3D bone Segmentation and analysisx-ray image.ppt

Image Segmentation

1. Segmentation

2. 2D DXA measurement has limitation.
Extract BMD distribution of proximal femur from
single DXA image by constructing 3D bone shape.
A statistical model of the combined shape and BMD
distribution is presented, together with a method
for its construction from a set of quantitative
computed tomography (QCT) scans.
A reconstruction is acquired in an intensity based
3D-2D registration process whereby an instance of
the model is found that maximizes the similarity
between its projection and the DXA image.
Construct 3D of BMD Distribution

3. Osteoporosis is currently diagnosed from dual-energy X-ray absorptiometry (DXA), which
results in an image of the projected bone mineral density (BMD).
To provide a diagnosis, a T-score is computed relating to the number of standard
deviations above or below the mean for a population of healthy adults of the same sex
and ethnicity as the patient.
In current clinical practice, DXA derived BMD remains the common measure for
diagnosis.
Although DXA gives an accurate planar representation of the BMD, it is limited by its
two-dimensionality, and therefore does not represent the 3D shape or spatial
distribution of the BMD.
To overcome this limitation, a 3D representation of the femur bone with BMD
distribution can be acquired by quantitative computed tomography (QCT).
High financial costs, a high radiation dose for the patient.
Osteoporosis

4. Uses a statistical model of the combined shape and density
distribution.
Model is first constructed from a large dataset of QCT scans.
Data Preprocessing:
To prevent the soft tissue structures from interfering with the
registration process, a thresholding set to 60 mg/cm
3
.
Pelvis area and soft tissue structures are removed.
Shape Model Construction:
The intensity based registration process to construct the shape model.
A reference volume is chosen manually based on the smoothness of its
bone surface.
Segment the femur bone using active couture method.
To reduce the computation load of the registrations, a mask is
generated of the bone boundary (boundary mask) by dilating the
segmentation and subtracting an image mask of the erosion. In
addition, a mesh is extracted from the segmentation onto which some
additional processing is done to generate a smooth regular surface
mesh consisting of 7147 vertices.
3D Reconstruction From Planar
Images

5. Development of the proximal femur atlas.
The data for the atlas came from the QCT scans.
Volumetric DXA (VXA) compare four DXA images with projections of an iteratively modified
tetrahedral subject model.
The process begins by assuming an average femur derived from the femur atlas.
At each iteration, the pose (location and orientation in space), scale, and modes of variation
of the statistical model are varied to minimize the difference between the simulated DXA
images from projections of the model and the actual acquired DXA images.
aBMD in each pixel of the VXA projection calculated from the subject’s model is calibrated to
the measured aBMD value in the DXA image.
National Institute of Standard and Technology (NIST) (1850 mg/cm
3
) value for calibration.
Extract 3D Information From
Multiple In Vivo DXA Images

6. Femur is divided in following parts:
Head
Neck
Trochanter
Shaft
Current System Automatic
Segmentation of Femur

7. Calculate the threshold of each region in image.
Thin spline interpolation to calculate area of each bone part.
Area varies image to image.
X-ray image have strong and weak edges.
Canny edge detector to detect strong and weak edges in
image.
Segmentation

8. A pixel fall in between two threshold.
High and Low.
If pixel value lower then lower threshold set to 0.
If greater then higher threshold set to 1.
Measure the pixel value for each region like neck, ward,
trochanter and shaft region in image.
Area defined for neck region is rectangular, for ward is square
and for shaft and trochanter is triangular.
Head region is removed by detecting a spherical object in head
region.
Segmentation

9. Thanks!