Skip Navigation

European Journal of Echocardiography 2004 5(5):335-346; doi:10.1016/j.euje.2003.12.003
© 2004 by European Society of Cardiology
This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrowRequest Permissions
Right arrow Disclaimer
Google Scholar
Right arrow Articles by Frouin, F.
Right arrow Articles by Diebold, B.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Frouin, F.
Right arrow Articles by Diebold, B.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?

Copyright © 2004, The European Society of Cardiology

Factor analysis of the left ventricle by echocardiography (FALVE): a new tool for detecting regional wall motion abnormalities

Frédérique Frouina, Annie Delouchea, Hanna Raffoulb, Hervé Dieboldb, Eric Abergelb and Benoît Diebolda,b,*

aINSERM U494, CHU Pitié-Salpêtrière, Paris, France
bService d'Echocardiographie, Hôpital Européen Georges Pompidou, Paris, France

Received 31 July 2003; received in revised form 17 December 2003; accepted after revision 22 December 2003.

* Corresponding author. Service d'Echocardiographie, Hôpital Européen Georges Pompidou, 20 Rue Leblanc, 75015 Paris, France. Tel.: +33-1-56-09-38-04; fax: +33-1-56-09-38-10. benoit.diebold{at}egp.ap-hop-paris.fr


   Abstract
Top
 Abstract
Introduction
 Principles
 Material and methods
 Results
 Discussion
 Conclusion
 References
 
Background: Factor analysis of the left ventricle in echography was developed to study the regional wall motion. Two factors and associated factor images were estimated using specific constraints: one "constant" factor and another "contraction–relaxation" factor. The constant factor was encoded in green, the positive component of the contraction in red and the negative in blue.

Methods: The evaluation was carried out on 12 patients with LBBB or pacemaker (group A), and on 26 others (group B). The segments were graded separately on the cine-loops by three experienced echocardiographers. Similarly, the three-color combination of the factor images was read at the endocardial border and each segment was scored.

Results: An absolute concordance was obtained for 64.8% of the segments and a relative concordance (within one grade) for 97.2%. They were 71% and 99.6% in group B. Most of the discordant cases were explained by the global motion during the cardiac cycle. The standard deviation of the difference between the mean wall motion scores was 0.38 for all the patients; it was reduced to 0.30 in group B.

Conclusion: Factor analysis is a promising tool to study the regional wall motion. It might become useful for assessing segmental wall motion in 2D and 3D echo.

Keywords: Segmental wall motion; Echocardiography; Image sequence; Factor analysis


   Introduction
Top
 Abstract
 Introduction
Principles
 Material and methods
 Results
 Discussion
 Conclusion
 References
 
The visual assessment of the segmental wall motion of the left ventricle is being used since decades in patients with suspected or demonstrated coronary artery disease (CAD).1 It has been proposed with the number of segments varying from 20 (Guyer et al.2) to nine (Heger et al.3). The segmental analysis relies both on wall displacement and wall thickening.4,5 Although widely used, this method remains subjective and has a reproducibility which depends on the interpreter's experience and the number of segments.6,7 Several methods have been tested in order to establish more objective approaches. The first attempts were based on traces of the endocardium.5 Sophisticated automatic detections of the endocardium based on the radio frequency signal have been developed more recently.8,9 The color kinesis approach studies the endocardial motion and depends on the gain setting. Quantitative two-dimensional tissue Doppler studies have been reported10 but they strongly depend on the orientation of the ultrasonic beam with respect to the direction of the wall motion and on the signal-to-noise ratio.11 Finally, strain rate calculations are currently tested.12,13 These various approaches have not, so far, gained acceptance for the automatic detection of wall motion abnormalities.

Some image processing methods have been developed to analyze regional wall motion a posteriori. One of them is based on a robust detection of both endocardial and epicardial borders,14 but its clinical evaluation is still under progress. A second one is based on Fourier transform to synthesize phase and amplitude images from the echocardiographic image sequence.15 Some correlation has been found between fractional area shortening and phase values, but the phase shows significant dispersion even in normal cases.15

The present work has been designed to test the potentialities of the factor analysis of the left ventricle by echocardiography (FALVE) for detecting segmental wall motion abnormalities. Indeed, the use of factor analysis was initially developed for dynamic tracer studies in nuclear medicine.16 It is well adapted to myocardial perfusion studies in PET,17 as well as in MRI,18 and has already been successfully tested for studying the segmental wall motion in MUGA.19 The major improvements which are presented in this paper are related to the mathematical introduction of specific constraints. For the present work, these constraints were derived from established knowledge in terms of wall and heart motion.

This study has been designed to evaluate: (1) the feasibility of the factor analysis based on two factors; (2) its potentials for detecting segmental wall motion abnormalities; and (3) the sources of misinterpretation.


   Principles
Top
 Abstract
 Introduction
 Principles
Material and methods
 Results
 Discussion
 Conclusion
 References
 
The factor analysis method processes a sequence of images, based on the analysis of the evolution of the signal intensity S(x,y,t) of each pixel (x,y) throughout the time t (Fig. 1). It describes the whole set of time signal intensity curves by a limited number of curves, fk(t) called factors. Thus, each individual time intensity curve is a linear combination of these factors:Formula


Figure 1
View larger version (20K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		Figure 1 Principle of the FALVE method: decomposition of an echocardiographic sequence corresponding to a cardiac cycle into two factors and factor images—one "constant" and another "contraction–relaxation". For each pixel, the decomposition is performed on the same set of factors. Each weighting parameter yields a coefficient for the corresponding factor image.

 
The set of coefficients Ik(x,y), which gives the contribution of each pixel (x,y) to the factor fk(t), constitutes a parametric image, called factor image. The term e(x,y,t) represents an error term for each pixel (x,y) at time t. It takes into account both noise and modeling error. One objective of the factor analysis is to keep the global error term Err={sum}x{sum}y{sum}te(x,y,t) as low as possible.

The first step of the factor analysis is called "orthogonal analysis". It enables to minimize the global error term Err, for a given number, K, of orthogonal components uk(t). This number of orthogonal components can reach the number of images included in the initial image sequence. But, in medical applications, the majority of the information is included in the first two, three, or four orthogonal components. In other terms, the image sequence S(x,y,t) can be approximated by Formula , which is a linear combination of the K most important orthogonal components uk(t):Formula being minimal.

This mathematical decomposition is optimal in the least-squares sense, but the orthogonal images Vk(x,y) and the orthogonal curves uk(t) are difficult to interpret in physiological terms. The orthogonal analysis performs a double filtering of curves and images.

We have hypothesized that, when processing sequences of left ventricular echocardiographic images, (i) the most significant information describing the contraction is included in the first two orthogonal factors and (ii) an oblique analysis under ad hoc constraints can extract an almost constant factor and a "contraction–relaxation" factor.

The "oblique analysis" defines two factor curves f1(t) and f2(t), and two factor images I1(x,y) and I2(x,y) with the use of linear combinations of the first two orthogonal components, u1(t), and u2(t). This second step is used to add constraints reflecting prior knowledge derived from physiology or physiopathology. It allows a further interpretation, which is not only based on statistical considerations. Therefore, the factor f1(t) was estimated with a constraint of stability over time, and the factor f2(t) with positive values, a minimal value equal to zero, and showing an increase during systole and then a decrease. Using such constraints, the first factor represents the continuous component and the second one the motion component.

Using a mathematical formulation, the factor f1(t) was equal to a1u1(t)+a2u2(t), a1 and a2 being real values such that {sum}t[a1u1(t)+a2u2(t)–1]2 was minimum. The factor f2(t) was f2(t)=b1f1(t)±u2(t), depending on the orientation of the component u2(t): +u2(t) if u2(t) showed an increasing pattern during the systolic part of the cycle, followed by a decreasing pattern during the diastolic part of the cycle, –u2(t) if it was the opposite. The coefficient b1 was estimated as the minimum value of f2(t) was set to 0. The factor f2(t) was then normalized by a factor {lambda} such that its maximum value was set to 1.

After the evaluation of these two factors f1(t) and f2(t), the two associated factor images I1(x,y) and I2(x,y) were computed in the least-squares sense, from initial raw data and factors. The first factor image, I1(x,y), has positive values, while the second one, I2(x,y), has either positive or negative values depending on whether the time signal intensity curve increases or decreases during the systole.


   Material and methods
Top
 Abstract
 Introduction
 Principles
 Material and methods
Results
 Discussion
 Conclusion
 References
 
Population
Loops from 38 patients referred for standard 2D and Doppler echocardiographic examination were collected and digitally recorded. Good quality apical two-chamber views were obtained in 29 of them. All the loops were acquired during routine examinations in order to be representative of in-hospital patients in terms of pathology, echogenicity, and sources of misinterpretation. The description of the population with a predominance of ischemic heart disease and abnormal LV function is given in Table 1. The population is heterogeneous but representative of the real life. Patients with left bundle branch block (LBBB) or pacemaker defined the group A (n = 12). The 26 other patients defined the group B which comprised a significant number of post-operative patients (n = 4).


View this table:
[in this window]
[in a new window]

  		Table 1 Main characteristics of patient population

 
Image acquisition
The apical two-dimensional harmonic gray scale sequences of patients were acquired using an HDI 5000 device (Philips-ATL, Bothell, USA) and digitally recorded with the use of HDI Lab software (Philips-ATL, Bothell, USA). Four-chamber (n = 38) and two-chamber (n = 29) views were used. For conducting the present study, series of three cycles were acquired and directly transferred from the cine loop of the ultrasound machine to an external PC.

Separated cycles were identified by selecting the onset of the QRS complex, and the associated end-diastolic images. For each patient, the cycle giving the best superimposed initial and final images was selected for the rest of the analysis. Corresponding AVI loops were extracted for the reference visual analysis. The raw numerical data were saved for the post-processing.

Image processing: factor analysis of the left ventricle by echocardiography, FALVE
The mitral valve, the right ventricle and the atria were excluded with a manually drawn mask. No other pre-processing method was used. The factor analysis was performed with a customized commercially available software (Pixies, Apteryx, Issy-les-Moulineaux, France). The orthogonal analysis was first achieved as shown in Fig. 2. The relative importance of the first two orthogonal factors was studied. These factors were kept for the oblique analysis. Two constraints were imposed: one factor f1(t) with a constraint of stability over time, and another factor f2(t) with positive values, a minimal value equal to 0, and showing an increase during systole and then a decrease. After the estimation of these two factors f1(t) and f2(t), the two associated factor images I1(x,y) and I2(x,y) were computed.


Figure 2
View larger version (82K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		Figure 2 Example of the first 10 components of an orthogonal analysis obtained for a normal patient. The images (in gray color for positive coefficients and in blue color for negative coefficients) with the superimposed curves (in red color) are displayed from left to right and from top to bottom according to their decreasing contribution to the entire sequence. Relative contributions to the first image are shown for the following images.

 
Fig. 3 shows the oblique factors and the corresponding factor images derived from the orthogonal analysis presented in Fig. 2. The first factor image (Fig. 3a) is roughly similar to the averaged image of the sequence. The second factor image (Fig. 3b) is made of positive (encoded in gray) and negative (encoded in blue) values depending on the increase or decrease of the gray scale values during the contraction. The exact timing of the "contraction–relaxation" factor with respect to valve opening and closing has not been studied.


Figure 3
View larger version (65K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		Figure 3 Example of the two oblique factors with the corresponding factor images, derived from the orthogonal analysis shown in Fig. 2. The first factor (a) is more constant than the first orthogonal curve and the second factor (b) is always positive and shows the contraction–relaxation pattern. Three-color display of the factor images. (c) The constant image is displayed according to a green color scale, the positive values of the contraction–relaxation image are coded according to a red color scale and the negative values according to a blue color scale.

 
For each loop, a three-color image was generated as the superimposition of the constant factor image encoded in green, the positive values of the contraction–relaxation factor image encoded in red and the negative values of the contraction–relaxation factor image encoded in blue (Fig. 3c).

Data analysis
The anonymous AVI gray scale loops were read by three experienced echocardiographers with the use of the 16 segments of the American Society of Echocardiography. The wall motion was scored from –1 (dyskinetic) to 2 (normal). Discordant results were solved by consensus. The echogenicity was scored for each segment from 0 (nonvisible) to 2 (well delineated). For each view, a mean contraction score was calculated as the mean of the visible segments.

Illustrative examples of three-color images were obtained during a pilot study and provided to the readers at the beginning of the present study (Fig. 4). The three-color images were read according to the following rules:

- normal (2) in the case of wide red signal towards the cavity (case a: all segments, case b: latero-basal segment, case c: anterior segments but the apex, case d: basal segments);
- hypokinetic (1) in the case of narrow red signal towards the cavity (case b: all segments but the latero-basal);
- akinetic (0) in the case of green or mosaic signal towards the cavity (case c: inferior segments but the apex);
- dyskinetic (–1) in the case of blue signal towards the cavity (cases c and d: apical segments).


Figure 4
View larger version (150K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		Figure 4 Typical examples of three-color images in case of normal and pathological segmental wall motions: (a) normal wall motion, (b) diffuse hypokinesy, (c) akinetic inferior wall, (d) dyskinetic apex.
 
The anonymous three-color images were scored by the three experienced echocardiographers without the knowledge of the AVI loops and the discordances were solved by consensus.

The comparisons between visual readings and the factor images computed by FALVE were made on the entire population and on the group B patients without LBBB or pacemaker. Contingency tables were plotted. The absolute and relative (within one grade of discordance) agreements were calculated. The visual and the FALVE mean wall motion scores were compared with the use of Bland–Altman diagrams. Finally, the discordant cases were studied in terms of echogenicity and with the use of additional image processing.


   Results
Top
 Abstract
 Introduction
 Principles
 Material and methods
 Results
Discussion
 Conclusion
 References
 
Orthogonal analysis (Fig. 2)
The first component of the orthogonal analyses was always predominant. Its mean contribution to the initial raw sequences was about 80%. It was found to be almost flat, thus supporting the first constraint of the oblique analysis. The second component of the orthogonal analyses corresponded to about 6% of the information in the initial sequences. It exhibited a pattern similar to the contraction–relaxation sequence. The cumulative contribution of the first two factors corresponded to about 86% of the information in the initial sequences and the relative weight of the third factor (and the followings) was below 3%, thus supporting the oblique analysis with only the first two components of the orthogonal step.

Three-color images (Figs. 4 and 5Go)
Fig. 4 shows typical cases which were obtained during the pilot study and were used to define the guidelines for the interpretation of the three-color images. Normal segments are visible in cases a–d. Hypokinetic segments are seen in case b. Akinetic segments are seen in case c and dyskinetic in cases c and d.

Fig. 5a shows an example of an almost nonvisible segment with a very weak signal clearly identified as corresponding to a low echogenicity. The almost nonvisible segment on the initial loop was encoded as nonvisible and FALVE did not provide misleading information. The inferior wall is, otherwise, normal.


Figure 5
View larger version (52K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		Figure 5 Example of a three-color image in case of a low signal for the lateral wall corresponding to a low echogenicity (a). Example of mitral annular motion (b).

 
Fig. 5b shows a nice example of mitral annular motion: the basal segments that were never excluded by the mask exhibit a clearly visible upward longitudinal motion. The other segments but the infero-apical are severely hypokinetic.

Comparisons
A total of 398 segments were read. Four segments were almost nonvisible on the loops and on the results of FALVE and could not be quoted. Table 2 is the contingency table between the FALVE image reading on the abscissa and the visual reading on the ordinate when including all the patients. An absolute agreement was obtained for 64.8% of the segments and a relative agreement (within one grade) was obtained for 97.2% of the segments. Among the 11 segments with a discordance exceeding one grade, two were basal septal segments with an outward motion but a clear thickening on the right side of the septum.


View this table:
[in this window]
[in a new window]

  		Table 2 Concordance table between visual analysis of cine-loops and visual analysis of three-color factor images for the whole population

 
Table 3 is the contingency table between the FALVE image reading on the abscissa and the visual reading on the ordinate when excluding the patients with LBBB and/or pacemaker. An absolute agreement was obtained for 71.0% of the segments and a satisfactory agreement (within one grade) was obtained for 99.6% of the segments.


View this table:
[in this window]
[in a new window]

  		Table 3 Concordance table between visual analysis of cine-loops and visual analysis of three-color factor images for the population after exclusion of patients with LBBB or pacemaker

 
The Bland–Altman representation between the visual mean contraction score and the mean contraction score derived from FALVE was plotted (Fig. 6). In the entire population (67 views), the mean difference was equal to 0.05, with a standard deviation of the differences equal to 0.38. The two points clearly deviating from the distribution corresponded to patients of group A with a large global motion of the heart. In group B (46 views), when excluding the patients with LBBB and/or pacemaker, the mean difference was equal to 0.01, with a standard deviation equal to 0.30. No outlier was observed in this group.


Figure 6
View larger version (19K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		Figure 6 Bland–Altman plot to compare mean contraction scores per view by visual inspection of cine-loops and by interpretation of three-color FALVE images: (a) for all the views, (b) for the views in group B.

 
Discordant cases
The 11 most significant discordant results were related to visually normal or hypokinetic segments encoded as dyskinetic or akinetic. A review of the initial loops confirmed the accuracy of the initial visual score. Only one was found in group B; it was a basal septal segment. The 10 others were found in group A (patients with LBBB and/or pacemaker). A careful analysis of the initial loops showed a translation of the heart during the cycle despite properly superimposed end-diastolic images. Several additional post-processing methods were attempted. Different color displays or oblique studies derived from three instead of two orthogonal factors did not provide satisfactory encoding. On the contrary, an automatic alignment of the entire image sequence (correlation method included as an option in the Pixies package) before the FALVE analysis did correct most of the misleading encoding. As shown in Fig. 7a, a septal dyskinesy was found on a normal left ventricle with wide LBBB (top); this problem was almost totally solved after automatic alignment (bottom). The same alignment has been tested on other cases and it never deteriorated the results (Fig. 7b and c). The influence of echogenicity was studied by comparing the score of echogenicity of the perfectly agreeing segments with the others. We did not find any difference: 1.68±0.37 versus 1.65±0.37.


Figure 7
View larger version (115K):
[in this window]
[in a new window]
[Download PowerPoint slide]
  		Figure 7 Typical examples of three-color images obtained on raw echocardiographic data (top), and after automatic alignment (bottom). (a) False dyskinesis due to a large global motion of the heart (patient with LBBB). (b, c) Other cases without global motion.

 

   Discussion
Top
 Abstract
 Introduction
 Principles
 Material and methods
 Results
 Discussion
Conclusion
 References
 
Mathematical background
This new application of factor analysis depicts the wall motion by studying the entry, the presence or the exit of the wall at the level of each pixel and is based on the variations of the time signal amplitude curves. In this sense, the approach is innovative. Furthermore, it might become a fully automatic post-processing.

The results obtained during this study showed that the segmental wall motion can be mainly described by the use of the first two orthogonal components, which summed up more than 86% of the information of the raw image sequence. The time behavior of the orthogonal components has prompted the subsequent oblique processing and the introduction of the a priori constraints: a constant factor and a "contraction–relaxation" factor. Moreover, this a priori knowledge derived from physiology was introduced in the mathematical process, which guarantees the uniqueness of the solution. These two factors enable to separate the information between the mean gray level and the variations which are due to wall motion. The interpretation of the three-color factor images can thus be the following: pixels which correspond to the regions where there is no motion effect or no motion at all are mainly coded in green color, pixels which show a contraction–relaxation pattern are coded in red color, they are located at the border of the endocardium in the normal cases. Inversely, the pixels located at the outer side of the wall show an opposite pattern and are coded in blue color. Finally, in the case of systolic expansion, pixels located at the endocardium show a negative contraction–relaxation pattern which is encoded in blue.

As a result, the factor images provided a satisfactory description of the segmental wall motion and could be interpreted for the calculation of a global wall motion score.

Motion artifacts
A global heart motion can be induced by breathing and/or probe motion during acquisition. This type of motion could deeply modify the constant factor image that is very close to the mean image. A strict apnea during the acquisition can partially eliminate this problem. In order to avoid this type of artifact, the studied cycles were carefully selected according to the end-diastolic images: the selected cycles had almost superimposed end-diastolic images. A numeric subtraction of end-diastolic images could easily automatically detect this source of artifact. Another global motion can occur during the heart cycle, in particular in the case of LBBB, pacemaker or in post-operative patients. Indeed, this study showed that most discordant cases occur in the group A. A realignment procedure has improved the reliability of the results but further studies are necessary to systematically test this hypothesis. The entry of structures in the scanning plane is also of "false" motion. The present "2D+t" approach does not allow to solve this problem but an extension to a "3D+t" solution can be easily conceived and it will have to be tested as soon as the real-time echocardiography will provide a sufficient frame rate.

Clinical results
The results obtained on the entire population have clearly been deteriorated by its heterogeneity. The results of the clinical evaluation have been good in the group without LBBB and/or pacemaker. This group is similar to the populations included in most of the comparative studies devoted to the left ventricular function. Indeed, LBBB and/or pacemaker induce an asynchronism with a global motion of the heart which affects the endocardial displacement but preserves the wall thickening. In the present study, this problem has been almost totally solved by an image realignment. Further studies are necessary to validate a single robust tool giving satisfactory results in each type of patient. This type of tool might be very useful in routine.

Comparison with other methods
Color kinesis approach8,9,20 is derived from a complex post-processing of the integrated backscatter. Unlike FALVE, it strictly depends on the detection of the endocardium. On one hand, it does not allow different types of pre-processing like the alignment and, in case of low signal-to-noise ratio, it may deviate from the endocardial interface and provide misleading information. On the other hand, color kinesis is a real-time process (FALVE can be achieved in less than 1 min) and it provides a detailed description of the systolic events. Other approaches have been tested with the use of the integrated backscatter.21,22 So far, they do not provide a parametric imaging but they lead to curves that look like the second oblique factor. These curves show variations in the intensity of the echo generated within the myocardium whilst the second oblique factor of the factor analysis extract variations related to the transition between the cavity and the myocardium. Theoretically, the fluctuations of the intensity of the myocardium might influence the factors but, in the present study, they have been limited by the use of harmonic imaging and the orthogonal analysis.

Color Doppler10,11 or strain rate studies12,13 depend on the orientation of the motion with respect to the ultrasonic beam and a black color Doppler pixel is an ambiguous information11 since it may equally depict a zero velocity or a poor signal-to-noise ratio. In addition, displays based on the velocity rather than on the spatial velocity gradient (strain rate) are strongly affected by the physiological differences between base and apex.23

Fourier analyses separate a continuous component from periodic factors. They impose a sinusoidal shape to all the factors. Left ventricular analyses are restricted to the first sinusoidal component but they measure different phases. By comparison, FALVE does not impose a sinusoidal shape but it does not measure different phases. A Fourier analysis of gray scale echo sequences has been tested on an animal model.24 Results on patients have been reported.15 The authors showed a relationship between the mean phase shift of a segment and the wall motion score. On the other hand, they found for each segment, a rather wide histogram leading to noisy parametric images. Fourier analysis has also been used for detecting left ventricular asynchrony:25 a semi-automatic edge detection has been applied on sequences of several cycles acquired during apnea, curves of the mean displacement of the septum and the lateral wall have been constructed, a Fourier analysis has been used to measure the phase shift between these two curves. This processing has required sophisticated smoothing and averaging. It has provided a value for the phase shift but not a parametric imaging.

Practical implementation
In the present study, FALVE has been used as a semi-automatic approach with the selection of an appropriate heart cycle and with a manually drawn mask. Further development could allow an automatic detection of the cycles having the best superimposed end-diastolic images: calculation of the difference between initial and final end-diastolic images, and selection of the cycle providing the lowest difference. As concerns the mask, further studies are needed to establish its real impact. On the other hand, the identification of the end-diastolic image of the mitral annulus by two points and of the apex by one point could lead to a very fast determination of this mask. Finally, an automatic quantification of the segmental wall motion could be derived from the width of color bands in the three-color image. However, a learning stage should be implemented on a larger database, in order to take into account the possible variations which are due to the location or the echogenicity of the segments.

Further applications
The FALVE method is easy to implement and the selection of the best cycle can be automated. Provided the acquisition of loops correspond to complete cycles, this method could help in the reading of stress echo examinations.6 The same post-processing could be used on sequences of MRI images of the left ventricle. More complex approaches specially designed to identify post-systolic shortening could be conceived. Finally, it could be used to separate wall motion from perfusion in real-time contrast sequences analyzing flash and refill kinetics.26


   Conclusion
Top
 Abstract
 Introduction
 Principles
 Material and methods
 Results
 Discussion
 Conclusion
References
 
A new post-processing method, based on factor analysis, has been proposed to detect segmental wall motion abnormalities. It has allowed the development of a parametric imaging encoding normal and abnormal wall motion. This first clinical evaluation has shown a good agreement with the visual analysis in the absence of LBBB or pacemaker. It appears to be able to lead to an automatic detection of segmental wall motion abnormalities. Further technical improvement and automation could be conceived.


   Acknowledgements
 
This work was supported in part by grants from the Fédération de Cardiologie, Fondation de France, and ACI programme.


   References
Top
 Abstract
 Introduction
 Principles
 Material and methods
 Results
 Discussion
 Conclusion
 References
 

  1. Jacobs J., Feigenbaum H., Corya B., Phillips J. Detection of left ventricular asynergy by echocardiography. Circulation (1973) 48:263–271.[Abstract/Free Full Text]
  2. Guyer D., Foale R., Gillam L., Wilkins G., Guerrero J., Weyman A. An echocardiographic technique for quantifying and displaying the extent of regional left ventricular dyssynergy. J Am Coll Cardiol (1986) 8:830–835.[Abstract]
  3. Heger J., Weyman A., Wann L., Rogers E., Dillon J., Feigenbaum H. Cross-sectional echocardiographic analysis of the extent of left ventricular asynergy in acute myocardial infarction. Circulation (1980) 61:1113–1118.[Free Full Text]
  4. Pandian N., Skorton D., Collins S., Koyanagi S., Kieso R., Marcus M., et al. Myocardial infarct size threshold for two-dimensional echocardiographic detection: sensitivity of systolic wall thickening and endocardial motion abnormalities in small versus large infarcts. Am J Cardiol (1985) 55:551–555.[CrossRef][Web of Science][Medline]
  5. McGillem M., Mancini G., DeBoe S., Buda A. Modification of the centerline method for the assessment of echocardiographic wall thickening and motion: a comparison with areas of risk. J Am Coll Cardiol (1988) 11:851–866.[Abstract]
  6. Oberman A., Fan P., Nanda N., Lee J., Huster W., Sulentic J., et al. Reproducibility of two-dimensional exercise echocardiography. J Am Coll Cardiol (1989) 14:923–928.[Abstract]
  7. Badano L., Stoian J., Cervesato E., Bosimini E., Gentile F., Giannuzzi P., et al. Reproducibility of wall motion score and its correlation with left ventricular ejection fraction in patients with acute myocardial infarction. Am J Cardiol (1996) 78:855–858.[CrossRef][Web of Science][Medline]
  8. Lau Y., Puryear J., Gan S., Fowler M., Vagelos R., Popp R., et al. Assessment of left ventricular wall motion abnormalities with the use of color kinesis: a valuable visual and training aid. J Am Soc Echocardiogr (1997) 10:665–672.[CrossRef][Web of Science][Medline]
  9. Mor-Avi V., Collins K., Korcarz C., Shah M., Kirk T., Spencer K., et al. Detection of regional temporal abnormalities in left ventricular function during acute myocardial ischemia. Am J Physiol Heart Circ Physiol (2001) 280:H1770–H1781.[Abstract/Free Full Text]
  10. Wilkenshoff U., Sovany A., Wigström L., Olstad B., Lindström L., Engvall J., et al. Regional mean systolic myocardial velocity estimation by real-time color Doppler myocardial imaging: a new technique for quantifying regional systolic function. J Am Soc Echocardiogr (1998) 11:683–692.[CrossRef][Web of Science][Medline]
  11. Garot J., Diebold B., Derumeaux G., Monin J., Bosio P., Duval-Moulin A., et al. Comparison of regional myocardial velocities assessed by quantitative 2-dimensional and M-mode color Doppler tissue imaging: influence of the signal-to-noise ratio of color Doppler myocardial images on velocity estimators of the Doppler tissue imaging system. J Am Soc Echocardiogr (1998) 11:1093–1105.[CrossRef][Web of Science][Medline]
  12. Urheim S., Edvardsen T., Torp H., Angelsen B., Smiseth O. Myocardial strain by Doppler echocardiography. Validation of a new method to quantify regional myocardial function. Circulation (2000) 102:1158–1164.[Abstract/Free Full Text]
  13. Jamal F., Strotmann J., Weidemann F., Kukulski T., D'Hooge J., Bijnens B., et al. Noninvasive quantification of the contractile reserve of stunned myocardium by ultrasonic strain rate and strain. Circulation (2001) 104:1059–1065.[Abstract/Free Full Text]
  14. Jacob G., Noble J., Kelion A., Banning A. Quantitative regional analysis of myocardial wall motion. Ultrasound Med Biol (2001) 27:773–784.[CrossRef][Web of Science][Medline]
  15. Hansen A., Krueger C., Hardt S., Haass M., Kuecherer H. Echocardiographic quantification of left ventricular asynergy in coronary artery disease with Fourier phase imaging. Int J Cardiovasc Imaging (2001) 17:81–88.[CrossRef][Medline]
  16. Di Paola R., Bazin J., Aubry F., Aurengo A., Cavailloles F., Herry J., et al. Handling of dynamic sequences in nuclear medicine. IEEE Trans Nucl Sci (1982) 29:1310–1321.[Web of Science]
  17. Wu H., Hoh C., Buxton D., Schelbert H., Choi Y., Hawkins R., et al. Quantification of myocardial blood flow (MBF) using dynamic N-13 ammonia dynamic PET studies and factor analysis of dynamic structures. J Nucl Med (1995) 36:2087–2093.[Abstract/Free Full Text]
  18. Janier M., Mazzadi A., Lionnet M., Frouin F., André-Fouet X., Cinotti L., et al. Factor analysis of medical image sequences improves evaluation of first-pass MR imaging acquisitions for myocardial perfusion. Acad Radiol (2002) 9:26–39.[CrossRef][Web of Science][Medline]
  19. Cavailloles F., Bazin J., Pavel D., Olea E., Faraggi M., Frouin F., et al. Comparison between factor analysis of dynamic structures and Fourier analysis of segmental wall motion abnormalities: a clinical evaluation. Int J Cardiovasc Imaging (1995) 11:262–272.
  20. Vitarelli A., Sciomer S., Penco M., Dagianti A., Pugliese M. Assessment of left ventricular dyssynergy by color kinesis. Am J Cardiol (1998) 81:86G–90G.[CrossRef][Web of Science][Medline]
  21. Barzilai B., Vered Z., Mohr G.A., Wear K.A., Courtois M., Sobel B.E., et al. Myocardial ultrasonic backscatter for characterization of ischemia and reperfusion: relationship to wall motion. Ultrasound Med Biol (1990) 16:391–398.[CrossRef][Web of Science][Medline]
  22. Pasquet A., d'Hondt A.M., Melin J.A., Vanoverschelde J.L. Relation of ultrasonic tissue characterization with integrated backscatter to contractile reserve in chronic left ventricular ischemic dysfunction. Am J Cardiol (1998) 81:68–74.[CrossRef][Web of Science][Medline]
  23. Cain P., Baglin T., Khoury V., Case C., Marwick Th. Automated regional myocardial displacement for facilitating the interpretation of dobutamine echocardiography. Am J Cardiol (2002) 89:1347–1353.[CrossRef][Web of Science][Medline]
  24. Kuecherer H., Schoels W., Sterns L., Freigang K., Kleber G., Brachmann J., et al. Echocardiographic Fourier phase and amplitude imaging for quantification of ischemic regional wall asynergy: an experimental study using coronary microembolization in dogs. J Am Coll Cardiol (1995) 25:1436–1444.[Abstract]
  25. Breithardt O.A., Stellbrink C., Kramer A.P., Sinha A.M., Franke A., Salo R., et al. Echocardiographic quantification of left ventricular asynchrony predicts an acute hemodynamic benefit of cardiac resynchronization therapy. J Am Coll Cardiol (2002) 40:536–545.[Abstract/Free Full Text]
  26. Masugata H., Peters B., Lafitte S., Strachan G.M., Ohmori K., De Maria A.N. Quantitative assessment of myocardial perfusion during graded coronary stenosis by real-time myocardial contrast echo refilling curves. J Am Coll Cardiol (2001) 37:262–269.[Abstract/Free Full Text]

Add to CiteULike CiteULike   Add to Connotea Connotea   Add to Del.icio.us Del.icio.us    What's this?



This Article
Right arrow Abstract Freely available
Right arrow FREE Full Text (PDF) Freely available
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrowRequest Permissions
Right arrow Disclaimer
Google Scholar
Right arrow Articles by Frouin, F.
Right arrow Articles by Diebold, B.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Frouin, F.
Right arrow Articles by Diebold, B.
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?