Research topic (Back)

X-Ray imaging

Pathologies in the vascular system (angiopathies), and in particular those concerning the heart (cardiopathies) are the first death cause in Europe. They are mostly due to stenosis formations: lipids deposits reduce or occlude the vessels. The consequences can be very serious especially when they impact arteries that are as crucial to the organisms life as coronaries (which provide the heart in oxygen).
X-Ray imaging offers a minimally invasive mean to diagnostic and to treat such malformations. The clinician introduces through the patient's groin the interventional tools he needs, and handles them remotely, guiding his gesture based on the images acquired with X-Rays.

General Electric Innova imaging system

X-Ray image formation

Difficult interventional images

Diagnostic exams allow the clinician to visualize the patient's vascular tree by contrast injection. To have a precise idea of the 3D organisation of the vascular system, he images it in different angulations. This exam being quite short, high radiations are allowed, leading to contrasted images with little noise.

On the other hand, Interventional exams can last much longer. The clinician must set his tools correctly at the stenosis' location and perform his intervention. He can for instance push the plaque by inflating a balloon, and possibly lay out a stent (spring whose surface has been chimically treated).
Such an operation can last long, especially when complications occur, and can imply dozens of minutes of exposition. As a result, the X-Ray radiation has to be limited to protect patients and medical staff. But this implies the formation of difficult images with high quantic noise levels. To make the images more readable, and to ease the task of working with them daily, they must be digitally processed.

Typical image for diagnostic exams

Typical image for difficult fluoroscopic exams

An efficient way of addressing the noise issue is to average successive measures of the same pixel. Such a temporal denoising sheme has the advantage over spatial filters not to introduce certain artifacts.
Spatial filters namely tend to colour the noise (it is gathered in beans). This artifact, hardly noticeable on static images, can be extremely disturbing at video frame rates, especially when the noise level is high. Moreover, since the phenomenon is linked to the psychovisual perception of a noise as being disturbing, it is difficult to formalize and to handle. And finally, spatial filters also change the image look, which can disconcert the users.

The main drawback of the temporal denoising filter is its unability to address scene motions: an adaptative temporal filter would decide at each pixel which averaging strength offers the best compromise between noise reduction and information conservation. This discussion leads to conservative parameter tuning which limits greatly the denoising power of the filter.

My PhD offers to explore ways out of this alternative by developing an algorithm which would be robust to the motions. (Back)

Transparent images

Layer model

Transparent fluoroscopic image

Because of the physics of the X-Rays, the clinical images are transparent. For instance, we can see on the example above the heart superimposed to the diaphragm, the lungs, the spine, the ribs and the interventional devices. (More)

Thus, original methods, tackling specifically the transparency issue, must be developped to answer the two following questions:

How can we estimate the motions present in the image?

How can we use that information in an efficient temporal denoising?

Our contributions on these two topics are briefly introduced in the next page.

Back