using
hyperspectral image analysis,” J. Biomed. Opt. 16(10), 106008 (2011).

15. SNAP SHOT
IMAGER RESULTS
3/3/2023
BME 707 Course 2014
15
A. A. Fawzi et al., “Recovery of macular pigment spectrum in vivo
using hyperspectral image analysis,” J. Biomed. Opt. 16(10),
106008 (2011).

16. HSI
Classification
Acquisition
mode
Spectral
range
Resolution
Measured
phenomena
Image
Detector
Dispersive
element
3/3/2023
BME 707 Course 2014
16

17. 3/3/2023
BME 707 Course 2014
17
HSI
Classification
Acquisition
mode
Spectral
range
Resolution
Measured
phenomena
Image
Detector
Dispersive
element

18. 3/3/2023
BME 707 Course 2014
18
HSI
Classification
Acquisition
mode
Spectral
range
Resolution
Measured
phenomena
Image
Detector
Dispersive
element

19. HSI RANGE
Spectral range:
• It depends on the object optical properties to be investigated.
• Ultraviolet (UV): 200-400 nm
(Cervical Neoplasia reflectance and fluorescence)
• Visible (VIS): 400-750 nm
(1-2 mm penetration, it is concerned of subpapillary info.)
• Near-Infrared (NIR): 750-2500 nm
(surgical guidance due to its deep penetration)
• Mid-Infrared: 2500-25000 nm
(rich info. about Genomics, Proteomics,& Metabolomics of
a cell)
.
3/3/2023
BME 707 Course 2014
19

20. HSI RESOLUTION
Resolution:
• Spectral:
• Ability of system to resolve spectral bands (nm)
• FWHM of the resolved band
• Spatial:
• Dimension of smallest element required to be observed
3/3/2023
BME 707 Course 2014
20
www.cyberphysics.co.uk
advanced-microscopy.utah.edu
pinholeworks.com

21. 3/3/2023
BME 707 Course 2014
21
HSI
Classification
Acquisition
mode
Spectral
range
Resolution
Measured
phenomena
Image
Detector
Dispersive
element

22. 3/3/2023
BME 707 Course 2014
22
HSI
Classification
Acquisition
mode
Spectral
range
Resolution
Measured
phenomena
Detector
Dispersive
element

23. 3/3/2023
BME 707 Course 2014
23
HSI
Classification
Acquisition
mode
Spectral
range
Resolution
Measured
phenomena
Image
Detector
Dispersive
element

24. PHENOMENA & DETECTOR
What do you want to measure?
• Reflectance
• Fluorescence
• Transmission
Detector type:
• Charge coupled device (CCD) (expensive, complex, slow, High
sensitivity, and precision)
• Intensive Charge coupled device (ICCD)
• Complementary Metal Oxide semiconductor
(CMOS) (Cheap, Less complex, faster, less sensitive)
3/3/2023
BME 707 Course 2014
24
http://www.digitalbolex.com/global-shutter
http://electronics.howstuffworks.com/cameras-photography/digital/question362.htm
http://www.andor.com/learning-academy/intensified-ccd-cameras-the-technology-behind-iccds

25. 3/3/2023
BME 707 Course 2014
25
HSI
Classification
Acquisition
mode
Spectral
range
Resolution
Measured
phenomena
Image
Detector
Dispersive
element

26. 3/3/2023
BME 707 Course 2014
26
HSI
Classification
Acquisition
mode
Spectral
range
Resolution
Measured
phenomena
Image
Detector
Dispersive
element

27. HSI
Classification
Acquisition
mode
1-Spectral scanning
2- Spatial scanning
Spectral
range
1-UV-VIS
2-VIS-NIR
3-SWIR
Resolution
1- spectral
2-Spatial
Measured
phenomena
1-Reflectance
2-Transmission
3-Fluorescence
Image
Detector
1-CCD
2-ICCD
3-CMOS
Dispersive
element
1-Monochromator
2- Single shot imager
3-Prism Grating Prism
4-Optical Bandpass Filter
3/3/2023
BME 707 Course 2014
27

28. DISPERSIVE ELEMENTS
Prism
• Disperses light due to prism
material change of refractive
index
• Prism based HSI are complex
due to non-linear scanning
dispersion.15
Diffraction
Grating
• Surface-ruled grooves comparable
to light wavelength diffracts light.
• Less complexity
• Low throughput relative to prism
based HSI.
3/3/2023
BME 707 Course 2014
28
http://www.askamathematician.com/2011/05/cheap-experiments-and-
demonstrations

29. DISPERSIVE ELEMENTS
3/3/2023
BME 707 Course 2014
29
Computer
generated
holograms
Array of
square
pixels
functioning
as grating
Capable of
capturing
spectral and
spatial info.
In single
frame
http://www.jiscdigitalmedia.ac.uk/guide/digital-
cameras
2-

30. DISPERSIVE ELEMENTS
Grating volume
sandwiched
between two prisms
with a low pass filter
and long pass filter
to eliminate
unwanted
wavelengths
3/3/2023
BME 707 Course 2014
30

31. DISPERSIVE ELEMENTS
4- Optical
Bandpass
Filter
Tunable
AOTF LCTF
Fixed
Filter
wheel
3/3/2023
BME 707 Course 2014
31

32. DISPERSIVE ELEMENTS
Fixed Filter
(Interference filters)
-Mechanical vibration
-slow wavelength
switching
-Limited spectral
bands
-Low resolution
Wheel-mounted,
mostly 10 band
filters, rotated in front
of Light source or
detector, used with
MSI mainly
3/3/2023
BME 707 Course 2014
32

33. WHAT IS OPTICAL TUNABLE FILTER?
Optical tunable filter (TF) is a mean of controlling the spectral transmission
electronically by several ways such as applying voltage or acoustic signals, etc.
3/3/2023
BME 707 Course 2014
33
What are Ideal Tunable Filter properties?
No. Ideal Tunable Filter Characteristic
1 Infinite spectrum range with random access to wavelength
2 In sensitive to environment variation (Heat and humidity)
3 Minimum tunable time between selected wavelengths
4 Low power consumption
5 Small physical size
6 Large aperture size
7 In sensitive to light polarization
8 Infinitesimal and selectable bandpass wavelength
9 Constant bandpass to wide spectrum of wavelengths
10 In sensitive to the light incidence angle (wide field of view)

34. DISPERSIVE ELEMENTS
Tunable
Filter (TF)
Liquid Crystal
Tunable Filter
(LCTF)
Acousto-Optic
Tunable Filter
(AOTF )
3/3/2023
BME 707 Course 2014
34

35. DISPERSIVE ELEMENTS
LCTF
Birefingent crystal filter, applies
retardation principle, in phase,
between ordinary and
extraordinary illumination beam
passing through stack of its
units to develop constructive
and destructive interference, to
pass the selected wavelength
band of light.
Time separation between
wavelength tuning is depending
on the relaxation time of the
crystal. This time ranges from
50 to 150 ms. .
LCTF is highly sensitive
element to polarization.
3/3/2023
BME 707 Course 2014
35
https://www.cis.rit.edu/research/thesis/bs/1999/canning/thesis.ht
ml
http://en.wikipedia.org/wiki/Liquid_crystal_tunable_filter

36. DISPERSIVE ELEMENTS
3/3/2023
BME 707 Course 2014
36
• The spectral passband of LCTF is determined polarizers, optical
coatings, and crystal layers structures properties.
• Retardation is mainly dependent on crystal length and refractive
index variation produced between ordinary and extraordinary beam
at the wavelength of incident illumination.
http://www.perkinelmer.ca/en-ca/Catalog/Product/ID/VARISPC

37. ACOUSTO-OPTIC TUNABLE FILTER
(THEORY)
3/3/2023
BME 707 Course 2014
37
Near Infrared Spectroscopy: fundamentals, practical aspects and analytical applications, J. Braz.
Chem. Soc. vol.14 no.2 São Paulo Mar./Apr. 2003

38. AOTF (THEORY)
There are two types of AOTF:
1- Collinear AOTF
3/3/2023
BME 707 Course 2014
38
Olympus Microscopy Resource Center Acousto-Optical Tunable Filters (AOTF,s)
"www.olympusmicro.com/primer/techniques/confocal/aotfintro.html".

39. AOTF (THEORY)
Collinear AOTF was historically first
reported.
The start came from Bragg equations
V is the acoustic velocity, λ is the optical
wavelength in vacuum, and ni and nd
are the refractive indices of incident and
diffracted light.
The significant region of this plot for
AOTF operation is that corresponding to
the minimum frequency for which phase
matching can take place.
3/3/2023
BME 707 Course 2014
39
“Design and Fabrication of Acousto-Optic devices”, Akis P. Goutzoulis, Dennis R.Pape, 1994 by MARCEL DEKKER, INC.
http://nte-serveur.univ-lyon1.fr/spectroscopie/revuepresse_fichiers/Tunablefilters.pdf

40. AOTF (THEORY)
AOTF tuning
∆𝑛 = 𝑛𝑖−𝑛𝑑
Filter transmittance
P1 is the coupled signal versus incident P0, 𝜌 is the AO material density, PA is the acoustic
power density, and M2(ij) is AO figure of Merit corresponding to photoelastic coefficient Pij
3/3/2023
BME 707 Course 2014
40
“Design and Fabrication of Acousto-Optic devices”, Akis P. Goutzoulis, Dennis R.Pape, 1994 by MARCEL DEKKER, INC.

41. AOTF (THEORY)
Filter response
The frequency response of AOTF:
∆𝑦 : is the difference between filter central frequency and optical
frequency.
b: dispersive constant, intrinsic property of AO crystal.
b = 2𝜋(𝑛0 − 𝑛𝑒)
3/3/2023
BME 707 Course 2014
41
“Design and Fabrication of Acousto-Optic devices”, Akis P. Goutzoulis, Dennis R.Pape, 1994 by MARCEL DEKKER, INC.
http://nte-serveur.univ-lyon1.fr/spectroscopie/revuepresse_fichiers/Tunablefilters.pdf

42. AOTF (THEORY)
2- Non-Collinear AOTF
3/3/2023
BME 707 Course 2014
42
Olympus Microscopy Resource Center Acousto-Optical Tunable Filters (AOTF,s)
"www.olympusmicro.com/primer/techniques/confocal/aotfintro.html".

43. AOTF (THEORY)
The earliest report of non-collinear AOTF was in 1976.
The main idea in this configuration is the compensation of momentum mismatch
between wave vectors as a result of angular change between optical and acoustic
waves with the angular change of birefringence in interaction plane.
Extraordinary polarized light component wave vector is ellipse, while Ordinary one
is on circle.
3/3/2023
BME 707 Course 2014
43
“Design and Fabrication of Acousto-Optic devices”, Akis P.
Goutzoulis, Dennis R.Pape, 1994 by MARCEL DEKKER, INC.
Accurate Optical Design of an Acousto-optic Tunable Filter Imaging Spectrometer, Pengwei
Zhou, Huijie Zhao, Ying Zhang, ChongChong Li, © 2012 IEEE

44. AOTF (THEORY)
The shown figure is used to calculate different Non-Collinear
AOTF various parameters
1. Diffracted light angle
2. The tuning relationship
3. The output separation angle
3/3/2023
BME 707 Course 2014
44
“Design and Fabrication of Acousto-Optic devices”, Akis P. Goutzoulis, Dennis R.Pape, 1994 by MARCEL DEKKER, INC

45. AOTF (THEORY)
Filter Transmission
The filter resolution can be
calculated by letting
So we have, Filter passband FWHM
3/3/2023
BME 707 Course 2014
45
“Design and Fabrication of Acousto-Optic devices”, Akis P. Goutzoulis, Dennis R.Pape, 1994 by MARCEL DEKKER, INC.

46. AOTF (THEORY)
Non-Collinear AOTF angular aperture
The AOTF angular aperture is defined as that range of incident light beam
directions over which the deviation from perfect phase matching is sufficiently small
that for a fixed RF frequency the passband of the filter remains relatively
unaffected; i.e., the resolution is not degraded
Etendue Comparison
3/3/2023
BME 707 Course 2014
46
Optical
element
Etendue
relation
Parameters
definition
Typical numerical values
AOTF E = 2𝜋𝑛𝑒𝑛0𝑇 T: system optical
Transmission
EAOTF
Fabry-Perot
Filter (FPF)
E = 2𝜋𝑇 EAOTF > EFPF with a factor of nen0
This factor typically is in the range of
4x-10x
Grating
spectrometer
E = TA
ℎ
𝐹
2𝑡𝑎𝑛𝜃
h: slit height, F: grating-slit
distance , A: effective area,
θ: Blaze angle angle, typical
value 30o
Etendue is dependent on the blaze angle
and slit dimension ratio with respect to
AOTF
Prism grating 𝐸𝑝𝑟𝑖𝑠𝑚
𝐸𝑔𝑟𝑎𝑡𝑖𝑛𝑔
=
1
2𝑡𝑎𝑛𝜃
λ𝑑𝑛
𝑑λ
It has been estimated this factor to be about
16x for an NaCl prism at 10 µm
“Design and Fabrication of Acousto-Optic devices”, Akis P. Goutzoulis, Dennis R.Pape, 1994 by MARCEL DEKKER, INC.

47. AOTF (THEORY)
Comparison between Collinear & Non-Collinear AOTF
3/3/2023
BME 707 Course 2014
47
Parameter Non-collinear AOTF (TeO2)
Collinear AOTF
(Quartz)
Spectrum range 380-5500 nm 200-1000 nm
RF tuning frequency 20-200 MHz 50-220 MHz
Deflection angle 3-9 degree Zero degree
Diffraction efficiency
(Polarized Input)
10-90% 20-90%
RF input power 0.5-3 Watts 5-30 watts
Tuning speed 4-20 micro-sec. 14-35 micro-sec.
Optical aperture 0.1-1.5 cm2 0.1-5.0 cm2
Olympus Microscopy Resource Center Acousto-Optical Tunable Filters (AOTF,s)
"www.olympusmicro.com/primer/techniques/confocal/aotfintro.html".

48. BRIEF ON HSI IMAGE ANALYSIS
Image preprocessing:
-Normalization: reduce noise
-Reflectance
𝑅 𝜆 =
𝐼𝑟𝑎𝑤 𝜆 −𝐼𝑑𝑎𝑟𝑘(𝜆)
𝐼𝑤ℎ𝑖𝑡𝑒 𝜆 −𝐼𝑑𝑎𝑟𝑘(𝜆)
-Image Co-registration: fix movement artifacts
3/3/2023
BME 707 Course 2014
48
Captioned from “Basic PET Data
Analysis Techniques” Karmen K.
Yoder

49. HSI IMAGE ANALYSIS
Feature extraction& Selection:
The goal is to obtain the most relevant information from the
original HSI datasets, where larger number of spectral bands may
potentially make discrimination between more detailed classes
possible.
3/3/2023
BME 707 Course 2014
49
Dorra Nouri,
Yves Lucasa,
Sylvie
Treuillet,"
Calibration and
test of a
hyperspectral
imaging
prototype
for intra-
operative
surgical
assistance "
"Proc. SPIE
8676, Medical
Imaging 2013:
Digital
Pathology,
Lake Buena
Vista (Orlando
Area), Florida :
United States
(2013)".

50. HSI IMAGE ANALYSIS
Classification
• Hyperspectral image classification methods applied in the medical area mainly include
pixel and subpixel classification based on the type of pixel information used.
• Common Classification techniques are:
a) Support vector machines
b) Artificial neural networks
c) Spectral information divergence
d) Spectral angle mapper
e) Spectral unmixing
3/3/2023
BME 707 Course 2014
50
Zhi Liu, Hongjun
Wang, Qingli Li ”
Tongue Tumor
Detection in
Medical
Hyperspectral
Images” Sensors
2012, 12, 162-
174;
doi:10.3390/s1201
00162.

51. 3/3/2023
BME 707 Course 2014
51
. Guolan Lu, Baowei Fei, “Medical Hyperspectral imaging: a review” Journal of Biomedical Optics 19(1), 010901 (January 2014).
Medical Applications of Hyperspectral Imaging
#
Spectral
range
(nm)
Spectral
resolution
(μm∕pixel)
Detector Dispersive
device
Acquisition
mode
Measurement
mode
Application
1 400 to 1100 — Si CCD Filter wheel Staring Reflectance Burn wounds
2 200 to 700 5 CCD Filter wheel Staring
Fluorescence and
reflectance
Cervical
neoplasia
3 530 to 680 12 CCD Prism Pushbroom Transmission
Cutaneous
wound
4 5000 to 10,526 11 HgCdTe — FTIR Reflectance
Cervical
pathology
5 500 to 600 — CCD LCTF Staring Reflectance Diabetic foot
6 440 to 640 1 to 2
CCD;
ICCD AOTF Staring
Fluorescence and
reflectance
Skin cancer
7
400 to 1000;
900 to 1700;
950 to 2500
5
Si CCD;
InGaAs;
HgCdTe
Grating Pushbroom Reflectance Skin bruises
8 450 to 700 ∼ 1 FPA CGH Snapshot Reflectance Ophthalmology
9 1000 to 2500 6.29 HgCdTe PGP Pushbroom Reflectance Gastric cancer

52. LCTF & AOTF HSI APPLICATIONS
This section displays two applications for hyperspectral imaging, one
for each mentioned tunable filters:
Hyperspectral imaging for intra-operative surgical assistance
(LCTF-based spectrometer)
Hyperspectral Imaging for Tongue Tumor Detection
(AOTF-based spectrometer)
3/3/2023
BME 707 Course 2014
52
Dorra Nouri, Yves Lucasa, Sylvie Treuillet," Calibration and test of a hyperspectral imaging prototype for intra-operative surgical assistance “
"Proc. SPIE 8676, Medical Imaging 2013: Digital Pathology, Lake Buena Vista (Orlando Area), Florida : United States (2013)".
Zhi Liu, Hongjun Wang, Qingli Li ” Tongue Tumor Detection in Medical Hyperspectral Images” Sensors 2012, 12, 162-174; doi:10.3390/s120100162.

53. HYPERSPECTRAL IMAGING FOR INTRA-OPERATIVE
SURGICAL ASSISTANCE
This application aims to assist the surgeon during intervention in
operating room to distinguish between critical tissues that may not be
clear by naked eyes, such as facial nerves. Insensitive deal may lead to
partial or complete paralysis
3/3/2023
BME 707 Course 2014
53
http://www.tabletsmanual.com/wiki/read/bells_palsy

54. HSI FOR SURGICAL ASSISTANCE
The system comprises of
1. Halogen Illumination source.
2. Spectral imager operating in range of
400-1100 nm, using two LCTF devices,
the first for range 400-720nm, the
second for 650-1100nm.
3. High sensitivity CCD camera.
4. 35 mm focal length lens.
5. Computer.
3/3/2023
BME 707 Course 2014
54

55. HSI FOR SURGICAL ASSISTANCE
Due to high sensitivity of operating room, system should be accurately
calibrated.
3/3/2023
BME 707 Course 2014
55

56. HSI FOR SURGICAL ASSISTANCE
1- spectral response of a standard reflection object (Spectralon).
3/3/2023
BME 707 Course 2014
56

57. HSI FOR SURGICAL ASSISTANCE
2-Illumination Calibration
3/3/2023
BME 707 Course 2014
57
(a) 0.9 m (b) 1.0 m
(c) 1.1 m

58. HSI FOR SURGICAL ASSISTANCE
3-Depth of field (DOF) Calibration
3/3/2023
BME 707 Course 2014
58

59. HSI FOR SURGICAL ASSISTANCE
4- Lens Achromaticity
3/3/2023
BME 707 Course 2014
59

60. HSI FOR SURGICAL ASSISTANCE
The clinical results in the operating room
3/3/2023
BME 707 Course 2014
60

61. HSI FOR SURGICAL ASSISTANCE
Achieved goals:
1. The new modality has achieved a good intervention with the
medical operation room with limited obstruction.
2. The technique provided surgeons ability of distinguishing critical
tissues like facial nerves in a complex media.
3/3/2023
BME 707 Course 2014
61

62. TONGUE TUMOR DETECTION
Tongue cancer should be early diagnosed to prevent its spreading
to neighbor tissues. It acts as a life danger if it metastasizes the
lymph nodes in the neck and go through to rest of the body.
3/3/2023
BME 707 Course 2014
62
https://www.healthxchange.com.sg/healthyliving/SpecialFocus/Page
s/tongue-cancer-more-young-women-falling-victim.aspx

63. TONGUE TUMOR DETECTION
Histopathology is the gold standard of Cancer diagnosis.
However it has several disadvantages, for instance high cost,
invasive technique, time consumption in addition to subjective
diagnosis.
3/3/2023
BME 707 Course 2014
63
http://www.histopath-lab.com/online-faqs.html

64. TONGUE TUMOR DETECTION
AOTF Spectral imager
3/3/2023
BME 707 Course 2014
64

65. TONGUE TUMOR DETECTION
Tumor detection technique
3/3/2023
BME 707 Course 2014
65

66. TONGUE TUMOR DETECTION
Database samples of tongue tumors
3/3/2023
BME 707 Course 2014
66

67. TONGUE TUMOR DETECTION
Spectral signature of tumors versus normal tissues
3/3/2023
BME 707 Course 2014
67

68. TONGUE TUMOR DETECTION
3/3/2023
BME 707 Course 2014
68
Comparison between the physician opinion(Left) and the AOTF
system classifier algorithm (right)

69. TONGUE TUMOR DETECTION
Results
1-This system has introduced a method for tongue tumor
detection based on AOTF-spectrometer.
2-The spectral characteristics of the tissues using sparse
representation has achieved 96.5% recognition ratio
between tumors and normal tissues.
3/3/2023
BME 707 Course 2014
69

70. HSI SUMMARY
Achievements
1- HSI acquires 3-D images (2-D spatial and 3rd is spectral ) for
samples.
2- Real time and Non-invasive modality of imaging.
3- It can be adapted to other modalities such as Microscopy,
Colposcopy, Endoscopy, … etc, to achieve better results.
Challenges:
1-Acquisition of high resolution HSI datasets in video rates.
2- Fast processing techniques deals with huge amount of data
acquired by HSI.
3-Establishment of database for the important biomarkers and
tissues such as skin, brain tissue, ocular tissue,….etc.
3/3/2023
BME 707 Course 2014
70

71. REFERENCES
1.“Mineralogical mapping in the Cuprite mining district, Nevada,” in Proceedings of the Airborne Imaging Spectrometer Data Analysis Workshop, A.
F. H. Goetz and V. Srivastava (JPL, 1985), pp. 22–29.
2.“Imaging spectrometry for earth remote sensing,” A. F. H. Goetz, G. Vane, J. E. Solomon, and B. N. Rock, Science 228, 1147–1153 (1985).
3.“Hyperspectral Fluorescence Imaging for Mouse Skin Tumor Detection”, Seong G. Kong, Matthew E. Martin, and Tuan Vo-Dinh, ETRI Journal,
Volume 28, Number 6, December 2006.
4.“Hyperspectral fluorescence lifetime imaging for optical biopsy”, Zhaojun Nie,Ran An, Joseph E. Hayward,cThomas J. Farrell, and Qiyin Fang,
Journal of Biomedical Optics 18(9), 096001 (September 2013)
5.. Zhi Liu, Hongjun Wang, Qingli Li ” Tongue Tumor Detection in Medical Hyperspectral Images” Sensors 2012, 12, 162-174;
doi:10.3390/s120100162.
6.“Hyperspectral imaging of hemoglobin saturation in tumor microvasculature and tumor hypoxia development”, Brian S. Sorg Benjamin J. Moeller,
Owen Donovan, Yiting Cao, Mark W. Dewhirst, Journal of Biomedical Optics 10_4_, 044004 _July/August 2005
7.“New method for detection of gastric cancer by hyperspectral imaging: a pilot study”, Shu Kiyotoki, Jun Nishikawa, Shin Satori, Isao Sakaida,
Journal of Biomedical Optics 18(2), 026010 (February 2013).
8.“In vivo skin chromophore mapping using a multimode imaging dermoscope (SkinSpect™)”, Nicholas MacKinnon, Daniel L. Farkas, Proc. of SPIE
Vol. 8587, 85870U · © 2013 SPIE.
9.“Hyperspectral imaging for early detection of oxygenation and perfusion changes in irradiated skin”, Michael S. Chin, Brian B. Freniere, Yuan-
Chyuan Lo, Jonathan H. Saleeby, Stephen P. Baker, Heather M. Strom, Ronald A. Ignotz, Janice F. Lalikos, Thomas J. Fitzgerald, Journal of
Biomedical Optics 17(2), 026010 (2012)
10.Hyperspectral imaging as a diagnostic tool for chronic skin ulcers”, Martin Denstedt, Lyngsnes Randeberg,, Proc. SPIE 8565, Photonic
Therapeutics and Diagnostics IX, 85650N (March 8, 2013); doi:10.1117/12.2001087.
11.“Hyperspectral Imaging for Measurement of Oxygen Saturation in the Optic Nerve Head”, Bahram Khoobehi, James M. Beach, and Hiroyuki
Kawano, Investigative Ophthalmology & Visual Science, May 2004, Vol. 45, No. 5 Copyright © Association for Research in Vision and
Ophthalmology
12.“Hyperspectral imaging of atherosclerotic plaques in vitro”, Eivind L. P. Larsen, Lise L. Randeberg, Elisabeth Olstad, Olav A. Haugen, Astrid
Aksnes, Lars O. Svaasand, Journal of Biomedical Optics 16(2), 026011 (February 2011).
13.," Calibration and test of a hyperspectral imaging prototype for intra-operative surgical assistance " Dorra Nouri, Yves Lucasa, Sylvie Treuillet
"Proc. SPIE 8676, Medical Imaging 2013: Digital Pathology, Lake Buena Vista (Orlando Area), Florida : United States (2013)".
14.“AOTF based Hyperspectral Tongue Imaging System and Its Applications in Computer-aided Tongue Disease Diagnosis”, Zhen Sun, Hongjun
Wang, Qingli Li ,2010 3rd International Conference on Biomedical Engineering and Informatics (BMEI 2010).
15.Guolan Lu, Baowei Fei, “Medical Hyperspectral imaging: a review” Journal of Biomedical Optics 19(1), 010901 (January 2014).
16.Dorra Nouri, Yves Lucasa, Sylvie Treuillet," Calibration and test of a hyperspectral imaging prototype for intra-operative surgical assistance
"Proc. SPIE 8676, Medical Imaging 2013: Digital Pathology, Lake Buena Vista (Orlando Area), Florida : United States (2013)".
3/3/2023
BME 707 Course 2014
71

72. 3/3/2023
Introduction to Biophotonics 6IO3
72