Single-fiber reflectance spectroscopy: Is it more accurate than

advertisement
Single-fiber reflectance spectroscopy:
Is it more accurate than radiography
and computed tomography in
identifying degenerated canine
intervertebral discs?
Nigar Sultana1, Kelci McKeirnan2, Melanie A. Breshears3, Anqi Zhang4, Daqing Piao1 and
Kenneth E. Bartels2
1School
of Electrical and Computer Engineering, Oklahoma State University, Stillwater, OK 74078, USA
of Veterinary Clinical Science, Oklahoma State University, Stillwater, OK 74078, USA
3Depertment of Veterinary Pathobiology, Oklahoma State University, Stillwater, OK 74078, USA
4Depertment of Biomedical Engineering, Johns Hopkins University, Baltimore, MD 21205, USA
2Depertment
1
OVERVIEW
 Background
 Motivation and Objective
 System configuration
 Imaging protocols
 Results
 Summery
2
Background
 Intervertebral discs (IVD), located
between the vertebrae in the spine
and act as “shock absorbing
cushions” which allows slight
movement of vertebrae
 Intervertebral discs are comprised
of two layers
o Annulus fibrosus ---Tough outer
fibrous lamellar layer
o Nucleus pulposus --- Gelatinous
inner core
Source: http://www.spineuniverse.com/conditions/herniated-disc
3
Background

Intervertebral disc herniation is a common disease that
affects the vertebra

The chondrodystrophoic dogs (dachshund, Pekingese,
etc.) mainly seem to have disc degeneration along the
entire vertebral column while it is even common in
humans also

Loss of nucleus fluid and deposition of a calcified mineral
component within the nucleus makes the disc
degenerated

Degenerated disc is less resilient to normal wear and tear

Degeneration can cause severe pain, nerve compression
and paralysis
Extrusion of degenerated disc
material through the dorsal
annulus
Source: http://www.mayfieldclinic.com/PEHLDisc.htm#.Ujy34sakquM
4
Background
Treatment
 Surgical disc fenestration--- most
widely used treatment for
intervertebral disc disease
 It’s a highly invasive procedure
 Very complex
 Recurrence rate as high as 19.2 % as
reported by several large scale studies
of a mixed population of dogs
Source: http://cal.vet.upenn.edu/projects/saortho/chapter_62/62mast.htm
5
Background

Percutaneous Laser disc Ablation (PLDA)--minimal invasive procedure for intervention of
intervertebral disc disease

This laser surgical treatment photothermally ablates
the disc material and fibrotic scar tissue forms which
presumably prevents the disc from herniating in the
future

PLDA is performed by passing an optical fiber
through a needle placed percutaneously in the center
of a disc to introduce laser energy, most frequently
from a 2100 nm holmium:YAG (Ho:YAG) laser

PLDA shown to have minimum recurrence rate as low
as 3.4 %
6
Motivation
 In current practice, PLDA is performed in thoracolumbar IVD
with same laser fluence applied to each disc, regardless of its
mineral content
 Knowledge of individual disc mineral composition would allow
treatment of degenerated discs that otherwise may have been
missed
 Mineralized volume quantization of individual disc would permit
laser energy dosage adjustments for specific degenerated disc
7
Motivation (Imaging)

Current standard for diagnostic imaging for canine
IVD degeneration is X-ray radiography, which usually
shows as opacification of the disc space

Radiography is also used to guide the needle
positioning into the center of each treated disc for
PLDA

However, radiography is not adequately sensitive in
identifying the degenerated discs, nor is it reliable
for characterizing the degree of disc degeneration

When compared with histopathology, radiography
has a sensitivity of 30 % and specificity of 100 % in
diagnosing disc degeneration
Source: http://www.vetsurgerycentral.com/myelogram.htm
8
Motivation (imaging)

X-ray computed tomography (CT) and magnetic
resonance imaging (MRI) usually able to pinpoint
the precise location of degenerated disc

However, it is not also highly sensitive in
characterizing the degree of disc degeneration and
sometimes provide misleading results in
identifying the degenerated discs

When compared with histopathology, computed
tomography (CT) has a sensitivity of 75% and
specificity of 60 % in diagnosing disc degeneration

Moreover, due to the cost and inconvenience it is
challenging to standardize PLDA surgery under CT
or MRI image guidance
Source: http://www.mayfieldclinic.com/PE-HLDisc.htm#.Ujy7jsakquM
http://canadawestvets.com/disc-disease
9
Motivation (Optical)
 An imaging/sensing technology that provides a more accurate preoperative in-situ assessment of the disc mineralization,
 and potentially rapid post-operative feedback, could optimize the
outcome of the PLDA procedure
 Single-fiber reflectance spectroscopy---o Sense the changes of tissue compositions in the disc
o A diagnosis modality that could be integrated to PLDA procedure
http://www.dfwvetsurgeons.com/downloads/Newsflash_rdvm.pdf
10
Hypothesis (scattering)

The hypotheses
o The mineral deposit within degenerated IVD
increases scattering
o IVD that contains mineral works as an
avascular medium
o The more degenerated disc is, the more
scattering the medium should become
11
Objective
 The objective of this study is to revaluate whether the needlebased single-fiber reflectance spectroscopy could have higher
accuracy in identifying degenerated intervertebral discs
compared to X-ray radiography and computed tomography (CT)
 If needle-based single-fiber reflectance spectroscopy could
provide a positive correlation with histopathology results
 If SfRS could have a higher sensitivity in grading degree of disc
degeneration
12
System configuration
The experimental setup includes:
Computer
Light Source
Spectrometer

Halogen -Deuterium Source

Spectrometer

Bifurcated fiber bundle

320μm single fiber with 15°
Bifurcated fiber bundle
single fiber
150 angle
polished fiber
in needle
angle polished tip

Computer

20 gauge needle
15°
Schematic of Single fiber reflectance
spectroscopy setup
13
Analytical Representation
4
5
x 10
4.5
4
Raw Spectrum
Intensity (a.u.)
3.5
R Norm (λ) =
3
2.5
R tissue λ − R water (λ)
R air λ − R water (λ)
2
1.5
1
0.5
0
550
600
650
700
750
800
wavelength (nm)
850
900
950
1000
8
7.5
Normalized intensity
absorption coefficient
7
R Norm (𝑑𝑓 ) = 𝑅0 Norm (𝑑𝑓 ) 𝑒 −μ𝑎,𝑡𝑖𝑠𝑠𝑢𝑒 ⟨L⟩
Normalized Intensity
6.5
6
Normalized Spectrum
5.5
5
Incident light intensity
photon path length
4.5
4
3.5
3
550
600
650
700
750
800
wavelength (nm)
850
900
950
14
Imaging protocol
15
Imaging protocol
Acquisition time: 2 second/measurement
-Step1: Measurements
-Step2: PLDA procedure
-Step3: Measurements
extending
~1mm
Reflectance
spectroscopy 16
Results (Dog#1 Radiography and CT)
L2-3, slight
nuclear
mineralization
L4-5, no
nuclear
mineralization
No mineral present on radiographs
17
Results (Dog#2 Radiography and CT)
T10-11
mineralized
T11-12
mineralized
T12-13
mineralized
L1-2
mineralized
L2-3
mineralized
L3-4
mineralized
L4-5
mineralized
L5-6
mineralized
T8-9
T12-12
L2-3 L3-4 L4-5
are mineralized in radiograph
L7-S1
Results (Dog#3 Radiography and CT)
T10-11
mineralized
T11-12
mineralized
T12-13
mineralized
T13-L1
mineralized
L1-2
mineralized
L2-3
mineralized
L3-4
mineralized
L4-5
mineralized
T10-11 mineralized
19
Results (Radiography and CT)

Radiographic and CT imaging were performed on all thoracic and
lumbar intervertebral discs of the two cadaveric dogs

The needles for Single fiber reflectance spectroscopy sensing and
subsequently for PLDA were placed successfully in “ Dog 1” from T8– 9
to L5– 6 and in “ Dog 2 and Dog 3” from T9– 10 to L5– 6. Therefore only
the radiography and CT images of those discs being evaluated by Single
fiber reflectance spectroscopy measurement are displayed
20
Results (Dog1)
disc
No
Radiograph
CT
T8-9
No mineralization
No mineralization
T10-11
No mineralization
Slight focal mineral in
nucleus
T11-12
No mineralization
T12-13
No mineralization
No mineralization
None
No mineralization
Pathology
2 small foci of light mineralization within
nucleus (<10%) with equivocal mineralization
at periphery of presumptive char tract
No mineralization
Large central focus of dense mineralization
(60%) in nucleus
No mineralization
2 distinct foci of mineralization/bone within
nucleus that composed approx. 35% of visible
nucleus
T13-L1
No mineralization
L1-2
No mineralization
No mineralization
No mineralization
L2-3
No mineralization
very focal mineral
(nucleus)
Focus of dense mineralization near center of
nucleus (in chondroid tissue) (30%)
L3-4
No mineralization
No mineralization
No mineralization
L4-5
No mineralization
No mineralization
No mineralization
L5-6
No mineralization
No mineralization
equivocal mineralization at edge of nuclear
material surrounding
described focus of cavitation (10%)
21
Results (Dog2)
disc
No
Radiograph
CT
Pathology
T10-11
No mineralization
rim mineralization of
nucleus
no mineralization on H&E
T11-12
No mineralization
rim mineralization of
nucleus
2 small foci of mineralization in nucleus material
(~5% of nucleus)
many foci of distinct mineralization
(approximately 60% of nucleus material)
admixed with granular debris interpreted to
represent mineralized tissue
T12-13
small amount of
mineral
T13-L1
No mineralization
L1-2
No mineralization
small amount of
mineral
no distinct mineralization in nucleus
L2-3
small amount of
mineral
small amount of
mineral
2 small and equivocal foci of mineralization in
nucleus (approximately 10% of nucleus)
L3-4
small amount of
mineral
small amount of
mineral
multiple mineralized foci comprise
approximately 40% of the nucleus
L4-5
small amount of
mineral
small amount of
mineral
basophilic coagulum interpreted to be charred
mineralized material composes approximately
70% of nucleus
L5-6
No mineralization
small amount of
mineral
mineralized tissue composes approximately 50%
of nucleus tissue present in section
heavily mineralized
No mineralization
no distinct mineralization in visible nucleus
22
Results (Dog3)
disc
No
Radiograph
CT
Pathology
T10-11
mineralized
large amount of mineral
nearly all of the nucleus material on slide
is VK+ (mineralized)
T11-12
no mineralization
moderate mineral central
nucleus
4 fairly large foci of mineralization in
nucleus (approximately 40% of nucleus)
T12-13
no mineralization
mineralized ventral aspect
of nucleus
large foci of mineralization in nucleus
comprise about 50% of nucleus
T13-L1
no mineralization
mineralized and ruptured
nucleus
large foci of mineralization in nucleus
(40%) and annulus
L1-2
no mineralization
small amount of focal
mineral
central focus of mineralization comprises
approx. 20% of the nucleus
L2-3
no mineralization
small amount of focal
mineral
central focus of mineralization comprises
approx. 20% of the nucleus
L3-4
no mineralization
small amount of focal
mineral
central focus of mineralization comprises
approx. 18% of the nucleus
L4-5
no mineralization
very slight focal mineral
<5% of nucleus is equivocally mineralized
L5-6
no mineralization
No mineralization
central focus of mineralization comprises
approx. 10% of the nucleus
23
Results

In histopathology result, 20 out of 28 discs
were determined to have different level of
25
mineral deposition (5/10 form dog1, 4/9
R19 identified
by SfRS
from dog2 and 9/9 from dog3)

Out of 20 degenerated discs
o Radiography identified 6 degenerated
No of disc
20
15
15
10
discs
o CT was successfully identified 15
20
6
5
degenerated discs and
o SFRS identified 19 degenerated discs
1
Radiography
2
CT
3
Histopathology
correctly
24
Results
 Normalization is applied to all 28 discs
8
with respect to air and water from
entire optical range of 530-1000nm
spectra of normal disc (T12-13-Dog1,
L3-4-Dog1 and L1-2 Dog1)
6
Normalized Intensity
 Blue dashed line represents normalized
7
Degenerated
5
4
 Red dotted line represents normalized
spectra of degenerated disc (T12-13Dog3, T13-L1-Dog3, L1-2-Dog-1, T1112-Dog1)
identified by CT and
histopathology
Normal
3
2
550
600
650
700
800
750
Wavelength (nm)
850
900
950
1000
25
Mineralized
Normal
Pathology (%
of mineral
volume)
SfRS (avg
normalized
intensity of
530-1000nm)
Computed
Tomography
(CT)
Radiography
T8-9 T10-11 T11-12 T12-13 T13-L1 L1-2
L2-3
L3-4
L4-5
L5-6
26
Results

The comparison of SfRS normalized intensity with radiography and CT
based on histopathology result for discs T8-9, T10-11, T11-12, T12-13,
T13-L1, L1-2, L2-3, L3-4, L4-5 and L5-6 are shown here for three
cadaveric dogs

The red triangle represents degenerated disc and blue circle is the
representation of normal disc (not having any mineral)

For SfRS, the normalized intensity with respect to air and water is
presented on an average of optical range of 530-1000nm

SfRS result showed a good correlation with histopathology result of
different percentage of mineral content by different normalized intensity
level
27
Results
8
7
6
5
4
3
2
0
1
T8-9
2
T10-11
3
T11-12
4
5
T12-13 T13-L1
6
L1-2
7
L2-3
8
L3-4
9
L4-5
10
L5-6
 Receiver operating characteristic (ROC) curve for average normalized spectral
intensity of SfRS from 530-1000nm is generated by varying threshold level to
determine SfRS classification in identifying degenerated disc
 Threshold is placed where SfRS could identified 19 out of 20 degenerated discs
28
Results
 ROC curve is the graphical presentation of
sensitivity and specificity to determine
system classification where,
Specificity =
0.9
Ntrue positive
Ntrue positive + Nfalse negative
Ntrue negative
Ntrue negative + Nfalse positive
 The dotted line represents a random guess
of 50% accuracy
 The solid line shows the sensitivity and
specificity when SfRS identified 19 out of 20
degenerated disc by threshold placement
 Area under ROC curve of SfRS determined to
AUC=0.841
0.8
0.7
Sensitivity
Sensitivit𝑦 =
1
0.6
0.5
0.4
0.3
0.2
random guess
Single fiber Rs
0.1
0
0
0.1
0.2
0.3
0.4
0.5
0.6
1-Specificity
0.7
0.8
0.9
1
be 0.841
29
Results
 SfRS normalized intesity averaged
10
9
are plotted against the percentage
8
amout of mineralized volume
7
determined by histopathology
 Up to 50% of mineralized volume,
the correlation coefficient, R
Normalized SfRS Intesity
from 530 to 1000 nm for all 28 discs
2
R =0.7625
6
5
4
3
obtained as 0.8732 and coefficient
2
of determination, R2 determined to
1
be 0.7625
0
Homogenous
Avascular
0
10
20
30
40
50
60
70
Mineralized volume (%) in histopathology
80
90
100
30
Summery
 In identifying and grading canine intervertebral disc degeneration,
Single fiber Reflectance Spectroscopy is performed
 A needle-based single-fiber reflectance spectroscopy system that is
compatible with PLDA procedure is constructed
 SfRS measurements indicate increase of light scattering intensity
across the entire 530 –100 nm spectral range in discs with
mineralization
 In grading disc mineral composition, SfRS showed a higher
sensitivity and specificity than both X-ray radiography and
computed tomography (CT)
31
Summery
 In identifying degenerated disc by varying threshold level, SfRS
showed a good classification with having area under ROC curve of
0.841
 At high percentage of mineral volume (above 50%) confirmed by
histopatology imaages, due to higher number of scattering particle,
deposited mineral acts as a homogeneous medium and only
accounts for fresnel reflection while applying SfRS, thus SfRS
could not provide as good correlation as for less than 50% of
mineral volume
32
Acknowledgement
 Endowment to Bartels KE, Kerr Foundation, Oklahoma
city, OK
 Oklahoma Center for the Advancement of Science and
Technology (OCAST) HR 11-043
33
Download