Assessment and characterization of endothelial function in the diagnosis of cardiovascular diseases is a current clinical research topic [1,2]. The endothelium shows measurable responses to flow changes [3,4], and flow-mediated dilation (FMD) may therefore be used for assessing endothelial health; B-mode ultrasonography (US) is a cheap and noninvasive way to estimate this dilation response [5]. However, complementary computerized image analysis techniques are still very desirable to give accuracy and objectivity to the measurements [1].

Several methods based on the detection of edges of the arterial wall have been proposed over the last 10 years. The first studies used a tedious manual procedure [5], which had a high interobserver variability [6]. Some interactive methods tried to reduce this variability by attracting manually drawn contours to image features, like the maximum image gradient, where the vessel wall is assumed to be located [7-9]. Some more recent efforts are focused on dynamic programming or deformable models [10-19] and on neural networks [20].

1 Computer Vision Lab, Aragon Institute of Engineering Research, University of Zaragoza, Zaragoza, Spain

2 Department of Technology, Pompeu Fabra University, Barcelona, Spain

3 Lozano Blesa University Clinical Hospital, Aragon Institute of Health Sciences, Zaragoza, Spain

All these methods present some common limitations. First, edge detection techniques are undermined by important error sources like speckle noise or the varying image quality typical of US sequences. Second, most methods require expert intervention to manually guide or correct the measurements thus being prone to introduce operator-dependent variability. Also, almost no method performs motion compensation to correct for patient and probe position changes. This could easily lead to measuring arterial dilation using wrong anatomical correspondences. Temporal continuity is another aspect that has not been exploited enough in previous work. Two consecutive frames have a high correlation, and only Newey and Nassiri [20] and Fan et al. [15] take advantage of this feature during edge detection. Finally, there is a general lack of large-scale validation studies in most of the techniques presented so far.

In this chapter a method is proposed that is based on a global strategy to quantify flow-mediated vasodilation. We model interframe arterial vasodilation as a superposition of a rigid motion (translation and rotation) and a scaling factor normal to the artery. Rigid motion can be interpreted as a global compensation for patient and probe movements. The scaling factor explains arterial vasodi-lation. The US sequence is analyzed in two phases using image registration to recover both rigid motion and vasodilation. Image registration uses normalized mutual information [21] and a multiresolution framework [22]. Temporal continuity of registration parameters along the sequence is enforced with a recursive filter. Application of constraints on the vasodilation dynamics is a natural step, since the dilation process is known to be a gradual and continuous physiological phenomenon.

Once a vasodilation curve is obtained, clinical measurementes must be extracted from it. Classically, FMD is quantifiedby measuring the peakvasodilation diameter relative to the basal diameter level, which are usually manually identified in the curve. Automatically extracting these two parameters is not trivial (e.g., a mere mean of several basal frames and a simple search for a maximum in the curve) given that the curve may also include artifacts. We examined the use of a robust principal component analisys to derive intuitive morphological parameters from the curves and relate them to classical FMD indexes and cardiovascular (CVD) risk factors.

The chapter is organized as follows: Section 5.2 describes the system for image acquisition and the protocol for a typical FMD study. It also describes the population used to evaluate our technique. Section 5.3 introduces the proposed method to assess FMD. The validation of the technique is reported in section 5.4. In section 5.5 a novel parameterization of the vasodilation curve is introduced, and correlation analyses are presented that relate these new parameters to CVD risk factors and classical FMD parameters. In section 5.6 the results are discussed and some concluding remarks are made in section 5.7.

Was this article helpful?

0 0

Post a comment