ImSimQA™

Independent virtual phantom, testing and validation software for deformable image registration and beyond.

Overview

“So much more than QA of deformable image registration”

ImSimQA is the only dedicated and independent system for testing, validating and commissioning modern image registration and fusion software as used for adaptive re-planning, atlas-based automatic segmentation, IGRT and tumour response assessment.

ImSimQA provides a wide range of functionality for testing across a range of modules, using a virtual phantom library and extensive DICOM editing tools, for example, 4D planning / respiratory gating, multi-modality imaging, and general geometric QA tests of imaging devices used in planning and onboard treatment machines.

Independent, TG 132 ready, as recommended by AAPM Task Group report

Deformable image registration, DIR is now embedded in many clinical applications for adaptive radiotherapy, multi-modality planning, automatic contouring and more. ImSimQA specialises in the QA and validation of DIR algorithms used in a wide range of clinical systems, providing all the testing recommended by the TG-132 report.

ImSimQA image datasets as provided by the Task Group 132 report, are starting sample images only, to implement QA and validation of deformable image registration and registration and fusion generally. Meaningful clinical validation, however, requires ImSimQA to be installed, so that the full range of virtual phantoms, DICOM editing tools and quantitative pass/fail analysis tools can be used.

Quantitative analysis

ImSimQA contains all the quantitative analysis tools recommended by TG 132 and provides an extensive, fully customisable library of digital, or virtual phantoms both geometric and anthropomorphic. Additionally, any patient or phantom image can be imported as DICOM and then edited to re-create the precise clinical scenario which will differ across a range of clinical techniques, imaging devices, and from clinic to clinic.

ImSimQA Benefits

“Test without the hard phantom, test without the scanner”

Extensively tests and validates modern radiotherapy systems in a way not possible with hard phantoms
Increases accuracy of testing
Provides a cost-effective toolkit of multiple phantoms
Offers flexible and infinite scope for test scenarios
Reduces the time needed to acquire test images on busy CT/MR/PET scanners, 4D and IGRT imaging systems

ImSimQA™ Base Module

This the underlying software to which optional modules can be added. This module includes a virtual phantom library, manipulation of DICOM-3 images and the creation of new DICOM images for testing clinical applications.

Virtual phantom library

The virtual phantom library contains a library of 25 3D anatomical and geometric virtual phantoms, all of which can be extensively edited and exported as DICOM-3 images, to create uniquely customisable phantoms which offer an infinite range of testing potential. All virtual phantoms can be transformed, deformed, re-orientated, structure density changed etc, to create DICOM-3 series in CT, simulated MR and PET which are then exported to the test application. DICOM headers can be edited. Parameters such as slice spacing and FOV can be modified in the exported DICOM-3 series.

Manipulation of DICOM-3 images

DICOM-3 images of any modality can be imported, edited and exported as new DICOM series. Supported modalities include CT, MR, PET, MVCT and cone beam CT. RT Structure Sets can be imported, and edited prior to export. A selection of noise patterns can be added to images.
DICOM images can be globally translated, rotated, and re-scaled under controlled conditions to assist with systematic validation of clinical systems.
More extensive deformations can be applied using the ImSimQAdform module.

Testing image registration algorithms without the scanner

ImSimQA is used to test rigid and deformable image registration algorithms found in today’s image fusion packages, IGRT treatment machines, auto-contouring and dose mapping in treatment planning applications.

For image registration testing, varying setups of an initial DICOM image series can be created, without having to use valuable scanner time to create multiple scans for testing. Initial high-resolution images e.g. 1mm slice spacing can be resampled to 2,3,5mm etc and exported to test the effect of slice resolution. RT Structure outlines can be edited, and structure densities can be allocated a new density prior to export – to test the effect of density differences on image registration accuracy.

Editing Real DICOM Images

Editing DICOM images imported into ImSimQA is a powerful way of creating an infinite range of new test scenarios, without needing to re-scan or source patient images, yet apply systematic changes in a scientific and methodical fashion.

Any DICOM-3 image modalities such as CT, MR, PET, and radiation therapy images such as MVCT and CBCT, can be imported and extensively edited.

DICOM parameter editing

DICOM header and image acquisition parameters such as name, UID, slice and FOV resolution can be edited and new parameters saved for export.

Density changes and RT Structures

An imported CT series with associated RT Structures can easily be edited to change the density of a contoured structure such as the bladder. A user-defined Hounsfield value can be assigned and the newly created images exported for testing. An example of this would be to apply a density for contrast media to the bladder in an IGRT-CBCT or planning CT image series, to test how image registration and fusion software performs with varying densities.

Virtual Phantom Library & Editing

ImSimQA provides a toolkit of 15 virtual phantoms that can be extensively edited and transformed, before being converted to DICOM CT, MR & PET simulated images. With the ability to add noise, change density, change slice spacing and re-orientate, the phantom DICOM images are then exported to the test application, minimising the use of the real scanner, and increasing the efficiency of testing. Real DICOM images can also be imported into ImSimQA for editing, without having to continually re-scan.

ImSimQA^dform Module

ImSimQA^dform allows users to implement the requirements of the AAPM Task Group 132 report on the ‘Image Registration and Fusion Algorithms and Techniques in Radiotherapy’. The user can create deformed changes to virtual or real image structures. Deformations can be local or global and applied to any user-defined anatomy or object. Once deformations have been applied, static or 4D image series of CT/MR/PET can be created and exported to a 3rd party system for testing of, for example, image fusion software. Examples could include CT-CBCT IGRT image matching assessing daily anatomy changes; CT-MR image fusion of pre and post-operative images; image fusion of MR images on curved couches – CT image on flat couch tops.

ImSimQA^dform users also have access to an additional library of 10 original and deformed synthetic head and neck phantoms created where the deformed dataset is based on actual anatomical changes seen during the radiotherapy treatment course. From these, a ground truth deformable vector field (DVF) was created for each phantom. This allows the phantoms to be used to assess the accuracy and precision of the DIR algorithm in any available system, with clinically relevant data. This was developed through work with clinical partners.

As well as creating ground truth deformed datasets for testing your clinical system. ImSimQA can also be used to compare the ground truth deformation with the clinical dataset.

Comparisons can be made in 4 ways:

Direct comparison of CT images visually and quantitatively using Mutual Information and correlation coefficient.
Contour Analysis Comparison
Comparison of the ground truth and clinical deformation vector fields (DVFs).
Jacobian matrix display (pictured below).

Deforming Images & Structures

Access to scanners for obtaining test images, either of phantoms or patients, can be challenging due to high clinical demand. ImSimQA can import DICOM-3 image series of any imaging modality, plus import DICOM RT Structure sets associate typically with CT data.

Once an initial set of images are imported, they can be extensively edited e.g. slice spacing, FOV changed but can be manipulated and deformed to produce newly created image series and RT structures that can be exported to a test application. A typical example would be to deform an MR image series and export it along with a reference CT series to image registration and fusion application, commonly used in radiation therapy and medical imaging. The test application would then perform its image registration, acting typically to transform (for rigid image registration) or deform (for deformable image registration) the MR. The capability within ImSimQA to generate an infinite range of new test images will provide the physicist or clinician, with a powerful tool to test the performance and limitations of the image registration software in clinical use.

Oncology Systems Limited's ImSimQA can further re-import the generated images from an image registration or auto-contouring application, and perform quantitative testing on the resulting images, to report and analyse the goodness of matching, and visually highlight areas of mismatch.

Deforming and Transforming Images

Virtual markers can be inserted into any modality image, mapped to a new location, and a local or global deformation applied. New realistic clinical scenarios can be created to produce new images that can be exported to test rigid and deformable image registration, auto-contouring, auto-margining algorithms plus a wide range of medical and oncology imaging systems that reconstruct images or handle image data generated at different points in time when a patient’s anatomy would have changed.

ImSimQA^4d Module

Create 4D image series for testing 4D RTP and gating systems.

ImSimQA^4d creates 4D image series of phantoms quickly and effortlessly, using the 4D Phase Wizard. Create multiple phases of whole phantoms or individual phantom structures. Export newly created DICOM image series to test 4D simulation, planning and gating systems.

Imported RPM - phantom plus CT

Step 1: Select ImSimQA thorax virtual phantom

Step 2: Apply motion from imported RPM or manually select motion patterns

Step 3: Export ImSimQA created images as 4D CT DICOM-3 series to test application

ImSimQA^contour Module

ImSimQA^contour compares and analyses pairs of imported contours from radiation therapy applications.

Analyse application performance, compare users

This module has been designed to analyse the performance of automatic contouring/segmentation algorithms commonly found in radiation therapy applications, and also to compare the performance of that personnel involved in the contouring process, such as radiation oncologists and dosimetrists.

Qualitative visualisation and quantifiable metrics are reported and presented for fast and efficient analysis.

Typically, this module is suited to testing atlas-based automatic segmentation performance, general RTP automatic contouring software, to compare operator contouring performance against a gold standard and inter-operator variability. A range of metrics are included such as Conformity Index, Mean Distance to Conformity and Centre of Gravity, all of which provide results of precisely where and by how much contours vary.

UF Health Cancer Center at Orlando Health

ImSimQA is a very useful tool for the commissioning and QA of deformable image registration algorithms. As DIR algorithms become more commonplace in clinics, it is imperative that clinicians understand the mechanisms and limitations of these algorithms. ImSimQA allows users to create their own deformations where the underlying deformation map is user-defined and known for every point in the image set. This allows physicists to perform more extensive testing than landmark analysis would permit, especially in homogenous regions where the anatomy may be difficult to differentiate (e.g. parotid glands). Additionally, ImSimQA overcomes the limitations of testing with physical phantoms by allowing the user to work with actual patient data.

Our research team used ImSimQA in conjunction with an in-house algorithm to create a library of ten virtual head and neck phantoms for testing the performance of multiple commercial DIR algorithms. These tools allowed for the modeling of head rotation and translation, mandible rotation, spine flexion, shoulder movement, hyoid movement, tumor/node shrinkage, weight loss, and parotid shrinkage in a clinically-relevant fashion.

Jason Pukala, PhD, Physicist
UF Health Cancer Center at Orlando Health

Minneapolis Radiation Oncology

Image Guided Radiation Therapy (IGRT) using daily kilovoltage Cone beam CT (kvCBCT) or Megavoltage CT (MVCT) imaging is emerging as the standard of care in Radiation Oncology. The technology, workflow, and clinical implementation have been on the fast track for the past year at most centers in USA.
One of the key components in IGRT workflow is the direct imaging of soft tissue and comparison with the treatment planning CT image to derive the shift of the day. This is being routinely done in most clinical institutions using a vendor supplied rigid registration auto-fusion algorithm. However, very little quality assurance has been carried out to validate this process and its accuracy in clinical work flow.
I have been using IMSIMQA in both clinical and research environments. It provides excellent tools to explore the validity of registration algorithms used in IGRT work flow. Specifically, it eliminates the need to use real phantoms since actual patient images and a library of supplied phantoms can be imported and used to test the image fusion registration accuracy.

I have been particularly impressed with the IMSIMQA software’s potential to evaluate the validity and accuracy of deformable registration algorithms. Deformable registration is the key technology or process to implement adaptive radiotherapy and 4D planning because one is interested in the total dose delivered over several fractions that incorporate different patient images (adaptive) or different time points (4D). There have been numerous publications in the literature describing various algorithms on deformable registration but very few papers have addressed the validity and accuracy of the algorithms to verify that the deformations produced are indeed correct. The problem arises from the absence of a gold standard on how this can be implemented. It is clear that building a deformable phantom is an enormous challenge for most clinical departments.

IMSIMQA software has tremendous potential for evaluating the deformable registration algorithm used against a known applied deformation. The software has tools to allow the user to apply known deformations on a patient DICOM study set thereby replicating a clinical scenario and then applying the deformable registration algorithm to evaluate what you recover after the deformation.
We are currently working towards the development of a framework within which IMSIMQA tools can be used to validate deformable registration algorithms. We believe IMSIMQA is a powerful software with applications in both clinical and research settings.

Raj Varadhan M.S.,M.Phil,D.A.B.M.P. , Senior Medical Physicist
Minneapolis Radiation Oncology

Northampton General Hospital

ImSimQA has been in use at Northampton General Hospital for several years. It has been an incredibly useful tool with many applications.

We originally procured it because we had acquired new Virtual Simulation software, and needed a way of commissioning and comparing the two image fusion packages we now had available. As with most departments, there is a distinct lack of time available on MR / PET scanner; and ImSimQA seemed an inspired idea! We were able to assess the advantages and limitations of each system in a controlled and methodical way; for each algorithm provided by the two systems. Following software upgrades we were able to easily re-test the planning systems.

Additional uses have been in training staff in the image fusion packages; and setting new scanning protocols in the diagnostic department to improve the chances of a successful fusion of images.
We have also been able to use it for other departmental development. For example, the advancement of breast planning, by showing the effect on a treatment plan of putting patients on lower wedge angles on breast boards.

ImSimQA is such an exceptionally valuable tool. We could not have implemented image fusion at NGH without it; and there have also been unforeseen advantages as shown above. We look forward to using it to commission respiratory gating, CT-PET fusion and other new developments in the future – all of which will be a lot less time consuming via ImSimQA.

Nicky Whilde, Head of Radiotherapy Physics
Northampton General Hospital