Note: Descriptions are shown in the official language in which they were submitted.
CA 02274363 1999-06-14
1
TITLE OF THE INVENTION
SYSTEM AND METHOD FOR SEGMENTATION OF
TWO-DIMENSIONAL MAGNETIC RESONANCE IMAGES
FIELD OF THE INVENTION
The present invention relates to magnetic resonance
imaging. More specifically, the present invention is concerned with
system and method for segmentation of two-dimensional magnetic
resonance images .
BACKGROUND OF THE INVENTION
As it is generally known, Magnetic Resonance Imaging
(MRI) is a non-ionizing process that can produce high contrast images of
anatomical structures. In the case of MRI, such structures can include
soft tissue. Using appropriate MR protocols, images of a selected
structure can thus be obtained with very high accuracy. The resulting
images can then be used to create a three-dimensional model of the
structure of interest.
However, since many other structures such as, for
example, synovial fluid, ligaments, and muscles are usually visible on the
images, the detection of a selected structure needs appropriate contour
CA 02274363 1999-06-14
2
selection. Conventional contour detection methods allow only the
detection of the contour of the complete image without any discrimination
between the different structures available on the image.
A method to accurately delineate tissues available on
MR images is thus desirable.
OBJECTS OF THE INVENTION
An object of the present invention is therefore to provide
an improved segmentation method for two-dimensional magnetic
resonance images.
BRIEF DESCRIPTION OF THE DRAWINGS
In the appended drawings:
Figure 1 is a schematic bloc diagram of a segmentation
system according to an embodiment of the present invention; and
Figure 2 is a flow chart of a method for segmentation of
magnetic resonance images according to an embodiment of the present
invention.
CA 02274363 1999-06-14
3
DESCRIPTION OF THE PREFERRED EMBODIMENT
Referring to Figure 1 of the appended drawings, a
segmentation system 10, according to an embodiment of the present
invention, will be described.
The segmentation system 10 includes a computer 12,
a storing device 14, an output device in the form of a display monitor 16,
and an input device 18. The storing device 14, the display monitor 16
and the input device 18 are all connected to the computer 12 via standard
connection means, such as, for example, wires.
The computer 12 can be a conventional personal
computer or any processing machine that includes a processor, a
memory and input/output ports (not shown). The input/output ports may
include network connectivity to transfer the images to and from the storing
device 14.
The storing device 14 can be, for example, a hard
drive, a cd-rom drive or other well known storing means. It can be directly
connected to the computer 12 or remotely via a computer network, such
as, for example the Internet. According to this embodiment of the
invention, the storing device 14 is used to store both the non-segmented
medical images as well as the resulting segmented images as computer
files. Those files can be stored in any format and resolution that can be
read by the computer 12.
CA 02274363 1999-06-14
4
The display monitor 16 is used to visualize the medical
images both before and after the segmentation process. With the input
device 18, the display monitor 16 also allows the input of guidance points
by the user as will be described hereinbelow. The display monitor 16 is
finally used to display a user interface, to facilitate the interaction
between
the user and the computer 12. It is believed within the reach of a person
of ordinary skills in the art to provide another output device that allows for
the visualization of the medical images.
The input device 18 can be a conventional mouse, a
keyboard or any other well known input devices or combinations thereof.
Of course, the computer 12 runs a software that
embodies the method of the present invention thereof.
Other aspects and characteristics of the system 10 will
become more apparent upon reading of the following description of a
segmentation method according to an embodiment of the present
invention.
Referring now to Figure 2 of the appended drawings,
generally stated, the segmentation method of the present invention
consist in performing the following steps, in sequence:
100- Starting the system 10;
CA 02274363 1999-06-14
5
102- Receiving a set of MRI images including a substructure;
for the first image of the set (stews 104 to 108
104- Selection by the user of the substructure contour;
106- Filling of the substructure contour on the first image of the set;
108- Cleaning of the first image;
for all the other images of the set~steps 110 to 120)
110- Estimating the center of the substructure of the previous image;
for all the angle positions s anning the substructure (steps 112 and 114)
112- Predicting the contour position;
114- Finding the pixel position belonging to the contour;
116- Connecting the pixels found;
118- Filling the substructure contour; and
120- Stopping the system 10.
These general steps will now be described in further
details by way of an example where two substructures are to be
segmented on MRI images: a bone and its cartilage.
Before describing these general steps in more details,
it is to be noted that steps 104 to 118 must be performed for every
substructure to be identified in the images. Hence, for this example,
steps 104 to 118 are performed first to identify the bone, and then to
CA 02274363 1999-06-14
6
identify the cartilage.
After the segmentation process 10 has been started
(step 100), the step 102 consists in receiving, by the computer 12, a set
of MRI images of a structure that includes the bone and the cartilage. It
is to be noted that the two substructures are not perfectly defined in the
images received. Also, in other application, the substructure can be any
part of a structure that can be identified by different gray levels so that
they can be extracted from the structure.
The images received are two dimensional arrays of
pixels that has been previously produced by a MRI scanner. It is to be
noted that the set of images is provided sequentially in the order that they
appear in the three-dimensional object. In other words, successive
images come from adjacent slices of the three dimensional object.
As it will become apparent upon reading the following
description, the segmentation of the bone on an image is based on the
shape of the bone detected on the previous image which consists of the
a priory information. Hence, the system has to be initialized by the user.
The first image has to be segmented manually. The remaining of the
images are segmented automatically.
In step 104, the contour of the bone is selected by the
CA 02274363 1999-06-14
7
user. Once the contour of the bone is closed, the computer 12 then fill
the selected contour (step 106) and clean everything on the image,
except the bone (step 108).
After the segmentation of the bone has been performed
by the user on the first image of the set, the computer 12 can proceed
with the segmentation on all the other images of the set.
Knowing the position of the bone contour on the
previous image, the computer 12 determines the center of the
substructure in the current image (step 110). This is achieved by
determining the position of each of the four sides of a rectangle that
closely includes the bone contour on the previous image and then by
determining the center position of that rectangle.
The computer 12 then predicts, in step 112, the position
of the contour along a radius spanning form the center position in the
current image. This is achieved for different angular positions around the
center point determined in step 110. The predicted position is the
corresponding position in the previous image, for the current angular
position value.
Using the predicted position along the current radius as
a starting point, the computer 12 finds, in step 114, the exact position of
the contour along that radius.
CA 02274363 1999-06-14
The computer achieves this by verifying if the label
(color) of the pixel at the current predicted position corresponds to the
label of the pixel at the previous angle position of the current image. If so,
the pixel is accepted as belonging to the contour, after verifying if the
distance from the center is within a predetermined range. If not, the pixel
position is stored as a possible candidate, and the verification is
performed with the pixels at the right and at the left of the predicted pixel,
each time storing the position as a possible candidate if the pixel does not
belong to the contour.
If the computer 12 found that none of the three pixels
belong to the contour, the computer 12 calculates parameters to verify if
one of the possible candidates can be choosen as a good candidate for
the studied angle. These parameters are based on the continuity
between the studied pixel and the previous pixel found on the contour,
the label of the current pixel and its position. More precisely, the
computer 12 uses the parameters to find the best match among the
candidates and verifies if this best match responds to predetermined
criteria.
If, according to the parameters values, none of the
possible candidates belongs to the contour, the computer 12 tags the
current angular position and repeat steps 112 and 114 with the remaining
angular positions.
CA 02274363 1999-06-14
9
When all the angular position that spans the
substructure have been processed by the computer 12, the pixel positions
corresponding to the previously tagged angular positions are interpolated
between the pixel positions corresponding to the adjacent angular
positions. A weight factor, corresponding to these pixel position, is
dynamically adjusted depending on the position of the missing pixel.
In step 116, the computer so connects the pixels found
as to provide a smooth closed contour.
Knowing the bone contour, the computer 12 fills the
bone on the image.
The computer 12 repeat steps 110 to 118 for all the
images.
To segment the cartilage on each of the images of the
set, steps 104 to 118 are repeated.
Since the segmentation of the cartilage is very similar to
the segmentation of the bone, only the differences between the two
process will be described herein.
It is to be noted that, although the above described
method can be used to segment most substructures, it has been found
CA 02274363 1999-06-14
10
advantageous to adapt the method according to the substructure to be
segmented.
With the segmentation of the cartilage, it is assumed
that a segmentation of the bone has been performed before. Since the
cartilage is close to the bone, the predicted pixel point of the contour of
the cartilage for each angular position is the pixel on the contour of the
bone for corresponding to the same angular position.
It has been found advantageous to pre-identified the
substructures among a pre-determined group of substructures. This
information helps the determination of the contour of the substructure of
interest since the computer can use more appropriate criteria in step 114
to determined if the distance of a pixel is at a correct position from the
center and to calculate the parameters allowing to verify if one of the
possible candidates belongs to the contour.
Although the present invention has been described for
the segmentation of bone and cartilage images, it can also be used to
segment other structures.
It is finally to be noted that, although the present
invention has been described hereinabove by way of preferred
embodiments thereof, it can be modified, without departing from the spirit
and nature of the subject invention as defined in the appended claims.