Cost functions used in MIPAV algorithms and Creating a mask using the Paint Grow Segmentation method: Difference between pages

From MIPAV
(Difference between pages)
Jump to navigation Jump to search
m (1 revision imported)
 
m (1 revision imported)
 
Line 1: Line 1:
A similarity or cost function measures the similarity between two images. During the registration the adjusted image V is transformed using the various transformation functions. And the similarity S(U;Vt) between the reference image U and transformed image Vt is then calculated.
You can create a mask using the paint grow segmentation method, which uses voxel aggregation to group voxels into larger regions. You begin by selecting a voxel to serve as the ''seed point,'' or starting point. The software applies paint color to all voxels touching the seed point that fall within the intensity range that you specify.<br />'''Note:''' You cannot apply the paint grow segmentation method to RGB images.<br />


'''Note:''' in MIPAV, we use the term cost function to refer to the negative cost function. That means that we assume that the transformation that gives the smallest value of the chosen inverse cost function is the transformation that also gives the best alignment.
==== To create a mask using the Paint Grow tool ====
[[Image:FillsAreaIco.jpg]] - Fill an Area with Desired Color<br />
1 Click the Fill an Area with Desired Color icon. The Paint Grow dialog box appears (Figure 9).<br />
2 Select the seed point, which is used as a starting point for the paint grow operation. To do this, move the pointer to the image. As you move the cursor, the location and intensity of the voxel under the pointer tip appears in the Cursor position and voxel intensity text box. Click the voxel you want to designate as the seed point.<br />
3 Adjust the delta values and parameters. <br />
4 Click Apply when complete to begin the paint grow method. The Paint Grow dialog box closes.<br />''' To correct the mask<br /></font>'''</div>


== Background ==
If the results are not what you want, do the following:
[[File:CostFunctions1.jpg |200px|thumb|right|The plots of the Correlation Ratio cost function versus some of the individual parameter values. In each plot (A, B, C, and D), a single parameter is varied while the other are kept constant]]


For the registration algorithms, such as [[Optimized automatic registration 3D | OAR]], the main approach for determining an optimal transformation is to
1 Click the Paint Grow button. The Paint Grow dialog box appears.<br />
2 Select the seed point, which is used as a starting point for the paint grow operation. To do this, move the pointer to the image. As you move the cursor, the location and intensity of the voxel under the pointer tip appears in the Cursor position and voxel intensity text box. Click the voxel you want to designate as the seed point.<br />
3 Adjust the delta values and parameters. <br />
4 Click Apply when complete to begin the paint grow method. The Paint Grow dialog box closes.<br />
If the results are not what you want, do either of the following:<br />
Click the Undo last region paint icon, and start again.<br />
Click the Erase icon, or click, the Erase all paint icon, to erase all paint.<br />


<ol>
'''To commit the mask'''<br />
    <li>Calculate a cost function,</li>
    <li>Determine how it changes as individual transformation parameters are varied,</li>
    <li>And find the parameters that minimize the value of the cost function.</li>
</ol>


The following figure (right) shows a plot of a sample cost function for a selection of transformation parameters.
Click one of the following commit buttons:


The cost functions implemented in MIPAV:
[[Image:MasksInsidePaintedAreaIco.jpg]] - the Masks Inside Painted Area icon.<br />[[Image:MasksOutsidePaintedAreaIco.jpg]] - the Masks Outside Painted Area icon.<br />


*[[#CorrelationRatio | Correlation ratio]]
==== Paint Grow dialog box options ====
*[[#LaplacianZeroCrossing | Laplacian Zero Crossing]]
*[[#GradientMagnitude | Gradient Magnitude]]
*[[#LeastSquares | Least Squares]]
*[[#NormalizedCrossCorrelation | Normalized Cross Correlation]]
*[[#NormalizedMutualInformation | Normalized Mutual Information]]


== Powell algorithm ==
{| border="1" cellpadding="5"
|+ '''Figure 9. Paint Grow dialog box '''
|-
|
<div class="CellBody"><span style="font-style: normal; text-decoration: none; text-transform: none; vertical-align: baseline">'''<font color="#000000">Cursor position and voxel intensity</font>'''</span></div>
|
<div class="CellBody">Indicates the coordinates and intensity of the pixel under the mouse pointer tip. This pixel is the seed point.</div>
| rowspan="3" colspan="1" |
[[Image:PaintGrow_StaticThreshold.jpg]]
|-
|
<div class="CellBody"><span style="font-style: normal; text-decoration: none; text-transform: none; vertical-align: baseline">'''<font color="#000000">Delta above selected voxel intensity</font>'''</span></div>
|
<div class="CellBody">Once a seed point has been selected, MIPAV uses this value to determine whether to fill adjacent voxels with color. The voxels that are filled must have intensity levels that fall within the range of the seed point intensity minus the lower delta value and the upper delta value. For example, if seed point has an intensity of 100.0, and the <span style="font-weight: normal; text-decoration: none; text-transform: none; vertical-align: baseline">''<font color="#000000">Delta Above Selected Pixel Intensity </font>''</span>value is 10 and the <span style="font-weight: normal; text-decoration: none; text-transform: none; vertical-align: baseline">''<font color="#000000">Delta Below Selected Pixel Intensity </font>''</span>value is 15, MIPAV color-fills adjacent voxels whose intensities range from 85.0 to 110.0.</div>
|-
|
<div class="CellBody"><span style="font-style: normal; text-decoration: none; text-transform: none; vertical-align: baseline">'''<font color="#000000">Delta below selected voxel intensity</font>'''</span></div>
|
<div class="CellBody">Once a seed point is, MIPAV uses this value to determine whether to fill adjacent voxels with color. The voxels that are filled must have intensity levels that fall within the range of the seed point intensity minus the lower delta value and the upper delta value.</div>
|-
| rowspan="1" colspan="1" |
<div class="CellBody"><span style="font-style: normal; text-decoration: none; text-transform: none; vertical-align: baseline">'''<font color="#000000">Parameters:</font>'''</span></div> <div class="CellBody"><span style="font-style: normal; text-decoration: none; text-transform: none; vertical-align: baseline">'''<font color="#000000">Unrestricted size</font>'''</span></div> <div class="CellBody"><span style="font-style: normal; text-decoration: none; text-transform: none; vertical-align: baseline">'''<font color="#000000">Maximum size</font>'''</span></div> <div class="CellBody"><span style="font-style: normal; text-decoration: none; text-transform: none; vertical-align: baseline">'''<font color="#000000">Unrestricted distance</font>'''</span></div> <div class="CellBody"><span style="font-style: normal; text-decoration: none; text-transform: none; vertical-align: baseline">'''<font color="#000000">Maximum distance</font>'''</span></div>
| rowspan="1" colspan="2" |
<div class="CellBody">Constrains the growth of the paint grow operation. Select the <span style="font-style: normal; text-decoration: none; text-transform: none; vertical-align: baseline">'''<font color="#000000">Unrestricted size</font>'''</span> and <span style="font-style: normal; text-decoration: none; text-transform: none; vertical-align: baseline">'''<font color="#000000">Unrestricted distance</font>'''</span> check boxes to allow the paint grow operation to be applied without restraint. If the <span style="font-style: normal; text-decoration: none; text-transform: none; vertical-align: baseline">'''<font color="#000000">Unrestricted size</font>'''</span> check box is not selected, type the <span style="font-style: normal; text-decoration: none; text-transform: none; vertical-align: baseline">'''<font color="#000000">maximum size</font>'''</span> (in cubic meters) of the paint grow region in the text box. If the <span style="font-style: normal; text-decoration: none; text-transform: none; vertical-align: baseline">'''<font color="#000000">Unrestricted distance</font>'''</span> check box is not selected, type the maximum distance from the original seed point in the text box.</div>
|-
| rowspan="1" colspan="3" |
<div class="CellBody"><span style="font-style: normal; text-decoration: none; text-transform: none; vertical-align: baseline">'''<font color="#000000">Fuzzy connectedness</font>'''</span></div>
|-
|
<div class="CellBody"><span style="font-style: normal; text-decoration: none; text-transform: none; vertical-align: baseline">'''<font color="#000000">Fuzzy connectedness</font>'''</span></div>
| rowspan="1" colspan="2" |
<div class="CellBody">Check tis box to use the fuzzy connectedness coefficient instead of static threshold. Here, Fuzzy connectedness represents the idea of connection or "hanging-togetherness" of image elements in an object by assigning a strength of connectedness to every possible path between every possible pair of image elements. </div> <div class="CellBody">A fuzzy connected object is defined with a fuzzy threshold or the strength of connectedness. </div>
|-
|
<div class="CellBody"><span style="font-style: normal; text-decoration: none; text-transform: none; vertical-align: baseline">'''<font color="#000000">Initial variance from selected VOI</font>'''</span></div>
|
<div class="CellBody">Uses the initial intensity values from the selected region of interest (VOI).</div>
| rowspan="6" colspan="1" |
[[Image:PaintGrow_FuzzyConnectedness.jpg]]
|-
|
<div class="CellBody"><span style="font-style: normal; text-decoration: none; text-transform: none; vertical-align: baseline">'''<font color="#000000">Display fuzzy image</font>'''</span></div>
|
<div class="CellBody">Displays the result image in a separate frame. </div>
|-
|
<div class="CellBody"><span style="font-style: normal; text-decoration: none; text-transform: none; vertical-align: baseline">'''<font color="#000000">Fuzzy threshold</font>'''</span></div>
|
<div class="CellBody">is a threshold on the strength of connectedness of image elements.</div>
|-
|
<div class="CellBody"><span style="font-style: normal; text-decoration: none; text-transform: none; vertical-align: baseline">'''<font color="#000000">Close</font>'''</span></div>
|
<div class="CellBody">Closes this dialog box.</div>
|-
|
<div class="CellBody"><span style="font-style: normal; text-decoration: none; text-transform: none; vertical-align: baseline">'''<font color="#000000">Cancel</font>'''</span></div>
|
<div class="CellBody">Disregards any changes that you made in this dialog box and closes the dialog box.</div>
|-
|
<div class="CellBody"><span style="font-style: normal; text-decoration: none; text-transform: none; vertical-align: baseline">'''<font color="#000000">Help</font>'''</span></div>
|
<div class="CellBody">Displays online help for this dialog box.</div>
|}


In MIPAV, most of the methods and algorithms use [http://math.fullerton.edu/mathews/n2003/PowellMethodMod.htm the Powell algorithm] to find the global minimum of the chosen cost function. For more information about the Powell algorithm, refer to [http://math.fullerton.edu/mathews/n2003/PowellMethodMod.htm http://math.fullerton.edu/mathews/n2003/PowellMethodMod.htm].
==== Examples of using the Paint Grow tool ====


<div id="CorrelationRatio"></div>
Here, is a step-by-step guide for selecting a region(s) of interest using the Paint Grow tool options. In this example we will use an image of the human eye and we will try to segment the blood vessels located on the retina.
== Correlation Ratio ==
First, make sure to adjust the contrast of your image so that the different tissues appear visually distinctive. For images with more than 8 bits per voxel you may want to use the various LUT available in MIPAV.<br />


Given two images I and R, the basic principle of the Correlation Ratio method is to search for a spatial transformation T and an intensity mapping f such that, by displacing R and remapping its intensities, the resulting image f(R* T) be as similar as possible to I. This could be achieved by minimizing the following correlation ratio function:
'''To segment blood vessels using the static threshold'''<br />
1 Open the Paint Grow dialog box.<br />
2 Use the mouse cursor to select the point on the image (on a blood vessel) which has the desired intensity value.<br />
3 Use the Change Paint Color option to select the color (other than red) which will be used for painting.<br />
4 On the Static Threshold tab, adjust the intensity thresholds so that the tissue you want to preserve is highlighted.<br />5 Check the Unrestricted size and Unrestricted distance options. This will allow the paint grow operation to be applied without restriction.<br />6 Watch the paint region growing.<br />


<math>
For example, when selecting the blood vessels, the image might look like the panel A for grayscale images or panel B for images after applying Blue LUT in the Figure 10 below.
min(T,f) of \displaystyle\sum\limits_{k} {I (x_k)-f(R(T(x_k)))}
</math>


which integrates over the voxel positions <math>x_k</math> in the image I.
=== Image types ===


Correlation Ratio can be used in multimodal image registration of Magnetic Resonance (MR), Computed Tomography (CT), and Positron Emission Tomography (PET) images.
{| border="1" cellpadding="5"
|+ '''Figure 10. The Paint Grow tool was used to locate the blood vessels on the grayscale image (A) first, and then on the same image after applying the Blue LUT (B). '''
|-
|
[[Image:RetinaVesselsGrayscale.jpg]]
|<div class="CellBody">A - the painted region appears in red (which is the default color)</div>
|-
|
[[Image:RetinaVessels BlueLUT.jpg]]
|<div class="CellBody">B - the painted region appears in green, because we selected it as a color contrast to LUT colors</div>
|[[Image:RetinaVeselsDialogOptions.jpg]]
|Dialog Options box
|}
'''To segment blood vessels using the Fuzzy Connectedness option'''<br />
1 Open an image of interest.<br />
2 You might consider to delineate a VOI on a region of the image which is of your interest, first. This is optional.<br />
1 Open the Paint Grow dialog box, and then open the Fuzzy Connectedness tab.<br />
2 Check the '''Fuzzy Connectedness''' check box to activate the tool.<br />
3 Check the '''Initial variance from selected VOI''' box (optional).<br />
4 Check the '''Display fuzzy image option''' to view the result in a new frame.<br />
5 Use the mouse cursor to select the point on the image (on a blood vessel) which has the desired intensity value.<br />
6 Adjust the Fuzzy thresholds so that the tissue you want to preserve is highlighted.<br />
7 Watch the paint region appeared in a new frame.<br />


<div id="LaplacianZeroCrossing"></div>
For example, when selecting the blood vessels, the image might look like the panel A or panel B for images in the Figure 11 below.


== Laplacian Zero-Crossing ==
{| border="1" cellpadding="5"
|+ '''Figure 11. The Fuzzy Connectedness option.'''
|-
|
[[Image:RetinaVesselsFuzzyGrayscale.jpg]]
|<div class="CellBody">a - the painted region appears in red </div>
|-
|
[[Image:RetinaVesselsFuzzyResult.jpg]]
|<div class="CellBody">b -the painted region also appears in a new frame</div>
|-
|
[[Image:RetinaVeselsDialogOptionsFuzzy.jpg]]
|<div class="CellBody">c</div>
|}


The laplacian zero-crossing is a binary edge feature used for edge detection, see also [[Edge Detection: Zero X Laplacian | Edge Detection: Zero X Laplacian algorithm]]. Convolution of an image with a laplacian kernel approximates the 2-nd partial derivative of the image. The laplacian zero-crossing corresponds to points of maximal (or minimal) gradient magnitude. Thus, laplacian zero-crossings represent "good" edge properties and should, therefore, have a low local cost meaning edge detection. If IL(q) is the laplacian of an image I at pixel q, then,


<math>
\begin{cases}
f_Z(q)= 0; if I_L(q) = 0\\
f_Z(q)= 1; if I_L(q)\neq 0
\end{cases}
</math>


Though, application of a discrete laplacian kernel to a digital image produces very few zero-valued pixels. Rather, a zero-crossing is represented by two neighboring pixels that change from positive to negative values. Of these two neighboring pixels, the one closest to zero is used to represent the zero-crossing. The resulting feature cost contains single-pixel wide cost "canyons" used for boundary localization.
[[Segmenting Images Using Contours and Masks:Converting VOI contours to masks]]
 
<div id="GradientMagnitude"></div>
== Gradient Magnitude ==
 
Gradient Magnitude provides a direct correlation between the edge strength and local cost. If <math>I_x</math> and <math>I_y</math> represent the partials of an image I in X and Y directions respectively, then the gradient magnitude G is approximated with the following formula:
 
<math>
G=\sqrt{I^2_x + I^2_y}
</math>
 
The gradient is then scaled and inverted, so high gradients produce low costs and vice-versa. Thus, the gradient component function is
 
<math>
f_G =  \frac{max(G)-G}{max(G)} =1 - \frac{G}{max(G)}
</math>
 
Finally, gradient magnitude costs can be scaled by Euclidean distance. To keep the resulting maximum gradient at unity, <math>f_G(q)</math> can be scaled by 1, if q is a diagonal neighbor to p and by <math>\frac {1}{\sqrt{2}}</math>, if q is a horizontal or vertical neighbor.
 
<div id="LeastSquares"></div>
== Least Squares ==
 
Least squares measures the average of the squared difference in image intensities.
 
<math>
\frac{\sum_{i=1}^{N}\{R(p_i) - I(p_i)\}^2}{N}
</math>
 
Where,
*R(p_i) = reference image(p_i) - minimum reference image value
*I(p_i) = input image(p_i) - minimum input image value
*N = the number of values over which the sum is performed
 
It can be divided by count to normalize the cost function or make it invariant to the number of voxels in the region of overlap. If two images show large differences in image intensity, then we recommend using Scaled Least Squares with a global scaling term added:
 
<math>
\frac{\sum_{i=1}^{N}\{R(p_i) - s*I(p_i)\}^2}{N}
</math>
 
=== Image types ===
 
It can be shown that Least Squares is the optimum cost function when two images only differ by Gaussian noise. Images in two different modalities, such as PET and MRI, will never differ by only Gaussian noise. Even two images in the same modality, such as two PET images, will seldom only differ by Gaussian noise as medical scan noise is frequently not Gaussian and because the object often changes between the image acquisitions. The effectiveness of this cost function will be greatly diminished by small numbers of voxels having large intensity differences.
 
<div id="NormalizedCrossCorrelation"></div>
 
== Normalized Cross Correlation ==
 
The Cross-Correlation function can be defined as
 
<math>
CrossCorr(s,t) = \sum_{x}\sum_{y}R(x,y)I(x-s, y-t)
</math>
 
Where,
*R-reference image intensity
*I-input image intensity
 
The summation is taken over the region (s,t) where R and I overlap. For any value of (s,t) inside R(x,y) the cross-correlation function yields one value of CrossCorr. The maximum value of CrossCorr(s,t) indicates the position where I(x,y) best matches R(x,y).
 
Normalized Cross Correlation is 0 for identical images and approaches 1 for images that are very different.
 
=== Image types ===
 
The Cross-Correlation function can be used for aligning the images that have a linear relationship between the intensity values in the images. E.g. images acquired using the same modality. However, the Normalized Cross-Correlation may not work very well in two images that are identical except for a global scaling factor.
 
<div id="NormalizedMutualInformation"></div>
 
== Normalized Mutual Information ==
 
Mutual information (MI) measures how well one image explains the other. In medical image processing, MI is often used as
*a similarity measure for image registration for images that were acquired at different times or by different modalities and
*for combining multiple images to build 3D models.
 
The mutual information  of random variables A and B is defined as:
 
<math>
MI(A,B) = \sum_{ab}p(a,b) log \frac{p(a,b)}{p(a) p(b)}
</math>
 
Where, p(a,b) is the joint probability distribution function of A and B, and p(a) and p(b) are the marginal probability distribution functions of A and B respectively.
 
MI could also be defined as MI(A,B) = H(A) + H(B)-H(A,B),
 
Where H(A) and H(B) are the [http://www.pandasthumb.org/archives/2004/05/shannon-entropy.html Shannon entropy] of image A and B respectively, computed  based on the probability distributions of their grey values. And H(A,B) denotes the conditional entropy, which is based on the conditional probabilities p(a |b) - the chance of grey value a in image A given that the corresponding voxel in B has grey value b. MI measures the distance between the joint distributions of the images' gray values p(a,b) and the distribution when assume that images are independent from each other.
 
=== Normalized mutual information ===
Normalized mutual information can be calculated as normalized MI, where
<math>NMI(A,B) = (H(A) + H(B))/H(A,B)</math>.
 
=== Image types ===
 
The normalized mutual information has been shown to work very well for registering multi-modality images and also time series images. In MIPAV the normalized mutual information approaches 0 for identical images and approaches 1 for images that are very different.
 
== References ==
 
Bjorn Hamre "Three-dimensional image registration of magnetic resonance (MRI) head volumes" Section for Medical Image Analysis and Informatics Department of Physiology & Department of Informatics University of Bergen, Norway.
 
Woods R.P., Handbook of Medical Image Processing and Analysis, Chapter 33 "Within-Modality Registration Using Intensity-Based Cost Functions", Editor, Isaac N. Bankman, Academic Press, 2000, pp. 529-553.
 
Mortensen E., Barrett W., "Intelligent scissors for image composition", International Conference on Computer Graphics and Interactive Techniques archive. Proceedings of the 22-nd annual conference on Computer graphics and interactive techniques, pp. 191 - 198, 1995, ISBN:0-89791-701-4.
 
Josien P. W. Pluim, J. B. Antoine Maintz and Max A. Viergever. Mutual information based registration of medical images: a survey. IEEE TRANSACTIONS ON MEDICAL IMAGING, VOL. XX, NO. Y, MONTH 2003.
 
Powell method to find the global minimum: [http://math.fullerton.edu/ mathews/n2003/PowellMethodMod.html http://math.fullerton.edu/ mathews/n2003/PowellMethodMod.html]
 
== See also ==
*[http://mipav.cit.nih.gov/wiki/index.php/Main_Page MIPAV WIKI]
*[[Using MIPAV Algorithms]]
*[[Degrees of freedom]]
 
[[Category:Help]]
[[Category:Help:Algorithms]]

Latest revision as of 16:28, 9 February 2024

You can create a mask using the paint grow segmentation method, which uses voxel aggregation to group voxels into larger regions. You begin by selecting a voxel to serve as the seed point, or starting point. The software applies paint color to all voxels touching the seed point that fall within the intensity range that you specify.
Note: You cannot apply the paint grow segmentation method to RGB images.

To create a mask using the Paint Grow tool

FillsAreaIco.jpg - Fill an Area with Desired Color
1 Click the Fill an Area with Desired Color icon. The Paint Grow dialog box appears (Figure 9).
2 Select the seed point, which is used as a starting point for the paint grow operation. To do this, move the pointer to the image. As you move the cursor, the location and intensity of the voxel under the pointer tip appears in the Cursor position and voxel intensity text box. Click the voxel you want to designate as the seed point.
3 Adjust the delta values and parameters.

4 Click Apply when complete to begin the paint grow method. The Paint Grow dialog box closes.
To correct the mask

If the results are not what you want, do the following:

1 Click the Paint Grow button. The Paint Grow dialog box appears.
2 Select the seed point, which is used as a starting point for the paint grow operation. To do this, move the pointer to the image. As you move the cursor, the location and intensity of the voxel under the pointer tip appears in the Cursor position and voxel intensity text box. Click the voxel you want to designate as the seed point.
3 Adjust the delta values and parameters.
4 Click Apply when complete to begin the paint grow method. The Paint Grow dialog box closes.
If the results are not what you want, do either of the following:
Click the Undo last region paint icon, and start again.
Click the Erase icon, or click, the Erase all paint icon, to erase all paint.

To commit the mask

Click one of the following commit buttons:

MasksInsidePaintedAreaIco.jpg - the Masks Inside Painted Area icon.
MasksOutsidePaintedAreaIco.jpg - the Masks Outside Painted Area icon.

Paint Grow dialog box options

Figure 9. Paint Grow dialog box
Cursor position and voxel intensity
Indicates the coordinates and intensity of the pixel under the mouse pointer tip. This pixel is the seed point.

PaintGrow StaticThreshold.jpg

Delta above selected voxel intensity
Once a seed point has been selected, MIPAV uses this value to determine whether to fill adjacent voxels with color. The voxels that are filled must have intensity levels that fall within the range of the seed point intensity minus the lower delta value and the upper delta value. For example, if seed point has an intensity of 100.0, and the Delta Above Selected Pixel Intensity value is 10 and the Delta Below Selected Pixel Intensity value is 15, MIPAV color-fills adjacent voxels whose intensities range from 85.0 to 110.0.
Delta below selected voxel intensity
Once a seed point is, MIPAV uses this value to determine whether to fill adjacent voxels with color. The voxels that are filled must have intensity levels that fall within the range of the seed point intensity minus the lower delta value and the upper delta value.
Parameters:
Unrestricted size
Maximum size
Unrestricted distance
Maximum distance
Constrains the growth of the paint grow operation. Select the Unrestricted size and Unrestricted distance check boxes to allow the paint grow operation to be applied without restraint. If the Unrestricted size check box is not selected, type the maximum size (in cubic meters) of the paint grow region in the text box. If the Unrestricted distance check box is not selected, type the maximum distance from the original seed point in the text box.
Fuzzy connectedness
Fuzzy connectedness
Check tis box to use the fuzzy connectedness coefficient instead of static threshold. Here, Fuzzy connectedness represents the idea of connection or "hanging-togetherness" of image elements in an object by assigning a strength of connectedness to every possible path between every possible pair of image elements.
A fuzzy connected object is defined with a fuzzy threshold or the strength of connectedness.
Initial variance from selected VOI
Uses the initial intensity values from the selected region of interest (VOI).

PaintGrow FuzzyConnectedness.jpg

Display fuzzy image
Displays the result image in a separate frame.
Fuzzy threshold
is a threshold on the strength of connectedness of image elements.
Close
Closes this dialog box.
Cancel
Disregards any changes that you made in this dialog box and closes the dialog box.
Help
Displays online help for this dialog box.

Examples of using the Paint Grow tool

Here, is a step-by-step guide for selecting a region(s) of interest using the Paint Grow tool options. In this example we will use an image of the human eye and we will try to segment the blood vessels located on the retina. First, make sure to adjust the contrast of your image so that the different tissues appear visually distinctive. For images with more than 8 bits per voxel you may want to use the various LUT available in MIPAV.

To segment blood vessels using the static threshold
1 Open the Paint Grow dialog box.
2 Use the mouse cursor to select the point on the image (on a blood vessel) which has the desired intensity value.
3 Use the Change Paint Color option to select the color (other than red) which will be used for painting.
4 On the Static Threshold tab, adjust the intensity thresholds so that the tissue you want to preserve is highlighted.
5 Check the Unrestricted size and Unrestricted distance options. This will allow the paint grow operation to be applied without restriction.
6 Watch the paint region growing.

For example, when selecting the blood vessels, the image might look like the panel A for grayscale images or panel B for images after applying Blue LUT in the Figure 10 below.


Figure 10. The Paint Grow tool was used to locate the blood vessels on the grayscale image (A) first, and then on the same image after applying the Blue LUT (B).

RetinaVesselsGrayscale.jpg

A - the painted region appears in red (which is the default color)

RetinaVessels BlueLUT.jpg

B - the painted region appears in green, because we selected it as a color contrast to LUT colors
 RetinaVeselsDialogOptions.jpg Dialog Options box

To segment blood vessels using the Fuzzy Connectedness option
1 Open an image of interest.
2 You might consider to delineate a VOI on a region of the image which is of your interest, first. This is optional.
1 Open the Paint Grow dialog box, and then open the Fuzzy Connectedness tab.
2 Check the Fuzzy Connectedness check box to activate the tool.
3 Check the Initial variance from selected VOI box (optional).
4 Check the Display fuzzy image option to view the result in a new frame.
5 Use the mouse cursor to select the point on the image (on a blood vessel) which has the desired intensity value.
6 Adjust the Fuzzy thresholds so that the tissue you want to preserve is highlighted.
7 Watch the paint region appeared in a new frame.

For example, when selecting the blood vessels, the image might look like the panel A or panel B for images in the Figure 11 below.

Figure 11. The Fuzzy Connectedness option.

RetinaVesselsFuzzyGrayscale.jpg

a - the painted region appears in red

RetinaVesselsFuzzyResult.jpg

b -the painted region also appears in a new frame

RetinaVeselsDialogOptionsFuzzy.jpg

c


Segmenting Images Using Contours and Masks:Converting VOI contours to masks

Retrieved from "https://mipav.cit.nih.gov/pubwiki/index.php?title=Cost_functions_used_in_MIPAV_algorithms&oldid=72"