Design and validation of Segment - freely available software for cardiovascular image analysis
© Heiberg et al; licensee BioMed Central Ltd. 2010
Received: 21 September 2009
Accepted: 11 January 2010
Published: 11 January 2010
Commercially available software for cardiovascular image analysis often has limited functionality and frequently lacks the careful validation that is required for clinical studies. We have already implemented a cardiovascular image analysis software package and released it as freeware for the research community. However, it was distributed as a stand-alone application and other researchers could not extend it by writing their own custom image analysis algorithms. We believe that the work required to make a clinically applicable prototype can be reduced by making the software extensible, so that researchers can develop their own modules or improvements. Such an initiative might then serve as a bridge between image analysis research and cardiovascular research. The aim of this article is therefore to present the design and validation of a cardiovascular image analysis software package (Segment) and to announce its release in a source code format.
Segment can be used for image analysis in magnetic resonance imaging (MRI), computed tomography (CT), single photon emission computed tomography (SPECT) and positron emission tomography (PET). Some of its main features include loading of DICOM images from all major scanner vendors, simultaneous display of multiple image stacks and plane intersections, automated segmentation of the left ventricle, quantification of MRI flow, tools for manual and general object segmentation, quantitative regional wall motion analysis, myocardial viability analysis and image fusion tools. Here we present an overview of the validation results and validation procedures for the functionality of the software. We describe a technique to ensure continued accuracy and validity of the software by implementing and using a test script that tests the functionality of the software and validates the output. The software has been made freely available for research purposes in a source code format on the project home page http://segment.heiberg.se.
Segment is a well-validated comprehensive software package for cardiovascular image analysis. It is freely available for research purposes provided that relevant original research publications related to the software are cited.
Applied medical research is becoming more and more dependent on imaging for evaluation of the therapeutic effects of new drugs or therapies. Thus, dedicated image analysis software is needed for quantitative medical imaging. Commercially available software often offers limited functionality and frequently lacks the validation that is required for clinical studies. On the other hand, open source software offers transparency and the ability to modify the source code is well-suited to academic research since it gives researchers the ability to see exactly how the algorithms are implemented. The current trend among research grant organisations is that the results of government-funded projects should be published in open access journals or should be otherwise publicly available. Consequently, open access publishing has had a noticeable effect on the ease with which scientific results become available.
Publication of results in peer-reviewed journals is the traditional way of documenting and mediating progress in science. For classical sciences such as medicine and physics, this is normally sufficient since the reader can incorporate the information given in the publication into his or her own research. However, medical image analysis research often involves complex algorithms, and one cannot easily incorporate the results into one's own research since the algorithms described in the scientific papers usually need to be re-implemented for use by other research groups . Thus, we believe that releasing medical image analysis software to other research groups as freeware has the potential to have a profound effect on medical image analysis and applied medical imaging research. If the source code is reusable and written in a standardized way, the application can be modified or extended with new algorithms. This would permit the development of new algorithms or refinement of existing algorithms in order to satisfy requirements that arise in a clinical research setting. Furthermore, scripting capabilities for medical image analysis software may open up new lines of research that were previously untestable, since manual analysis would not have been feasible. For example, in a recent study it was possible to classify 72 regional myocardial sectors according to the neighbouring sectors and to track them in 22 patients over 5 points in time, thus generating over 50,000 classified data points . This task would have been impossible without advanced scripting capabilities in the analysis software.
The Segment cardiac image analysis software package, which is the subject of this article, was originally developed by the first author and was released in 2005 on a freely available basis. Since then it has been downloaded by more than 2,000 unique users in 74 countries, and approximately 300 research groups. To date, it has been referenced in more than 40 scientific publications. One advantage of freely available medical image analysis software is that it facilitates multi-centre clinical trials since all the participating sites can use the same software. The authors are aware of two ongoing multi-national and multi-centre studies that are using Segment software. Until recently, the software was distributed as a precompiled Windows application. If the software was available in a source code format, other researchers could contribute with their own modules and improvements. Image processing experts could then directly integrate new features into the software and make these improved algorithms available to other researchers. In this way, algorithm developers would be able to focus on algorithm development and make use of the common basic functionality of the software such as image loading, image display, user interactions etc. We believe that this may not only increase scientific productivity, but more importantly it may also provide a bridge between the very latest image processing ideas and applied clinical research. Today, many very promising image processing ideas never reach clinical routine because it is too cumbersome to write prototype software that is sufficiently user-friendly to be used by clinical researchers. Thus, the aim of this article is to present the design and validation of a cardiovascular image analysis software package and to announce its release in a source code format.
What follows is an outline of some of the important details concerning design and implementation of the software.
Automated or semi-automated image processing tools in Segment
- Left ventricle
2D, 2D+T, 3D, 3D+T
- Semi-automatic tools for right ventricle
2D, 2D+T, 3D, 3D+T
- Phase unwrapping algorithm
2D+T, 3D+T, 3+3D+T
- Phase background correction
2D+T, 3D+T, 3+3D+T
- Automated vessel tracking
- Flow visualization
- Quantification of infarct size
- Infarct extent
General Object Segmentation
- Fast levelset
3D, 2D+T, 3D+T
3D, 2D+T, 3D+T
- Left ventricle segmentation
- Defect size
- Gated SPECT segmentation
Manual image processing tools in Segment
Image visualization tools
- Contrast adjust + auto contrast
2D, 2D+T, 3D, 3D+T, 3+3D+T
- Multi view/panel support
2D, 2D+T, 3D, 3D+T, 3+3D+T
- Image plane intersection
3D, 3D+T, 3+3D+T
Manual contouring tools
2D, 2D+T, 3D, 3D+T
Region of interest analysis (ROI)
- Signal intensity quantification
2D, 2D+T, 3D+T
- Histogram analysis
2D, 2D+T, 3D+T
- Visual ROI analysis
- Area tools
- Volume tools
2D, 2D+T, 3D, 3D+T
Reformating image tools
- Multi planar reconstruction
2D, 2D+T, 3D, 3D+T
Export, import and reporting capabilities of Segment
DICOM import and manipulation
2D, 2D+T, 3D, 3D+T
Movie recording capacity
Wall thickening analysis
Polar plot of function and infarct
Batch export to statistical software
Communication module to facilitate multicenter trials
Each loaded image or image stack (a single image (2D), a time-resolved single image (2D+T), a stack of a images covering a multi-slice image volume (3D) or a time-resolved multi-slice image volume (3D+T)) is stored as a struct, and has fields for storing contours, image data, image orientation, resolution, image acquisition details, delineations, annotations and measurements. Handles to the graphical user interface and some temporary data are stored in a global data structure. A set of low-level input and output user interface routines was implemented to improve portability of the code. An object-oriented system for graphical user interfaces is employed to improve the ability to maintain the code and simplify development of new user interfaces.
B. Programming environment
The software package is written in Matlab and, for time-critical sections of the code, standard ANSI-C is used coded with Mex-wrappers so that they can be called from Matlab. The complete software project consists of about 90,000 lines of Matlab code and about 10,000 lines of C code. In total, there are 44 separate user interface panels. Proper version control software is necessary when managing a programming project of this size. Version control is managed using the open source solution Subversion (SVN, http://subversion.tigris.org) with Tortoise SVN http://tortoisesvn.tigris.org as a shell extension. Feature requests and bug reports are managed using the web-based open source software Trac http://trac.edgewall.org. Up until now, the whole application has been compiled as a stand-alone application and distributed together with the Matlab Compiler Runtime Environment. Upon publication of this article, the source code, the user manual and the technical manual will be available on the Segment home page http://segment.heiberg.se. Precompiled versions of Segment will still be available for Windows and Linux (Ubuntu distribution). Precompiled object files for the C code will be made available for 32-bit and 64-bit Windows operating systems, 32-bit and 64 bit Linux, and Mac OS X.
C. Stability, accuracy and validity
For clinical image analysis, the following requirements of a software package are crucial: (1) stability, (2) high performance, and (3) accuracy and validity.
To achieve stability, the software was designed so that even when run-time errors occur, the application should not crash and the user interface should not end up in a state in which the user must to restart the application.
To optimise performance, effective memory management and highly optimised routines are essential when designing the software. Cardiac imaging is particularly demanding since the data sets are frequently large and multidimensional with regard to space and time. Great care was taken to avoid duplicating image data unless absolutely necessary. Also, time-critical routines were optimised and coded in C.
To achieve accuracy and validity, validation considerations were incorporated into the design process by placing all calculations of distances, areas, volumes and region of interest mask generation in well-validated sub-functions. Care was taken at the design phase to avoid loss of accuracy due to loss of numerical precision. One such consideration was to represent contours, measurements and regions of interest with double-precision floating point numbers. Surprisingly, this is not often the case in commercially available software tools where pixel-based approaches are frequently used, which can be quite misleading--especially in small regions of interest. Even with a robust underlying design, it remains a challenge to maintain a strict quality policy that allows use for clinical research. The solution that we have chosen is to write an extensive test script that runs on archived test data. The test script output from the software is then compared with known accurate results from previously validated scientific publications. To ensure that the software maintains a high standard, the complete software repository will only be made available to a limited number of trained developers and incorporation of user-contributed code in the code base will only be done after careful testing and quality control. With the use of the test script, it is possible to quickly test the entire software project and uncover unexpected side effects when the code is modified. Coding standards and quality policies are given in the technical manual, which is available on the project home page.
D. Software maintenance
Medical imaging is developing rapidly and as a result medical image analysis software must be continuously refined and maintained. To ensure long-term maintenance of the code, we have chosen a solution whereby a company was formed to support and commercialise the software for use in clinical practice or by commercial users. Lund University is a shareholder in this company, and researchers have and will continue to have access to the current and future versions of the source code of the software.
E. Terms of licence
Segment is freely available for academic investigational research use (studies paid by government-derived funds or donations) provided that the original research publications relevant to the software are cited. The software is also free for educational purposes. Note that the license terms do not generally include trials paid by pharmaceutical companies. For commercial use, Segment is sold and supported by the company Medviso AB, Lund, Sweden. Individuals or organisations are not allowed to compile software products derived from Segment that are to be sold commercially or shipped together with other commercial products without written permission from Medviso AB.
Segment is a full-featured software tool for cardiovascular image analysis and to date, it has been used in a wide range of publications ranging from technical algorithm descriptions [3–5] to applied research on the effects of cardiac gene therapy , perfusion MRI , perfusion multidetector CT , applied human physiology , validation of an imaging technique in clinical cardiology , analysis of infarction with MRI  and MDCT , and analysis of microinfarction , regional cardiac function , for the first time quantitatively determine the infarct evolution in man , brain imaging  and also experimental imaging in rodents . The software has been developed with a view to its use in cardiovascular magnetic resonance imaging (MRI) and myocardial perfusion single-photon emission computed tomography (SPECT), but in principle it can be used for image analysis in any organ system, and it has also been used for image analysis in computed tomography (CT) and positron emission tomography (PET). The following sections deal with the main features of Segment. Each section presents an overview of validation results and procedures where applicable.
A. Loading of DICOM images
DICOM images from all major MRI vendors, including both human and animal scanners (Bruker, GE, Philips, Siemens and Varian) can be loaded into the software. Correctness of the loaded image data is validated by the test script, which loads a large number of image types from all vendors and compares the results to previously manually validated results using ImageJ or proprietary vendor software. Checks are made for correct image sorting, image resolution, time increment between time frames, slice thickness etc.
B. Image display
A large set of image display tools has been implemented. Examples of functionality include simultaneous display of multiple image stacks, viewing of contours and regions of interest, image stack intersections, adjustment of image contrast and brightness, and scrolling over time and between image slices. Each image stack can have its own colour scale or contrast settings, and 12-bit color mapping is used internally. The user interface was developed to maximise the image display area without compromising user friendliness. Furthermore, multiple monitors are supported to maximise the display area and improve the work flow. Image display tools cannot be evaluated quantitatively, but the functionality is carefully tested by the test script. Tools for recording movies are also incorporated in the software. This greatly facilitates preparation of scientific presentations that include movies.
C. Automated segmentation of cardiac ventricular dimensions in MRI
Automated segmentation of the left ventricle in MRI was the first image process algorithm implemented in Segment. The algorithm has been described and validated [5, 17]. There was an excellent correlation between automated segmentation and manual segmentation for end diastolic volume (EDV), R2 = 0.99, with a mean error of -1 ± 11 ml, and left ventricle mass (LVM), R2 = 0.94, with a mean error of 4 ± 15 ml . Tools for semi-automatic delineation of the right ventricle have also been incorporated. From the automatically or semi-automatically segmented surfaces, wall thickness, wall thickening, and fractional wall thickening can be calculated [5, 17]. Typical computational time for a standard Windows XP desktop PC (Intel Dual Core 2 GHz, Buss speed 770 MHz and 2 GB RAM) is about 10 seconds for a typical data set with 12 slices, and 30 timeframes.
D. Flow quantification
Flow measurements are of major importance in cardiovascular research, and velocity encoded phase contrast (PC) magnetic resonance imaging (MRI) is the golden standard for the in vivo quantification of blood flow in large vessels. It has been shown that modern MRI scanners may have phase offsets due to eddy currents, which can have a large effect on clinical flow measurements . Segment has tools for compensation of linear and higher-order background phase offsets due to eddy currents or Maxwell effects . It is possible to use automatic detection of stationary tissue based on the temporal standard deviation of the phase or to use manual regions of interest.
When imaging velocities higher than the chosen velocity encoding range, phase aliasing will occur and cause wrap-around artifacts in the quantitative visualization of velocities. Tools to compensate for such aliasing effects have been implemented and used successfully . The algorithm detects temporal discontinuities in the phase, and pixels with a temporal phase jump pair are unwrapped. The tool also includes a graphical user interface for manual correction.
E. Tools for drawing object contours
All the necessary tools for manually drawing object contours, and regions of interest, linear measurements and annotations are implemented. The same tools can be used transparently to correct automated delineated object contours. All drawing tools include full undo capabilities. All quantitative measures from object contours rely on the same low-level quantification algorithms. These low-level algorithms were validated using computer phantoms implemented in the software, and are included in the test script.
F. Region of interest analysis
Tools for analysing regions of interest (ROIs) are implemented. Details about signal intensities and area are available. Areas of the regions of interest are measured by accurate polygon calculations, but signal intensity and signal intensity statistics are calculated by performing statistics on discrete pixels inside the ROI. Basic statistics such as mean intensity and standard deviations over time are available. Typical calculation time for calculating statistics for one ROI in a typical data set with 35 time frames is about 0.2 s on an ordinary Windows XP desktop PC, Intel Dual Core 2 GHz, Buss speed 770 MHz and 2 GB RAM.
Image intensities are stored in Segment as single precision floating-point numbers and internally scaled into the range 0-1. Signal intensity measurements are done with single-precision arithmetics which give relative errors of about 10-6). However, note that the DICOM standard specifies storage of images using between 12 and 16 bits. For CT images, they are internally stored in Segment as 16-bit signed integers and are converted to single-precision floats before any arithmetics besides general object segmentation or display.
G. Linear measurements and annotation points
Linear measurements can be made in single time frame and the distances between start and end points are calculated as the standard Euclidian distance. Distances and positions are stored as double-precision floats, and thus positions and distances can be measured with sub-pixel accuracy. Annotation points, both static in time and time-resolved, can be placed in the image volume, and coordinates for these annotation points can be exported. This also allows measurement of distances in time-resolved three-dimensional space.
H. MRI viability analysis
Delayed contrast enhancement MRI can be used to differentiate between viable and necrotic or fibrotic tissue, owing to changes in the extracellular distribution volume after irreversible cell injury . Automated tools for delineation of hyper-enhanced regions are implemented and have been clinically validated  and used in several studies, such as  and . Hyper-enhanced regions can be expressed in ml, as a percentage of left ventricle mass, or as endocardial extent . Furthermore, the method for delineation of hyper-enhanced regions has been improved with a novel paradigm to account for the partial volume effect, where the hyper-enhanced pixels are weighted depending on the image intensity . The method was validated in 8 animals by in vivo MRI compared to high-resolution ex vivo MRI. Further validation was performed in 40 computer phantoms and 40 patients, by comparing the results to manual delineations from three experienced observers. Values of mean bias ± variability (or standard deviation), expressed as a percentage of left ventricular myocardium (%LVM), were -0.3 ± 1.3% (animals), -1.2 ± 1.7% (phantoms) and 0.3 ± 2.7% (patients). The new weighted algorithm had lower variability than the previously published approach of dichotomously classifying pixels as wholly infracted or not (2.7 vs 7.7%LVM, P < 0.01) and was not statistically significantly different from inter-observer variability for bias (P = 0.31) or variability (P = 0.38). Also, calculation of weighted infarct transmurality has been implemented and validated . Typical calculation time for automated infarct delineation for a typical data set with 12 slices is about 0.1 s on an ordinary desktop PC.
I. General image object segmentation tools
Tools for segmentation of general objects have been implemented, and they are based on an approximate fast-level set algorithm . The segmentation can be exported as a three-dimensional mesh. This method has been described in detail already . Moreover, a novel prototype-based segmentation algorithm has been developed. This algorithm is capable of learning how to segment an object based on few examples, typically 5-10. This can be contrasted to traditional statistical model-based approaches, which generally required 50-100 training cases. In a test case the aorta was delineated in 10 healthy volunteers imaged using a steady-state free precession sequence, resolution 1.6 × 1.6 × 1.0 mm. Five volunteers were used as a learning set and five as test case. The volumetric error was 4 ± 5% when measured as volumetric error and 0.58 ± 0.06 mm when measured as distance error.
J. Left ventricle segmentation in SPECT images
Automated delineation of the left ventricle in myocardial perfusion SPECT has been implemented. The result of the fully automated segmentation in myocardial perfusion SPECT images of 100 patients was compared to that from manual planimetry of MRI images of the same patients as a reference standard. The mean error and variability compared to MRI was 6 ± 15%LVM, which was significantly lower than for the commercially available algorithm Quantitative Perfusion SPECT (QPS, Cedars-Sinai Health Systems): 18 ± 19%LVM . Typical time for fully automated LV segmentation is about 8 s on an ordinary desktop PC.
K. Image fusion
A tool for manual rigid body co-registration of image stacks was implemented and was evaluated in an imaging study where ex vivo data from 19 pigs using myocardial perfusion SPECT and MRI were successfully co-registered .
L. Reformatting image tools
General image tools such as multi-planar reconstruction, re-sampling and rotation have been implemented. Accuracy of re-sampling was validated by re-sampling of computer phantoms with known geometric properties. The image quality of re-sampled images is maximized by using bi-cubic interpolation with anti-aliasing filters. Testing of the general image tools is included in the test script.
M. Polar plots
Quantitative results of regional functional parameters such as wall thickening and infarct transmurality from viability analysis can be displayed as polar plots, including segmentalisation of the left ventricle according to the 17-segment model endorsed by the American Heart Association . Development and validation of this functionality has been described , and an example of how this can be used to register quantitative data in consecutive MRI studies has been published .
N. Export capabilities
A wide range of exportation algorithms were implemented. Export direct to the system clipboard can be done from most graphical user interface panels. Data in the clipboard can then be pasted directly into statistical spread-sheet software such as Microsoft Excel. Tools for batch exportation from multiple files are also implemented. This avoids tedious manual interaction, and it also eliminates manual mistakes in the exportation process.
O. Communication module
A module to facilitate communication between research groups and to facilitate multi-national and multi-centre studies is included in the software. Since the software is freely available to researchers, all participating sites can use the same software. Images are loaded into the software by the imaging site. Using the software, all images from one CMR examination are (optionally) analysed, anonymised, compressed and electronically transferred to a central server set up by the coordinating centre.
The coordinating centre can control the total image flow by distributing "site keys" to each participating site. The site key includes instructions to the imaging centre and tells the software where images from a specific site should be stored. No passwords or usernames need to be sent out to participating sites. The site key also contains the encryption/decryption key.
P. Plug-in and scripting capability
Possibilities for writing one's own dedicated plug-ins have been implemented and details are given in the technical manual. A detailed description is outside the scope of this article. A plug-in template is available, consisting of less than 30 lines of Matlab code, and user plug-ins are available directly main menu in Segment. Users have access to the complete internal data structure and have control over all elements of the graphical user interface. Scripting can be performed directly in Matlab, thus providing powerful scripting possibilities.
Q. Test script
Testing includes user interface, display, and low-level algorithms such as automated LV segmentation, vessel delineation, flow quantification, DE-MRI viability analysis etc. Currently, the test script comprises 66 main test cases and a total of 316 tests. The test script is continuously refined as new features are added to the program. A script that checks for broken call-back links in the user interface is also implemented.
Previous work on extendable software for medical imaging can be divided into two major categories: (1) complete applications, and (2) toolkits such as MITK http://www.mitk.org . These two different approaches to extendable medical analysis software have their own strengths and weaknesses. Early on, we considered whether Segment should be a toolkit or a complete application. It was decided that an application approach would be most beneficial to clinical research since it will allow the great majority of users to be able to use it without having to write any source code. One pioneering and freely available extendable medical imaging software package is ImageJ http://rsbweb.nih.gov/ij/, written by Rasband at the U.S. National Institutes of Health. ImageJ has become widely popular, probably because of the open platform that allows users to write their own plug-ins, thus rapidly increasing the applicability of the software. Another open source general viewing and image processing application that also has a complete plug-in architecture is OsiriX http://www.osirix-viewer.com/, which runs under Mac OS X. Yet another well-known application is Slicer http://www.slicer.org/ . One example of extendable software for medical imaging that is not open source is Analyze (AnalyzeDirect, Lenexa, KS; http://www.analyzedirect.com/). The majority of freely available medical research software approaches have been designed for analysis of the brain, such as SPM http://www.fil.ion.ucl.ac.uk/spm/software/spm2/ and Internet Image Viewer http://james.psych.umn.edu/iiV/ , and not cardiovascular applications.
In summary, Segment differs compared to previous approaches in that to our knowledge it is the first source code extendable software application dedicated for cardiac image analysis. It is also to our knowledge by far the largest solution for applied medical imaging written in Matlab, a computer language that is widely used for image processing research. The strength of Segment lies in the combination of both a clinically applicable tool and a tool that easily can be further expanded by image processing experts and directly be used for clinical research.
Predicted use of Segment
We predict that Segment will continue to be used by cardiovascular researchers and in research groups with engineering teams. New functionality required to answer clinical research questions will be implemented and made available to the research community. The scripting functionality will enhance and facilitate larger clinical studies that are required to help cardiovascular imaging, and cardiovascular MRI in particular, to become an outcome-based imaging modality in medicine. Our hope is that Segment will function as a bridge between researchers in the field of image processing and researchers in cardiovascular research, both clinical and pre-clinical.
Importance of validation
Accurate and careful validation is of crucial importance in cardiovascular research. One feature that makes this project stand out from many other freely available medical image analysis software packages is the careful scientific validation that is performed when developing the new algorithms used in the software. However, one must remember that a chain is no stronger than its weakest link, and this is certainly true of quantitative image analysis. One such example is that flow quantification by MRI is validated in Segment, but it has been shown that each MRI scanner is unique; different pulse sequences have to be individually validated for each scanner since some scanners can introduce large sources of errors . Another example is automated segmentation algorithms, which need to be supervised by trained and experienced observers to achieve the highest accuracy in clinically relevant measures.
Segment is a cardiovascular image analysis software package that has been used in over 40 peer-reviewed scientific publications, indicating that the software has had an impact on cardiovascular research. Segment is a well-validated and comprehensive package that is freely available in an open source format for research purposes.
Availability and requirements
The project name is Segment and the project home page is http://segment.heiberg.se. Pre-compiled versions of the software will be made available for Windows and Linux. The Matlab source code version of the program requires Matlab R2008a or later. The software is known to run under Mac OS X, but at the moment this is not supported. Segment is freely available for academic investigational research use, provided that relevant original research publications related to the software are cited. The software is also free for educational purposes. The terms of the licence do not generally include trials paid by pharmaceutical companies. For commercial use, Segment is sold and supported by Medviso AB, Lund, Sweden. Individuals or organisations are not allowed to compile software products derived from Segment that are to be sold commercially or shipped together with other commercial products without the express written permission of Medviso AB.
Delayed Enhanced Magnetic Resonance Imaging
Magnetic Resonance Imaging
Left Ventricle Mass
Picture Archiving Communication System
Single Photon Emission Computed Tomography
Positron Emission Tomography.
The authors thank all the users of Segment who have contributed with bug reports and suggestions for future enhancements. We also thank all the individuals who have contributed to Segment with code: Erik Bergvall (mathematical routines for strain analysis), Andreas Sigfridsson (parts of multiplanar reformatting), Helen Soneson (image fusion, strain analysis and LV SPECT segmentation), Erik Södervall (debugging and improvements), Johannes Töger (debugging of flow coordinate systems and implementation of 3D flow tools) and Johan Ugander (debugging and implementation of the point curve tool). We are grateful to the Swedish Research Council, the Swedish Heart and Lung Foundation and Scania Region for financial support.
• DICOM is a registered trademark of the National Electrical Manufacturers Association for its standards publications relating to digital communication of medical information.
• Linux is the registered trademark of Linus Torvalds in the U.S. and other countries.
• Mac OS X is a registred trademark of Apple Inc.
• Matlab is a registered trademark of The MathWorks Inc.
• Windows is a registered trademark of Microsoft Corporation Inc.
- Wolf I, Vetter M, Wegner I, Bottger T, Nolden M, Schobinger M, Hastenteufel M, Kunert T, Meinzer HP: The medical imaging interaction toolkit. Medical image analysis. 2005, 9: 594-604. 10.1016/j.media.2005.04.005.View ArticlePubMedGoogle Scholar
- Engblom H, Hedström E, Heiberg E, Wagner GS, Pahlm O, Arheden H: Rapid initial reduction of hyperenhanced myocardium after reprefused first myocardial infarction suggest recovery of the peri-infarction zone: One year follow-up by MRI. Circulation Cardiovascular Imaging. 2009, 48: 47-55. 10.1161/CIRCIMAGING.108.802199.View ArticleGoogle Scholar
- Heiberg E, Engblom H, Engvall J, Hedstrom E, Ugander M, Arheden H: Semi-automatic quantification of myocardial infarction from delayed contrast enhanced magnetic resonance imaging. Scand Cardiovasc J. 2005, 39: 267-275. 10.1080/14017430500340543.View ArticlePubMedGoogle Scholar
- Heiberg E, Ugander M, Engblom H, Götberg M, Olivecrona GK, Erlinge D, Arheden H: Automated quantification of myocardial infarction from MR images by accounting for partial volume effects: animal, phantom, and human study. Radiology. 2008, 246: 581-588. 10.1148/radiol.2461062164.View ArticlePubMedGoogle Scholar
- Heiberg E, Wigström L, Carlsson M, Bolger AF, Karlsson M: Time Resolved Three-dimensional Automated Segmentation of the Left Ventricle. IEEE Computers in Cardiology 2005; Lyon, France. 2005, 599-602. full_text.View ArticleGoogle Scholar
- Carlsson M, Osman NF, Ursell PC, Martin AJ, Saeed M: Quantitative MR measurements of regional and global left ventricular function and strain after intramyocardial transfer of VM202 into infarcted swine myocardium. American journal of physiology. 2008, 295: H522-532.PubMedPubMed CentralGoogle Scholar
- Carlsson M, Wilson M, Martin AJ, Saeed M: Myocardial microinfarction after coronary microembolization in swine: MR imaging characterization. Radiology. 2009, 250: 703-713. 10.1148/radiol.2503081000.View ArticlePubMedPubMed CentralGoogle Scholar
- Furtado AD, Carlsson M, Wintermark M, Ordovas K, Saeed M: Identification of residual ischemia, infarction, and microvascular impairment in revascularized myocardial infarction using 64-slice MDCT. Contrast media & molecular imaging. 2008, 3: 198-206.View ArticleGoogle Scholar
- Carlsson M, Ugander M, Heiberg E, Arheden H: The quantitative relationship between longitudinal and radial function in left, right, and total heart pumping in humans. American journal of physiology. 2007, 293: H636-644.PubMedGoogle Scholar
- Carlsson M, Ubachs JF, Hedstrom E, Heiberg E, Jovinge S, Arheden H: Myocardium at risk after acute infarction in humans on cardiac magnetic resonance: quantitative assessment during follow-up and validation with single-photon emission computed tomography. JACC Cardiovasc Imaging. 2009, 2: 569-576. 10.1016/j.jcmg.2008.11.018.View ArticlePubMedGoogle Scholar
- Engblom H, Hedstrom E, Heiberg E, Wagner GS, Pahlm O, Arheden H: Size and transmural extent of first-time reperfused myocardial infarction assessed by cardiac magnetic resonance can be estimated by 12-lead electrocardiogram. Am Heart J. 2005, 150: 920-View ArticlePubMedGoogle Scholar
- Carlsson M, Ursell PC, Saloner D, Saeed M: Multidetector computed tomography for characterization of calcium deposits in reperfused myocardial infarction. Acta Radiol. 2009, 50: 396-405. 10.1080/02841850902756540.View ArticlePubMedGoogle Scholar
- Carlsson M, Martin AJ, Ursell PC, Saloner D, Saeed M: Magnetic resonance imaging quantification of left ventricular dysfunction following coronary microembolization. Magn Reson Med. 2009, 61: 595-602. 10.1002/mrm.21869.View ArticlePubMedGoogle Scholar
- Hedstrom E, Engblom H, Frogner F, Astrom-Olsson K, Ohlin H, Jovinge S, Arheden H: Infarct evolution in man studied in patients with first-time coronary occlusion in comparison to different species - implications for assessment of myocardial salvage. J Cardiovasc Magn Reson. 2009, 11: 38-10.1186/1532-429X-11-38.View ArticlePubMedPubMed CentralGoogle Scholar
- Wu EX, Wu Y, Nicholls JM, Wang J, Liao S, Zhu S, Lau CP, Tse HF: MR diffusion tensor imaging study of postinfarct myocardium structural remodeling in a porcine model. Magn Reson Med. 2007, 58: 687-695. 10.1002/mrm.21350.View ArticlePubMedGoogle Scholar
- Gilson WD, Kraitchman DL: Cardiac magnetic resonance imaging in small rodents using clinical 1.5 T and 3.0 T scanners. Methods (San Diego, Calif). 2007, 43: 35-45.View ArticleGoogle Scholar
- Heiberg E: Automated Feature Detection in Multidimensional Images. Thesis. 2004, Linköping University, Department of Biomedical Engineering, [http://www.heiberg.se/einar/publications/pdf_open/phd.pdf]Google Scholar
- Chernobelsky A, Shubayev O, Comeau CR, Wolff SD: Baseline correction of phase contrast images improves quantification of blood flow in the great vessels. J Cardiovasc Magn Reson. 2007, 9: 681-685. 10.1080/10976640601187588.View ArticlePubMedGoogle Scholar
- Bernstein M, Zhou X, Polzin J, King K, Ganin A, Pelc N, Glover G: Concomitant gradient terms in phase contrast MR: analysis and correction. Magnetic Resonance in Medicine. 1998, 39: 300-308. 10.1002/mrm.1910390218.View ArticlePubMedGoogle Scholar
- Petzina R, Ugander M, Gustafsson L, Engblom H, Sjogren J, Hetzer R, Ingemansson R, Arheden H, Malmsjo M: Hemodynamic effects of vacuum-assisted closure therapy in cardiac surgery: assessment using magnetic resonance imaging. The Journal of thoracic and cardiovascular surgery. 2007, 133: 1154-1162. 10.1016/j.jtcvs.2007.01.011.View ArticlePubMedGoogle Scholar
- Arheden H, Saeed M, Higgins CB, Gao DW, Bremerich J, Wyttenbach R, Dae MW, Wendland MF: Measurement of the distribution volume of gadopentetate dimeglumine at echo-planar MR imaging to quantify myocardial infarction: comparison with 99mTc-DTPA autoradiography in rats. Radiology. 1999, 211: 698-708.View ArticlePubMedGoogle Scholar
- Engblom H, Carlsson MB, Hedstrom E, Heiberg E, Ugander M, Wagner GS, Arheden H: The endocardial extent of reperfused first-time myocardial infarction is more predictive of pathologic Q waves than is infarct transmurality: a magnetic resonance imaging study. Clin Physiol Funct Imaging. 2007, 27: 101-108. 10.1111/j.1475-097X.2007.00723.x.View ArticlePubMedGoogle Scholar
- Heiberg E, Engblom H, Ugander M, Arheden H: Automated Calculation of Infarct Transmurality. IEEE Computers in Cardiology; Durham, USA. 2007, 165-168.Google Scholar
- Nilsson B, Heyden A: A fast algorithm for level set-like active contours. Pattern Recognition Letters. 2003, 24: 1331-1337. 10.1016/S0167-8655(02)00374-4.View ArticleGoogle Scholar
- Svensson J, Gårdhagen R, Heiberg E, Ebbers T, Loyd D, Länne T, Karlsson M: Feasibility of Patient Specific Aortic Blood Flow CFD Simulation. MICCAI; Copenhagen. 2006, 77-86.Google Scholar
- Soneson H, Ubachs JF, Ugander M, Arheden H, Heiberg E: An Improved Method for Automatic Segmentation of the Left Ventricle in Myocardial Perfusion SPECT. J Nucl Med. 2009, 50: 205-213. 10.2967/jnumed.108.057323.View ArticlePubMedGoogle Scholar
- Ugander M, Soneson H, Heiberg E, Engblom H, Pals Jvd, Erlinge D, Arheden H: A novel method for quantifying myocardial perfusion SPECT defect size by co-registration and fusion with MRI - an experimental ex vivo imaging pig heart study. Scand Cardiovasc J, Malmö. 2008, 42 (Suppl): 47-Google Scholar
- Cerqueira MD, Weissman NJ, Dilsizian V, Jacobs AK, Kaul S, Laskey WK, Pennell DJ, Rumberger JA, Ryan T, Verani MS: Standardized myocardial segmentation and nomenclature for tomographic imaging of the heart: a statement for healthcare professionals from the Cardiac Imaging Committee of the Council on Clinical Cardiology of the American Heart Association. Circulation. 2002, 105: 539-542. 10.1161/hc0402.102975.View ArticlePubMedGoogle Scholar
- Cain PA, Ugander M, Palmer J, Carlsson M, Heiberg E, Arheden H: Quantitative polar representation of left ventricular myocardial perfusion, function and viability using SPECT and cardiac magnetic resonance: initial results. Clin Physiol Funct Imaging. 2005, 25: 215-222. 10.1111/j.1475-097X.2005.00618.x.View ArticlePubMedGoogle Scholar
- Gering DT, Nabavi A, Kikinis R, Hata N, O'Donnell LJ, Grimson WE, Jolesz FA, Black PM, Wells WM: An integrated visualization system for surgical planning and guidance using image fusion and an open MR. J Magn Reson Imaging. 2001, 13: 967-975. 10.1002/jmri.1139.View ArticlePubMedGoogle Scholar
- Lee JT, Munch KR, Carlis JV, Pardo JV: Internet image viewer (iiV). BMC medical imaging. 2008, 8: 10-10.1186/1471-2342-8-10.View ArticlePubMedPubMed CentralGoogle Scholar
- The pre-publication history for this paper can be accessed here:http://www.biomedcentral.com/1471-2342/10/1/prepub
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.