The overall goal of this software tool is to extract quantitative image features that fall into shape or texture such as gray level co-occurrence matrix categories. This tool can be used to answer some key questions in the field of radiomics. For example, you can extract quantitative image features and look at correlations between them and patient outcomes in radiation therapy.
The main advantage of this technique is that it's open source, it's easy to use, and it has a long history of people using it for extracting features for use in prognostic models. Radiomics has potential to provide information about the likely response of a tumor to certain treatments, and therefore can provide additional information that clinical staff can use when determining the appropriateness of alternative treatments. Radiomics can provide insight into the tumor phenotype, it can also be applied to allosistance such as quanitfying changes of more tissue throughout treatment.
Generally, individuals new to this method will struggle, because selecting the right settings to the different features such as the bit depth resampling is important, so it is vital to pay attention to these. Visual demonstration is critical as there are many steps and details. Careful attention to detail is important for good outcomes in radiomic studies.
To begin, install IBEX described in the accompanying text protocol. Once installed, load IBEX and create a new user in order to import and save your data. Next, click on the import icon, and select either the DICOM or Pinnacle file version nine format by clicking on the name of the format, and then click next.
To import all of the patients at once, check the batch folder box in the top right corner, and then click on the more button next to the listed directory. Select the folder that contains all patient folders with their corresponding DICOM images and radiation DICOM structure files. Finally, click the exit button.
To begin, load a patient file by first clicking on the data icon. Then in the select a patient section, click on the desired patient file to open. Next, in the select image set section, click on the image set of the desired patient file, and then click the open button.
Use the arrows to scroll through the images which are shown in the axial, coronal, and sagittal views. Then, select the zoom button and draw a box on the image plane to zoom. Right click on the image to zoom out.
Next, click on the window level button. Another button will pop up, when it does, drag the left and right sides to the desired area, or manually enter the window level in window and level boxes at the bottom of the new window. Select different window levels using the first drop-down menu on the left.
There are eight options for CT, and one option for PET. Visualize the regions of interest by clicking on the box next to the region of interest to make it visible on the images, or click on the On All ROIs"button to turn on all regions of interest. Next, prepare the data set in order to view the contours.
To accomplish this, first click on the show data set button to view contours in the data set. Once data sets are viewed, close the window. This is a new user with just-imported patients, so we have no data sets created yet.
Then, select the contours to add to the data set by clicking the check box next to their name, and click on the add to data set button. Click on the name of the data set to add the contours, or click the new button for a new data set. If new is selected, enter the name of the data set in the new window, and then click the OK button.
When finished, click the exit button to return to the list of patients and scans. Click on the feature icon, and add preprocessing if desired, by clicking on the add button and selecting the preprocessing option from the drop-down menu in the new window. Next, click on the i under Parameters to select the preprocessing parameters.
Select the number under the value column, type in the new parameter value, and then click the OK button. For description of the preprocessing method and the specific parameters, click on the question mark button in the upper right corner. Once satisfied, click the add button to add the preprocessing step.
In order to delete undesired preprocessing steps, select the preprocessing step, and click delete. In feature categories menu, select the gray level coocurrence matrix 25 in the neighbor instensity difference 25 feature categories to compute the gray level co-occurrence matrix and neighborhood intensity difference matrix respectively in 2.5D. This is accomplished by calculating on each slice individually, and then summing all the matrices together.
Another option is to select the gray level co-occurrence three in the neighbor intensity difference three feature categories to compute the gray level co-occurrence matrix and neighborhood intensity difference matrix, respectively, in 3D. Click on the I to see the selected parameters for that feature category. To change a parameter, enter the new value and click the OK button.
For additional information, click on the question mark in the upper right corner. Once all of the parameters are properly set, click the test button to view a feature category or specific feature by clicking on the button next to the feature or category desired. Select the data set to view the test on, and then click the open button.
Check the box next to the patients from the selected data set, and click the test button. Close the window once satisfied with the test. Uncheck any unwanted features for the category selected.
To uncheck all features, click on the word features"under set two, features. In order to add all of the selected features with the selected preprocessing, click the add to feature set button. Select the name and the feature set, and click the open button, or, to add these features to a new set, click the new button.
Enter the name of the feature set in the new window, and click OK.Next, click the show feature set button to view features with corresponding preprocessing techniques. Select the feature set to view, and then click open. Observe the preprocessing parameters and feature categories in dropdown menus.
Edit them by clicking on the I button under parameters. Once finished, close the window. In the main feature window, click on the show data set button to view current data sets.
Then, select the data set to view, and then click the open button. When finished, click the exit button. In order to prepare the output, first click on the result icon, then click on the data set to run the features on, under step one, and click on the feature set to run on the selected data under step two.
Next, click on the view data button to view the selected data set, followed by the view feature button to view the selected feature set. Finally, click the compute and save result button. Enter the file name for the results and select save.
The output from IBEX is a spreadsheet with three sheets. Shown here is a more complicated output with more regions of interest, and features selected than was run in this tutorial video. The first sheet contains the feature values for each region of interest placed in the data set.
For the gray level co-occurrence matrix and gray level run length matrix features, a value is output for each direction that the matrix is computed in, as well as an average matrix across all directions that is designated as 333. The fourth row states the feature category, and the fifth row states the header for that column, either designated the index, image, region of interest, MRN, or specific feature. The second sheet contains the data information.
This is information about the image that the region of interest was obtained from, such as the X, Y and Z voxyl dimensions. The third sheet contains the feature information. The features are separated by their category.
The parameters for the feature category are listed in the second column. All the selected features for that category are listed in the third column. The fifth column shows the preprocessing that was applied to the features, and the sixth column shows the parameters for the preprocessing.
Once mastered, this tool can be used to efficiently extract the texture data from medical images. It's important to remember to carefully review all data, including making sure the preprocessing and other feature settings are important. For example, use the IBEX review tool to carefully review the post-process images.
After its development, this technique paved the way for researchers in the field of radiomics to explore the use of images to stratify patients for outcomes like local regional control and cancer patients. After watching this video, you should have a good understanding of how to use IBEX to extract image features from medical images. You can then put those image features into statistical or machine learning approaches to learn something about tumor phenotype.
