Artefact Correction within Digital Subtraction Angiography

In medical imaging, the process of allowing visualisation of blood vessels within a body is known as angiography. To allow this visualisation, a contrast medium is injected and spreads throughout the body. By using contrast, the increased visualisation heavily increases the diagnostic value of the image. It increases the contrast of strcutres or fluids. Due to bony structures and high amounts of soft tissue, there may be superimposition of the blood vessels (Meijering, Zuiderveld & Viergever, 1999). Due to this superimposition, it allows isolation of blood vessels from surrounding anatomy.

My specific issue of interest is as follows, the fluoroscopy technique of digital subtraction angiography (DSA), requires the patient to always  remain motionless throughout the entire procedure. This is a difficult practice and isn’t achieved often, usually pateint motion is unavoidable and can make the image un-diagnostic. If this happens, white streaking will appear over the imaging. When this occurs, it is known as motion artefact (Meijering, Niessen & Viegever, 1999). Motion arefact is common in all forms of medical imaging due to anatomical factors such as respiratory or cardiac motion.

The computer techniques to improve this specific practice is pixel shifting. Pixel shifting for gross translational motion can be used (Bentoutou, Taleb, Chikr El Mezouar, Taleb & Jetto, 2002). Motion artefacts in digital subtraction angiography can result in misdiagnosis or refusal of the imaging. This will require another examination and is to be avoided. Due to a repeat examination, another radiation dose will be required. Due to the nature of fluoroscopy, there is a production of instant and continuous imaging which gives off a large radiation dose to the patient. Due to this being performed ever more frequently as a popular treatment choice for a variety of neurovascular diseases, radiation dose in all procedures must be minmised, Okamato (2000).

Computing technique for digital subtraction angiography

When using DSA, a ‘mask image’ is required. A mask image is the first image from a angiographic sequence and is the area of interest just before the contrast is injected. The following exposures show the passage of the contrast medium throughout the body’s vessels. These images are known as a ‘live image’. The live images separately outlay a basic set of steps where corresponding pixels from the mask image are subtracted from each pixel within it (Okamoto, Ito, Sakai & Yoshimura, 2000). In the software we used previously, imageJ, this process can be done by opening files of a live image and a mask image. One of the options under ‘process’, is ‘image calculator’, there is an operation to ‘subtract’ one image from the other. Once these options have been selected, the result should be viewed in a new window.

Results and evidence of digital subtraction angiography techniques

Mask and live image source: Wood, D. T. (2016).

Figure 1. the above are results of ImageJ processing. From left to right; mask image, live image, resulting subtracted image, image after contrast adjustment

These are the results by employing computer technology. From what we can see in the third image, the subtraction process has removed the superimposing anatomy. This has also given a lowered contrast.

The reduced contrast is due to values of the pixels in the mask image corresponding to the pixels of the contrast filled vessel in the live image were subtracted from each other. This effectively lightening the appearance of the vessel.

Source: Wood, D. T. (2016).

Figure 2. From top left going clock wise is; mask image, live image, subtracted image and image after contrast enhancement

As we can see from figure 2, this method is very effective in visualising and locating structures with low subject contrast. In the original image, the vessel is very difficult to differentiate. After mask subtraction, the image effectively shows the structures that were initially superimposed. We can see demonstrated in the examples that there is increased noise output in the images post subtraction. This is due to the noise being randomly distributed throughout the mask and live images respectively, and therefore is passed on from both original images and into the secondary outputs images. Even though there is noise present, they can still be utilised and deemed diagnostic, ‘the small differences between the two images original images, pre and post contrast, are very noticeable and the small contrast-laden vessels are easily seen’ (Dougherty, G. (2009). Motion artefact is seen in both, in the form of white streaking. This is due to the live and maks images not exactly registering, pixel motion artefact correction can be applied in this situation.

Motion artefact correction computer technique

To reduce motion artefact, pixel shifting can also be used. It is similar to image subtraction in regards to its ease of use. Pixel shifting is a shift the mask image, in any plane, by each individual pixel. This is repeated until both the live and mask images match up accordingly. Once this occurs, subtraction will take place. Although pixel shifting is well utilised to correct motion artefact, one issue that has been indentified with this, is that pixel shifting will reduce the artefacts in some parts of the image, but “the remainder of the image artifacts will inevitably be reinforced or even newly created.

To correct for more complex patient motion, registration techniques should be designed so as to have more local control” Bentoutou,  Y., Taleb, N., Chikr El Mezouar, M., Taleb, M., & Jetto, L. (2002). Based on this, there has been development of ‘edge-based selection’ of control points. This is where the displacement of anatomy between the two image layers is computed and corrected by alignment of set points. Control points are selected at important edges in images by thresholding the gradient magnitude of the mask and live images otherwise known as edge detection (Bentoutou & Taleb, 2005). Edges can be detected and matched better than more similar regions of the images.

Results and evidence of motion artefacts correction techniques

The following images show artefact removal processes and results.

Source: Dougherty (2009)

Figure 3. Left is DSA image after pixel shifting and right is a DSA image with motion artefact

As seen in figure three above, pixel shifting is effective at minimising motion artefact. A gross translational shift was the cause of this motion pictured above.

Source: Meijering, Zuiderveld & Viergever (1999)

Figure 4. left to right; image after processing with control points, image with motion artefact and

image after pixel shift

Based off the pictures of figure 4, the edge selection and control point technique, is more effective than solely pixel shifting. The picture in the middle shows how the motion artefact has only been selectively removed and is still present in the left portion of the image. Where the control point process has been utilised, motion artefact has been removed and image quality

Conclusion

To conclude, in digital subtraction angiography, motion artefact is a very prominent feature during imaging. This common problem has therefore needed a process to overcome the diagnostic issues it causes. Through introduction of correction of motion artefacts, images overall quality has increased drastically and therefore will allow DSA to be a helpful diagnostic tool. Multiple techniques have been intorduced but pixel shifting has proved most effective, Meijering (1999). With most pateints presenting with movements of a complex nature, there will always be a need for more effective automated processes.

References

• Bentoutou, Y ., Taleb, N., Chikr El Mezouar, M.,  Taleb, M., & Jetto, L. (2002). An invariant approach for image registration in digital subtraction angiography. PatternRecognition, 35, 2853-­‐2865.
• Dougherty, G. (2009). Digital image processing for medical applications. Cambridge:Cambridge University Press.
• H. W. Meijering, Erik & Niessen, W.J. & Bakker, Jeannette & van der Molen, Aart & A. P. de Kort, Gerard & T. H. Lo, Rob & Mali, Willem & A. Viergever, Max. (2001). Reduction of Patient Motion Artifacts in Digital Subtraction Angiography: Evaluation of a Fast and Fully Automatic Technique1. Radiology. 219. 288-93. 10.1148/radiology.219.1.r01ap19288.
• Meijering, E., Niessen, W., & Viegever, M. (1999). Retrospective motion correction in digital subtraction angiography: a review. IEEE Transactions On Medical Imaging, 18(1)
• Okamoto, K., Ito, J., Sakai, K., & Yoshimura, S. (2000). The principle of digital subtraction angiography and radiological protection. Interventional neuroradiology : journal of peritherapeutic neuroradiology, surgical procedures and related neurosciences, 6 Suppl 1(Suppl 1), 25–31. doi:10.1177/15910199000060S102
• Wood, D. T. (2016). Fluoroscopy equipment, operation and digital subtraction angiography, Hull and East Yorkshire Hospitals:80.