Micro X-ray Computed Tomography for Mouse Phenotype Screening

ugliestmysticΤεχνίτη Νοημοσύνη και Ρομποτική

14 Νοε 2013 (πριν από 3 χρόνια και 8 μήνες)

67 εμφανίσεις

Micro X
-
ray Computed
Tomography for Mouse
Phenotype Screening

M. J. Paulus

Oak Ridge National Laboratory

P.O. Box 2008

Oak Ridge, TN 37831
-
6006

paulusmj@ornl.gov

ORNL MicroCAT Team


M.J. Paulus


Hardware Development


H. Sari
-
Sarraf

Software Development


S.S. Gleason


Software Development


D. K. Johnson

Mouse Genetics


J.S. Hicks


Motion Control


J. Turner


Mechanical Design


J. K. Behel


Graduate Student


L. Thompson


Graduate Student


D. Austin


Undergraduate Student


A. Symko


Undergraduate Student


Outline


ORNL Functional Genomics Program


MicroCAT System Description


Recent Studies


Open Issues


Summary

Functional Genomics at ORNL


Objectives:


Support R&D in genome
-
scale function
determination


Sequence
-

and mutagenesis
-
based FG


Focus on comparative genomics


Demonstrate proof
-
of
-
principle


The ORNL Functional Genomics Initiative
integrates efforts in mouse genetics,
genome informatics/ computational
biology, and analytical technologies.


Assumptions:


Sequences are becoming available at
astonishing pace
-

megabases/day


Discovery of several hundred genes
possible every day


Function determination on a large scale is
the current challenge


Understanding complex biological
systems is the next challenge

ORNL is in the Mouse Business


>72,000 Mice in colony.


More than 1000 different
strains.


Colony stems from early Atomic
Energy Commission Research
on Radiation Effects.


Current research focuses on
mouse mutagenesis.


Phenotype Driven Approach


Germ
-
cell mutations are induced using X
-
rays and N
-
ethyl
-
N
-
nitrosourea (ENU)
.


X
-
rays most often cause large lesions, such as
deletions and translocations that can involve multiple
genes.


ENU treatment typically produces intragenic
mutations, such as point mutations.


X
-
ray and ENU mutations occur with relative low
frequency.


Phenotypic screens are required to identify mutants.


A number of models have been developed including
epilepsy, depression, Angleman syndrome and
obesity.

Problem: The Screening
Bottleneck


Approximately 1 out of every
500 mutagenized mice exhibit a
new phenotype.


Screening is time consuming and
costly.


Subtle phenotypes such as
abnormal organ development or
skeletal abnormalities are often
difficult to detect with traditional
screens.

Solution: Dedicated X
-
ray CT
Mouse Screening System


X
-
rays provide highest resolution of all medical
imaging probes.


Availability of bright x
-
ray sources provide
potential for rapid image acquisition.


X
-
ray CT systems are among the least expensive
imaging tools.


Anatomic phenotypes are of significant interest
for the ORNL functional genomics program.

MicroCAT Design Goals


High throughput: ~1 minute per mouse
image acquisition.


High resolution: ~50 mm FWHM in a 3
-
D
tomographic image, for
in vivo

embryo and
small organ studies.


Sensitivity to both skeletal and soft tissue.


Computer aided image screening.


Direct extension to multi
-
modality imaging.

Key System Components

Mechanical


Assembly


and Motion


Control

Detector

Read
-
out


Electronics

Image

Processing

Optical Breadboard and Radiation
Enclosure

Detector, X
-
ray Source and Bed

Motion Control System


Independent rotating stage, patient bed and
detector assembly motion.


Stepper motors drive precision motion stages.


Bed and detector stages: <10 micron positional
accuracy.


Rotating stage: <0.001 degree positional
accuracy.


National Instruments LabWindows motion
control drivers.

X
-
ray Source

X
-
ray Source Characteristics

Emission Spectra

Focal Spot

FWHM
x
=168
m
m FWHM
y

= 119
m
m

Detector


MedOptics Digital Mammography Camera.


Kodak KAF
-
1000 CCD (1024 x 1024 pixels).


2:1 Fiber Optic Reducing Taper.


Kodak minR Phosphor Screen.


~50 x 50
m
m
2

pixels.


~35 x 35
m
m
2

sampling in FOV.


~1 second per frame readout.

Spatial Resolution is Limited by
the Phosphor Screen

0
0.2
0.4
0.6
0.8
1
1.2
0
5
10
15
20
25
Pixel Size
Focal Spot Size
MinR Phosphor Screen
Measured Data
Spatial Frequency (mm
-
1
)

MTF

Image Reconstruction


Cone beam treated as pseudo
-
parallel fan
beams (10 degree cone).


Fanbeam
-
to
-
parallel resort.


Standard Ramp Filter.


Standard Filtered Back Projection
Reconstruction.


ML Reconstruction also developed for low
-
count
studies.


User
-
Friendly Reconstruction Shell

Area Detector Produces
Volumetric Images

Soft Tissue Resolution Improved Through IP
Injection of Contrast Agent

No Contrast Agent

Iodine Contrast Agent

Three Week Old (~6.5 cm long) Wild Type Mice

Recent Case Studies


Polycystic Kidney Disorder


Agouti Obesity


Osteo*** Model


Prostrate Cancer


MGP Knockout


Amyloidosis


Polycystic Kidney Disorder


ORNL “ORPK” polycystic
kidney disorder insertional
mutation (R. Woychik).


Specimen received IP contrast
media injection 30 minutes
prior to scan.


Cysts lead to uneven
distribution of contrast media
uptake in the kidneys.

Lean vs. Obese Mouse

Mutation of the agouti gene
induces obesity in the mouse
model.

(R. Woychik, ORNL)

Lean vs. Obese Mouse


In vivo

measurement of fat pad volume.


In vivo

determination of whole body adipose
tissue density.


Work in progress to identify a subset of slices
suitable for assessing obesity.


Preliminary Lung Tumor Study


BALB/c mouse injected with
lung carcinoma.


Specimen received IP contrast
media injection 30 minutes
prior to scan.


Targeted therapy using MoAb
linked to alpha particle
emitter.


Work under way to correlate
CT data with histology.



(S. Kennel, ORNL)

Prostate Cancer Model

(U.C. Berkeley and Baylor University)


Baylor University TRAMP
model for prostate cancer
developed in the C57BL/6
inbred strain (N. Greenberg).


Specimen received IP contrast
media injection 30 minutes
prior to being sacrificed.


Specimen was frozen and
shipped to ORNL on dry ice.

MGP Knock
-
out Mice

(U.T. Memphis and Baylor University)


Baylor University
Matrix gla
protein

(MGP) knockout (G.
Karsenty).


MGP knockout exhibits
calcification of the arteries
and cartilage.


Specimen shipped to ORNL
on dry ice.

Amyloidosis Model

(University of Tennessee Medical Center)

MRI

X
-
ray CT

Automatic Organ Recognition


Initial study focuses on ORPK polycystic kidney
disorder.


Recognition methodology selected with the following
objectives:


combine segmentation and classification processes into one.


hybrid approaches, i.e., not purely bottom
-
up nor purely
top
-
down.


facilitate the incorporation of a priori information in a
manner that is simple and intuitive.


handle significant variability in image orientation and quality.


Selected algorithm based on “deformable contours”
approach.

Subject
-
to
-
Subject Variability

Training Process

16

“Leave one Out” Test (Normal Mouse)

10

10

“Leave one Out” Test (Normal Mouse)

8

“Leave one Out” Test (Normal Mouse)

8

“Leave one Out” Test (Normal Mouse)

Preliminary Results (PKD)

5

Preliminary Results (PKD)

5

Healthy Kidneys


Segmented Kidney Image


Local Variance (LV) Tests




Left


Right

Texture

(mean of LV)

29.9

34.2


Homogeneity

14.5

33.6

(Std. Dev. of LV)

PKD Kidneys



Local Variance (LV) Tests




Left


Right

Texture

(mean of LV)

104.0

215.2


Homogeneity

81.5

207.3

(Std. Dev. of LV)

Segmented Kidney Image

Healthy Kidneys

PKD Kidneys

104.0

215.2

65.17

130.7

29.9

34.2

27.8

37.9

Left

Right

Left

Right

Texture

Texture

Planned Improvements


Cone
-
beam reconstruction algorithm.


Artifact reduction.


Anesthesia (“patient motion” and
repeated studies.)


Increased throughput.


Improved contrast media administration.


Enhanced automated screening.


Summary


X
-
ray CT is a viable modality for low cost,
high throughput anatomic screening.


A working prototype small animal
microCT system has been developed.


Several studies have been performed.


Opportunities for improvement exist.

Concluding Remarks


Currently, we are addressing
the shortcomings of the
employed techniques by


incorporating additional
information into the training
process


looking into more robust
optimization techniques


incorporating criteria for
automatic termination

Rat Femur Studies

Re
-
Sliced MicroCAT Images

Sagittal Slice

Coronal Slice

Iodine contrast agent IP injected

Comparison of Modes

20 minute scan, IP contrast
agent, dead animal

7 minute scan, no contrast agent,
live animal

Hydronephrosis (Kidney failure)

Rat Backbone Studies

3
-
D Image Segmentation