Dhanashree Palande Daqing Piao School of Electrical and Computer Engineering Oklahoma State University Stillwater OK 74078 USA Outline O bjective Methods Results Conclusion and future ID: 384687
Download Presentation The PPT/PDF document "Trans-rectal near-infrared optical tomog..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Trans-rectal near-infrared optical tomography reconstruction of a regressing experimental tumor in a canine prostate by using the prostate shape profile synthesized from sparse 2-dimentional trans-rectal ultrasound images
Dhanashree
Palande
,
Daqing
Piao
School of Electrical and Computer Engineering, Oklahoma State University, Stillwater, OK 74078, USASlide2
OutlineObjective
Methods
Results
Conclusion and future work
2Slide3
ObjectiveNear infrared(NIR) optical imaging: well suited for non-invasive quantification of hemoglobin oxygen saturation(StO2)
p
rovides unique information regarding optical properties
Limitation of NIR: low spatial resolution due to high scattering in tissue
Solution: compensate optical imaging with spatial prior information extracted from high resolution trans-rectal ultrasound (TRUS) images to improve the reconstruction outcome of trans-rectal DOT
o
btain a 3D prostate profile from 2D TRUS images using segmentation which is used as a structural spatial prior in optical tomography reconstruction
3Slide4
OutlineObjective
Methods
Results
Conclusion and future work
4Slide5
NIR Optical Tomography(DOT) Non-invasive imaging technique: aims to reconstruct images of tissue function and physiologyBiological
tissue is highly scattering at NIR
wavelengths
(650-900 nm)Also known as diffuse optical tomography(DOT)NIR light is applied through optical fibers positioned to surface of the tissue
Emergent light is measured at other locations on the same tissue surface NIR optical tomography along with reconstruction algorithm, produces images of tissue physiology for detection and characterization of malignancy
5Slide6
HbT and StO2 measurementIn the range 700-900nm, absorption of water is much lower than that of oxygenated hemoglobin and deoxygenated
hemoglobin
Multi-wavelength data:
Rendered extracting oxygen saturation and hemoglobin
concentration.705 nm, 785 nm, 808 nm Absorption coefficients recovered at 3 specific bands are:
(
)
6Slide7
HbT and StO2 measurementThey are used to calculate
HbO
and
Hb as
Where,
was matrix of molar extinction coefficients
Total hemoglobin:
Hb
T
=
HbO+Hb
(in
milliMole
)
Oxygen saturation:
S
t
O
2
=
HbO
/
Hb
T
x 100 (in %)
7Slide8
NIR Reconstruction GeometryOuter rectangular mesh:
equivalent to tissue surrounding the prostate
Required to match NIR reconstruction geometry
8Slide9
The Forward ModelThe technique to determine what a given sensor would measure in given environment by using theoretical equations for sensor response
Diffusion approximation in frequency domain
Where
: absorption coefficient
: reduced scattering coefficient
: isotropic source term
: photon fluence rate at position r and modulation frequency
: diffusion coefficient
: speed of light in medium
9Slide10
The Inverse Model
Using
the results of actual observations to infer the values of the parameters characterizing the system under investigation.
Goal: recovery of optical properties
at each spatial element
Tikhonov
minimization:
: measured fluence at tissue surface
: calculated data using forward solver
Where, NM: number of measurements from imaging device
NN: number of spatial elements
: regularization parameter
: initial estimate of optical properties
10Slide11
The Inverse Model
The minimization with respect to μ in equation
: Jacobian matrix, J
Using linear approximation and solving it as iterative scheme,
Where,
: update of optical properties
: data-model misfit at current iteration I
I: identity matrix
Slight modification gives,
Where,
and
11Slide12
The Inverse Model
NIRFAST is used for inverse problem solving
12Slide13
TRUS Images of a Canine ProstateA canine prostate was used for study
Transmissible
Venereal
Tumor(TVT) cells was injected in right lobe of a prostateDog was monitored over the 63-days period, at weekly
intervalsTRUS images were taken at:Right edge plane
Right middle plane
Middle sagittal plane
Left middle plane
Left edge plane
13Slide14
TRUS Images of a Canine Prostate Axial view Sagittal view
rectum
rectum
Left lobe
Right lobe
Caudal side
Cranial side
14Slide15
TRUS Images of a Canine Prostate Axial view Sagittal view
rectum
rectum
Left lobe
Right lobe
Caudal side
Cranial side
15Slide16
TRUS Image SegmentationTRUS image segmentation is challenging due to
Complexity in contrast
Image artifacts
Morphological featuresVariation in prostate shape and size
Manual contour trackingInteractive program takes inputfrom user
Sagittal images segmented manually
Used as reference for 3D profile
16Slide17
Approximating Axial Plane PositionsWe have set of sparsely acquired axial imagesWe use 3 images at cranial side, middle and caudal side of the prostate
A program is written
t
o find approximate positions of these axialplanes
17Slide18
3D Profiling of a ProstateInterpolationSpline type of interpolation for smooth profile along the curve
Using the points on axial contours
New data points are interpolated depending on required mesh density
18Slide19
Mesh GenerationGeneration of a 3D mesh prostate profile using Delaunay triangulationInput: interpolated data points from 3D profile
Output: elements of all the tetrahedrons
This mesh is now used as a spatial prior for NIR image reconstruction
19Slide20
Prostate Mesh within Rectangular MeshMesh used for reconstruction
20Slide21
OutlineObjective
Methods
Results
Conclusion and future work
21Slide22
Manually Segmented ImagesFor axial images
For sagittal images
22Slide23
3D Prostate Profile3D profile of prostate
3D mesh profile of a prostate
23
Slide24
Rectangular Mesh With spatial prior Without spatial prior
24Slide25
Reconstruction: Right LobeBaselineWith spatial prior Without spatial prior
25
90
40
90
40
60 mm
60 mm
40 mm
40 mm
Ultrasound image
HbT
StO2Slide26
Reconstruction: Right LobeDay 49
With spatial prior Without spatial prior
26
90
40
90
40
60 mm
60 mm
40 mm
40 mm
Ultrasound image
HbT
StO2Slide27
Reconstruction: Right LobeDay 56
With spatial prior without spatial prior
27
90
40
90
40
60 mm
60 mm
40 mm
40 mm
Ultrasound image
HbT
StO2Slide28
Right Lobe
40
40Slide29
29
Right Lobe
(weeks) 7 8 9
7 8 9 (weeks)Slide30
Reconstruction: Left LobeDay 49
With spatial prior without spatial prior
30
90
40
90
40
60 mm
60 mm
40 mm
40 mm
Ultrasound image
HbT
StO2Slide31
Left LobeDay 63
With spatial prior without spatial prior
31
90
40
90
40
60 mm
60 mm
40 mm
40 mm
Ultrasound image
HbT
StO2Slide32
Left Lobe
32
90
10
90
10
40
40Slide33
OutlineObjective
Methods
Results
Conclusion and future work
33Slide34
Conclusion and future workContribution of this workThis work explores combination of structural and functional imaging for the study of prostate cancer
3D prostate profile was generated from sparse 2D axial TRUS images of a canine prostate
A prostate mesh developed was used a spatial prior to NIR optical tomography for image reconstruction
Reconstructed images with and without prior were compared qualitativelyThis approach helps to interpret results for good understanding of position of tumor lesion developed in prostate.
To our knowledge, this is the first attempt to use TRUS guided structural spatial prior for image reconstruction of a prostate using NIR optical tomography
34Slide35
Conclusion and future work-Future directionsExtending this study to other animals and eventually to human prostates
Applying spectral prior information along with spatial prior
To make this system work real-time, so as to be used during clinical exams
35Slide36
Thank youQuestions/suggestions
36