Creation of DICOM--aware applications using ImageJ

Abstract

The demand for image-processing software for radiology applications has been increasing, fueled by advancements in both image-acquisition and image-analysis techniques. The utility of existing image-processing software is often limited by cost, lack of flexibility, and/or specific hardware requirements. In particular, many existing packages cannot directly utilize images formatted using the specifications in part 10 of the DICOM standard ("DICOM images"). We show how image analyses can be performed directly on DICOM images by using ImageJ, a free, Java-based image-processing package (http://rsb.info.nih.gov/ij/). We demonstrate how plug-ins written in our laboratory can be used along with the ImageJ macro script language to create flexible, low-cost, multiplatform image-processing applications that can be directed by information contained in the DICOM image header.

Fig 1

Importing a series of images in order of image title may lead to poorly ordered image stacks. A. When the character-equivalent of the image title is used, the title “image10” sorts between “image1” and “image2.” B. This problem is frequently solved by appending leading zeros to the numeric portion of the image title. The Import Dicom Sequence plug-in uses this solution when image number information (group:element 0020:0013) is not available in the DICOM header.

Fig 1

Importing a series of images in order of image title may lead to poorly ordered image stacks. A. When the character-equivalent of the image title is used, the title “image10” sorts between “image1” and “image2.” B. This problem is frequently solved by appending leading zeros to the numeric portion of the image title. The Import Dicom Sequence plug-in uses this solution when image number information (group:element 0020:0013) is not available in the DICOM header.

Fig 2

Use of Query Dicom Header to extract data from the DICOM header. A. A portion of the DICOM header from an image is displayed. Two entries, Echo Time and Receiving Coil are expanded for display. (Display from NeoLogica Dicom Dumper, v.1.1, http://www.neologica.it). The corresponding image is opened in ImageJ, and Query Dicom Header is used to query the Echo Time (B) and Receiving Coil (C). The result of the query is placed in ImageJ’s Results table. Type is coded 1 for numeric data, 2 for text data, and 9 for missing or invalid data. Query Dicom Header can also be triggered within a macro, and the result of the query retrieved within the macro from the results table.

Fig 2

Use of Query Dicom Header to extract data from the DICOM header. A. A portion of the DICOM header from an image is displayed. Two entries, Echo Time and Receiving Coil are expanded for display. (Display from NeoLogica Dicom Dumper, v.1.1, http://www.neologica.it). The corresponding image is opened in ImageJ, and Query Dicom Header is used to query the Echo Time (B) and Receiving Coil (C). The result of the query is placed in ImageJ’s Results table. Type is coded 1 for numeric data, 2 for text data, and 9 for missing or invalid data. Query Dicom Header can also be triggered within a macro, and the result of the query retrieved within the macro from the results table.

Fig 3

First application: windowing and leveling based on DICOM header fields. Dual echo spin-echo T2-weighted images are imported into a stack using the Import Dicom Sequence plug-in. The macro script used the Query Dicom Header plug-in to query the Window Center and Window Width fields in the DICOM header, and automatically sets the minimum and maximum signal intensities displayed in ImageJ based on this information.

Fig 3

First application: windowing and leveling based on DICOM header fields. Dual echo spin-echo T2-weighted images are imported into a stack using the Import Dicom Sequence plug-in. The macro script used the Query Dicom Header plug-in to query the Window Center and Window Width fields in the DICOM header, and automatically sets the minimum and maximum signal intensities displayed in ImageJ based on this information.

Fig 4

Macro script used to perform first application. The script clears the results table, runs the Query Dicom Header plug-in from within the macro script to extract the Window Center (group:element 0028:1050) and Window Width (group:element 0028:1051) fields in the DICOM header, performs error checking to ensure the values for these fields are present and valid, extracts the values from the results table, and uses these values to set the minimum and maximum pixel intensities displayed.

Fig 4

Macro script used to perform first application. The script clears the results table, runs the Query Dicom Header plug-in from within the macro script to extract the Window Center (group:element 0028:1050) and Window Width (group:element 0028:1051) fields in the DICOM header, performs error checking to ensure the values for these fields are present and valid, extracts the values from the results table, and uses these values to set the minimum and maximum pixel intensities displayed.

Fig 5

Second application: T2 parameter maps generated from dual echo spin-echo T2 weighted images. A. Dual echo spin-echo T2-weighted images are imported into a stack using the Import Dicom Sequence plug-in. B. The macro script uses the Query Dicom Header plug-in to extract Echo Time fields (group:element 0018:0081) from the DICOM header, and divides the stack into two substacks representing the long TR long TE and long TR short TE images respectively. C. The T2 parametric map is calculated using a standard formula by performing image mathematics on the two substacks and multiplying by the difference in echo times.

Fig 5

Second application: T2 parameter maps generated from dual echo spin-echo T2 weighted images. A. Dual echo spin-echo T2-weighted images are imported into a stack using the Import Dicom Sequence plug-in. B. The macro script uses the Query Dicom Header plug-in to extract Echo Time fields (group:element 0018:0081) from the DICOM header, and divides the stack into two substacks representing the long TR long TE and long TR short TE images respectively. C. The T2 parametric map is calculated using a standard formula by performing image mathematics on the two substacks and multiplying by the difference in echo times.

Fig 6

Third application: T1 parameter maps generated from multiple 3D spoiled gradient echo in steady state (SPGR) images using variable flip angles. A. In this example, five series of 3D SPGR images are imported into separate stacks using the Import Dicom Sequence plug-in. B. For the purposes of performing a linear correlation between images, each of the five stacks is converted to a pair of ordinate and coordinate stacks using Flip Angle information obtained using Query Dicom Header. C. The slope and intercept stacks obtained from the linear correlation of ordinate and coordinate stacks (not shown), along with information derived from the DICOM header placed in the log table (the number of pairs and the repetition time obtained using Query Dicom Header are listed), are used to calculate stacks of 32-bit quantitative maps of linear correlations (Rsq), equilibrium magnetization (S0) and T1.

Fig 6

Third application: T1 parameter maps generated from multiple 3D spoiled gradient echo in steady state (SPGR) images using variable flip angles. A. In this example, five series of 3D SPGR images are imported into separate stacks using the Import Dicom Sequence plug-in. B. For the purposes of performing a linear correlation between images, each of the five stacks is converted to a pair of ordinate and coordinate stacks using Flip Angle information obtained using Query Dicom Header. C. The slope and intercept stacks obtained from the linear correlation of ordinate and coordinate stacks (not shown), along with information derived from the DICOM header placed in the log table (the number of pairs and the repetition time obtained using Query Dicom Header are listed), are used to calculate stacks of 32-bit quantitative maps of linear correlations (Rsq), equilibrium magnetization (S0) and T1.