BrainMagix.dev

CT/PET/MRI and more with layer coregistration

2010-05-02T19:07:00.002+02:00

Radiology has several tomographic imaging techniques offering the insight into a brain. The most popular are Computed Tomography (CT), Positron Emission Tomography (PET), and Magnetic Resonance Imaging (MRI). They all are non-invasive, deliver great images, for radiological diagnostics, however due to the different phenomena they rely on, have different constrains and are more predestined to imaging particular aspects of human anatomy.

The CT, based on X-ray attenuation, delivers very clear images of bones and other dense tissues, and by using contrast windows allows to visualize brain tissues. PET, which is based a radioactive isotope decay, injected prior to the examination, is highly sensitive to brain physiological processes and functions, however lacks the information about the anatomy. MRI on the other hand provides beautiful images of hydrogen rich tissues (like brain - 78% H20), but is less sensitive for imaging of the "dry" skeleton structure, it may also be used for functional imaging.

It is becoming a common practice, that in order to get a full assessment for the diagnosis. different imaging modalities are used, but... 3 different machines produce 3 different image results! When looking at them individually it may be very difficult to grasp the common context.

This is why with BrainMagix, we are trying to let you see all available information simultaneously, as a cross-modality image fusion. You can freely navigate through the entire brain volume, change each modality layer properties, like transparency, or colour or contrast, to have a clear idea about the relationships between the bone, brain anatomy, and the function.

Currently we are working on automated co-registration algorithms that will allow you to easily merge different exams of different modalities, or acquired during different sessions (pre-op, post-op). But already made a simple tool available to perform the image co-registration by hand. You can control the geometry of each volume layer. While modifying parameters, the viewer is continuously updated, so you can get quite accurate.

The parameters you can modify respectively for x,y, and z directions are:

translation - the position of the volume in 3D space
rotation - the volume angle
scale - is the pixel spacing (you may find it useful when images are deformed, for example a common issue in EPI sequence)
shear - the volume screw

On the following screen shot, you can see a case study, where PET/CT examination was co-registered with MRI.

The next step would be adding fMRI, DTI and PWI layers, each revealing another dimension of the patient's brain.

Colour menu

2010-04-29T20:30:00.007+02:00

The viewer context menu gained a new member - the colour menu. It allows you to quickly change the colour map of the currently selected layer. We also added new colour maps which may help you create Andy Warhol's brain impressions (as seen below), or just give you a better insight into image contents.

We found it particularly useful when several fMRI exams are processed at once. Now you can assign a different colour to each, and make overlays easier to distinguish visually.

Another useful feature we added is the ability to control the background transparency. When applied, the background colour (most of the times black) will be erased from the layer to reveal the contents below. You can use it in conjunction with the brightness and contrast tool , to threshold the background.

fMRI Motion Correction Reports

2010-04-21T01:08:00.006+02:00

Motion correction, or volume realignment is an important step of the fMRI analysis. The algorithm calculating Statistical Parametric Maps (SPM) is using a single voxel approach, which means that it observes signal changes of an individual volume point in time. This of course requires every point of image be a representation of the same point in space, or brain to be more precise. The assumption is not easy to assert due to the inevitable patient's head movement during the long period of fMRI image acquisition. These small movements incorporate into the signal variance and generate artifacts (signal changes are not the effect of the blood oxygenation levels, but voxel's spacial displacement).

BrainMagix tires to compensate this effect by adjusting position of every volume with reference to the first one acquired. For every volume a 3D transformation is calculated that minimizes the error between them. The transformation has six parameters: translation in x, y, z direction and rotation in x, y, z direction. The result of this procedure is very important to asses the quality of the fMRI examination. While relatively small errors (translation ~1-2 mm and rotation ~1 deg) may be corrected (and neglected), larger head movements will most likely influence the SPM results - especially if the motion artifacts follow the experiment's task paradigm.

Recently we were working on improving the usability of the fMRI analysis report, especially for motion correction. Now you can view the calculated translation and rotation correction factor graphs for every fMRI examination individually, by navigating the tree on the left side of the results dialog window. You also may play a movie (cine) combined of images after the motion correction procedure, to visually confirm the data quality. This may help you in discarding tests which were not performed well by the patient.

We also added a toolbar icon that displays the results report window.

Introduction to Layers

2010-04-12T20:36:00.007+02:00

In BrainMagix you can display several examinations, or image volumes simultaneously. In order to use this feature it is necessary to understand the concept of layers.

You can think of layers as a stack of glass tiles, on which a particular brain image is painted. The stack is organized in such way so that the result image you see (the top image of the stack) is the sum of everything which is underneath.

You have several tools to control how layers are being displayed by BrainMagix:

You can add new layers using which opens the import wizard
You can change the ordering of the stack by moving layers up and down, using icons , or by drag-and-dropping a layer on a desired position
You can simply turn the visibility on and off for an individual layer with
You can change the transparency of each single layer, to control how well it covers layers below using the alpha tool
You can modify the brightness and contrast of each layer, for example to make layers which are below brighter and more visible using
If you no longer need a layer you can remove it from the stack with

You can use these features in layer supported views: orthogonal and MPR.

BrainMagix and Medtronic StealthStation integration

2010-04-12T13:44:00.000+02:00

Medtronic is one of the bigger manufacturers of the treatment guidance solutions. Recently we made an effort to integrate our software with their neuro-navigation system called StealthStation. While we are still in discussion about providing integration using proper DICOM data exchange (unfortunately we are facing some restrictions), we already made the integration possible by taking the advantage of the Analyze format which is supported by Medtronic (please contact them or us to obtain the document with details how to import images to StealthStation).

The new image format is available under a name "Medtronic Analyze". Note that you should not confuse it with the Analyze 7.5 format already available for some time in BrainMagix . The main differences between the existing and the new Medtronic Analyze, are:

Medtronic Analyze does not save additional information about volume orientation in a *.mat file (note that files exported this way loose a lot of important data tags)
Medtronic Analyze saves image data in a radiological left/right convention (the same as DICOM)

To export images in Medtronic format use the convenience of Export Profiles. Below is a screenshot of a sample profile's configuration.

Note that you need to export your data sets without colour.

Next you need to make sure that the BrainMagix's exported exam exactly matches the anatomical exam used by the guidance system with respect to resolution, orientation, slice thickness, and number of slices. You can easily assert that by just setting the required anatomical volume as the reference layer (single click), as explained here.

Now you are ready to go.

One advice is that when you transfer data on a CD to the navigation you need to make sure that you burn one Analyze exam (one set of *.hdr and *.img files) per one disk, otherwise StealthStation will not be able to read your data. Also you should use a burning software that enables data verification after recording, and is capable of writing disks in a "joliet CD format" (important!).

Finally you need to remember that even though we test our software to provide the best quality in terms of the accuracy of the spatial registration between the BrainMagix processed images and anatomical MRI, you [the User] take full responsibility for ensuring their validity and interpretation!

New Export

2010-04-09T18:09:00.007+02:00

Working with medical images is often not obvious. Take MRI data - depending on the imaging sequence parameters acquired images may have different spacial resolution (reconstruction matrix size and voxel dimensions), different orientation (i.e. transversal, coronal, sagittal), different field of view, etc. Proving the image viewing software is capable of image reconstruction using this specific information, they may be difficult to compare, when viewed side by side.

BrainMagix can provide image visualisation in a common coordinate system (see Reference Volume). However, in order to be compatible with other applications (for example neuro-navigation software) we needed to provide an export feature, that would produce images which may be opened correctly even with the simplest viewer and still be directly comparable (same size, orientation, etc.)

The way to do it is by:

reslicing - reconstructing volume slices in a modified geometry (angle, z-position)
resampling - reconstructing volume slices using another reconstruction matrix (interpolated)

with the same parameters as in the reference volume.

We figure the ability to export processed data in a easy and robust way should be the key feature of BrainMagix.

Let us see an example, where we have a 3D T1 volume (512x512x120), a high resolution image, and a series of temporal EPI volumes (96x96x40) that will be used to calculate Statistical Parametric Maps (SPM). T1 will be directly used by the neuro-navigation station. We will also use it as a reference volume in BrainMaigx to fuse with low resolution EPI.

By changing the visibility of volume layers (eye icon in the layer manager), you can directly export the re-sliced and re-sampled results from the MPR or orthogonal view, here:

3D T1 - the same as original (may used as a fail-safe, to check that no artefacts were generated while processing the original volume)
SPM on 3D T1 - the high resolution anatomical data with SPM overlays (may be used for surgical planning)
SPM - just the SPM layer (may be directly fused with T1 using another software without the need of additional co-registration or re-formatting)
EPI - a volume interpolated from the low resolution EPI (may be used to check the co-registration between anatomical and functional images)

Just keep in mind which is your Reference Volume!

We hope you will find more use cases of the new, improved export engine.

Behind the Scenes of the Image Import Hell

2010-04-07T21:08:00.012+02:00

While developing BrianMagix, one the first problems we encountered was image import. Even thought it may seam that

the Digital Imaging and Communications in Medicine (DICOM) is a standard for distributing and viewing any kind of medical image regardless of the origin ^*

should by definition provide unified way for displaying images, we quickly realized how bold a statement it was.

Let us say that BrainMagix differs from a regular DICOM viewer, a lot. Its objective is not to display DICOM images per se, but process them. Majority of algorithms in our software use single voxel processing, and operate on the series of temporal volumes (stacks of slice images captured at the particular moment of time) acquired with MRI. It means that in order to work, we need to have our data organized as volumes as well. Among other, we need to have an information about volumes' geometry, location in space, and location on the time axis. These requirements were not so easy to fulfill when we discovered the huge variability in the way different scanner manufacturers export their data.

Since our goal was to become "manufacturer independent" image processing workstation, we had to deal with all cases where

every slice of the volume is represented as an single image file
every volume is represented as a mosaic in a single image file
every series of volumes is represented as a big multi-frame file
slices in series are sorted based on their z-location
slices in series are sorted based on their position in volume
slices in series are not sorted at all
or maybe it's a DICOMDIR file, which just points to referenced images
...
or completely different format that needs another processing strategy

and many more (volume orientation, slice thickness, ... the list continues).

After many many tests and sample data sets, coming from dozens of different scanners we think we nailed all cases, so that you, the BrainMagix user, may just point, click next, and not bother such irrelevant issue as DICOM. Behind the scenes BrainMagix crawls though files and directories, quietly yet scrutinizingly, decomposing, sorting, re-organizing data to undo this mess, as the progress bar approaches 100%.

Also you don't have to bother that now this process is using parallel threads to squeeze the most of your n-core processor's performance ;)

next

Export Profiles

2010-04-07T18:12:00.009+02:00

Most of the times image processing workstation (i.e. BrainMagix) is a part of a hospital network. Image sources like scanners are connected to the Picture Archiving and Communication System (PACS) or other systems that are capable of receiving DICOM images (so called DICOM nodes). BrainMagix provides network connectivity and may be used as a DICOM server, to which images from MR scanner may be sent directly for post processing.

When the hard, image processing, work is done, and image maps are generated in normal scenario they are being archived. This time the connectivity in inverse direction is necessary. BrainMagix becomes an image provider for other DICOM nodes - for example PACS or neuro-navigation station.

Normally the hospital network has are one or more locations to which images are being archived or sent to undergo further processing. This is why we we created Export Profiles, which make the life a little more easier. Once defined in Settings, they may be quickly accessed in the export wizard.

Export profile may be defined for a network location (DICOM node) or a local database on the hard disk. In the second case you can specify the file format (Analyze, Nifti, DICOM, PNG, ... ). You can also define a file naming convention using image tags which will be substituted upon actual image writing, for example:

{PersonName}-{PersonID}-{StudyName}-{Num}

Values in brackets will be replaced with person name, id, study name and image number.

Another feature that we added is a possibility to export images with or without colour, as some 3rd party software only support grey scale images.

Fusion

2010-04-07T03:15:00.009+02:00

The ultimate goal of BrainMagix is to combine every possible information from every possible source, and present it in a clear form. Here is a sneak peek of one of the interesting features we are working on: mult-modal fusion. We combined an anatomical image with processed DTI and fMRI, to get this interesting result.

Reference Volume

2010-04-07T01:24:00.009+02:00

Introducing the Reference Volume! This neat mechanism allows you to choose a layer (volume) which features (geometry, position, etc.) are to be considered as a base.

Using BrainMagix stack viewers (orthogonal, mpr) you can easily display one layer on top of the other, and if the patient did not move a lot during the scanning session they more or less overlay each other. The reason why these images are correctly displayed is that we take into account the position of each layer in 3D space as described by DICOM tags (i.e. Image Position Patient and Image Orientation Patient). However this information is in reference to an arbitrary coordinate system of the scanner. It means the reconstructed image (viewed in the orthogonal planes) would be showing the actual patient's head position inside the gentry. This is not very convenient, as in most cases of radiological interpretation a well defined orientation (for example AC-PC) is desired. With the Reference Volume you can now quickly choose the layer of the reference geometry. BrainMagix will calculate the appropriate rotation/translation transformations and apply it to all layers in order to align them to the selected one.

To illustrate let us assume that during a single scanning session two image volumes were acquired:

1. Sagittal
2. Transversal in AC-PC orientation

The image shows the orthogonal view of the described situation. The reference layer is now the T1 Sagittal view (marked with orange icon). This orientation may be not useful for finding landmarks. However now you can quickly restore it to AC-PC by just clicking the reference layer icon of the T2 layer.

Layers are instantly aligned to the reference AC-PC orientation of the T2 volume.

Easy?

Welcome

2010-03-27T13:01:00.002+01:00

Welcome to BrainMagix developers' blog.

In this blog we will be presenting you new features of our software, write about what keeps us busy, share some useful tips and use cases.