Technical Note. CORRECT ANALYSIS of regional ventricular function is central to the diagnosis of coronary heart disease and to

Transcription

1 Technical Note JOURNAL OF MAGNETIC RESONANCE IMAGING 26: (2007) Evaluation of Regional Myocardial Function Using Automated Wall Motion Analysis of Cine MR Images: Contribution of Parametric Images, Contraction Times, and Radial Velocities Nadjia Kachenoura, MS, 1,3 * Alban Redheuil, MD, 1 3 Daniel Balvay, PhD, 1,3 Cinta Ruiz-Dominguez, PhD, 1,3 Alain Herment, PhD, 1,3 Elie Mousseaux, MD, PhD, 1 3 and Frédérique Frouin, PhD 1,3 Purpose: To develop fast and robust procedures for a clinical evaluation of regional myocardial contractile function. Materials and Methods: Parametric analysis of main motion was applied to steady-state free-precession (SSFP) cine MR images. From the time signal intensity curve associated with each pixel, parametric maps of mean high and low amplitudes and transition times between muscle and cavity were automatically computed. Then, regional time to first contraction, T fc, mean contraction time, T mc and radial component of the endocardial velocity, V m were estimated from these parametric maps and a user-defined endocardial end-diastolic contour. The method was applied to short-axis slices in 22 subjects: eight controls, 13 myocardial infarctions (MIs), and one left bundle branch block (LBBB). Results: Typical patterns of normality and pathology on parametric maps are indicated. For controls, the mean values standard deviations (SDs) of T fc,t mc, and V m were: msec, msec, and cm second 1. An apex to base gradient of T fc, a significant septal delay in T fc and T mc, and a decrease of V m between the lateral and septal walls were observed. For MI, T fc and T mc increased and V m decreased significantly in pathological segments. For LBBB, large delays were estimated in the septal wall. Conclusion: The proposed method is promising for clinical assessment of regional wall contraction. Key Words: cine MR imaging; wall motion; contraction time; radial velocity; parametric image J. Magn. Reson. Imaging 2007;26: Wiley-Liss, Inc. CORRECT ANALYSIS of regional ventricular function is central to the diagnosis of coronary heart disease and to 1 Institut Nationale de la Santé et de la Recherche Médicale (INSERM), U678, Paris, France. 2 Assistance Publique Hôpitaux de Paris (AP-HP), Hôpital Européen Georges Pompidou, Service de Radiologie Cardiovasculaire, Paris, France. 3 Université Pierre et Marie Curie, Faculté demédecine Pitié-Salpêtrière, Paris, France. *Address reprint requests to: N.K., U678 INSERM, CHU Pitié-Salpêtrière, 91 Boulevard de l Hôpital, F Paris cedex, France. Nadjia.Kachenoura@imed.jussieu.fr Received April 6, 2006; Accepted July 3, DOI /jmri Published online in Wiley InterScience ( the assessment of myocardial viability (1,2). Cardiac MR has become a relevant modality for studying segmental myocardial perfusion, and contractile function (3). Clinical assessment of segmental function of the left ventricle is mostly based on visual interpretation of cine MR images. Accurate diagnosis requires the ability of the reader to efficiently integrate both temporal and spatial information on endocardial wall motion and myocardial wall thickening. Such visual analysis requires an extensive training. To reduce the variability of visual wall motion analysis, quantitative postprocessing methods have been proposed. Among them, the two-dimensional centerline method, initially developed in angiocardiography (4), estimates wall motion and wall thickening from the endocardial and epicardial borders of the myocardium. This method has proved to be accurate in MRI (5), thanks to the excellent depiction of the myocardium. Some different approaches summarized the contraction information contained in cine MR into one or few synthetic still frames. As such, the first harmonics of the discrete Fourier transform image (6), and methods based on factor analysis (7) supplied parametric images in relation to the wall motion. In this study, the parametric analysis of main motion proposed in echocardiography (8) was applied to cine MR images. It is based on a nonlinear transition model defined by an adaptive window function and estimates four parameters from the time signal intensity curve of each pixel. This model has already proved to fit contraction data better than the Fourier or the factor analysis models (8). The four parametric images provide information related to both temporal and amplitude changes due to the wall motion and myocardial wall thickening during the cardiac cycle from which motion abnormalities can be easily identified. The parametric analysis of main motion was adapted to short-axis MR images and a new quantitative process was proposed to extract the segmental mean contraction times, the times to first contraction, and the mean radial velocities from the parametric images. Results obtained on eight control subjects and on 14 patients with myocardial infarction (MI) or left bundle branch block (LBBB) are given. Their physiological relevance is 2007 Wiley-Liss, Inc. 1127

2 1128 Kachenoura et al. Figure 1. Three patterns of time signal intensity curves according to the distance of the corresponding pixels to the endocardial wall on the end-diastole image (d0 d1 d2): first transition times, T ON, are smaller for outer pixels, while second transition times, T OFF, are larger for those pixels. Furthermore there is a larger difference between T ON, times than between T OFF times. underlined, before indicating the limitations and potential interests of the method. MATERIALS AND METHODS Study Population A total of eight subjects (male, mean age 44 years, range years) with normal cardiac MR (CMR) examination and normal ventricular function were selected. This group was referred to as control subjects. A total of 13 patients (11 male, two female, mean age 61 years, range years) with MI and one patient with LBBB (male, age 61 years) were also studied. Imaging Protocol All studies were performed according to a standard clinical protocol, approved by the institutional review board. Images were obtained with a 1.5-T MRI system (Signa LX; General Electric Medical Systems, Waukesha, WI, USA) using electrocardiogram gating with fiber-optic leads and a thoracic phased-array surface coil. Breathholding was used to reduce respiratory motion. Cine data were acquired according to a steadystate free-precession (SSFP) technique; cine segmented fast imaging employing steady-state excitation (FI- ESTA) was performed in short-axis views from the atrioventricular rings to the apex using the following parameters, pulse repetition time msec, echo time 1.7 msec, flip angle 40, views-per-segment 10 14, cardiac phase after view sharing, slice thickness 8 mm, acquisition matrix size pixels, partial phase FOV cm, and temporal resolution msec. Image Processing Preprocessing Step For each patient, three short-axis series were selected at apical, midventricular, and basal levels. For all the studies, the regional wall motion score was assessed by a consensus of two experts, according to a four-point scale (normokinesia, hypokinesia, akinesia, and dyskinesia). The center of the left ventricle and the anterior intersection between the two ventricles were determined in order to further divide the apical levels into four segments, the midventricular and basal levels into six segments, in conformity with the standardized segmentation (9). Finally, the end-diastolic endocardial contour was manually traced using the MASS software (Medis, Leiden, Netherlands). Time Signal Intensity Curve Modeling A total of three types of time signal intensity curves P(x,y,t) associated with the pixel (x,y) were observed. A decreasing then increasing shape was found in the pixels that remained within the cavity at the end of diastole, close to the endocardial border; during systole their signal decreased as they became part of the myocardium (Fig. 1). The second one was roughly flat; it represented pixels that were within the myocardium during the whole cardiac cycle. An increasing then decreasing curve with low amplitude was found in the pixels that stayed inside the cavity during the whole cardiac cycle. This variation was due to the inflow effect during rapid changes in flow velocity within cavity during ejection and rapid diastolic filling in early diastole. The parametric analysis of main motion based on the window function assumed a high contrast between the cavity and the myocardium, implying that they could be distinguished by two distinct signal intensity levels, and that the transition between these two states was fast: P x,y,t A B x,y A V x,y g 2 t,t ON x,y,t OFF x,y e P x,y,t, (1) with g 2 t,t ON x,y,t OFF x,y 1 if T ON x,y t T OFF x,y 0 otherwise. (2) The coefficient A B (x,y) was the baseline signal intensity, corresponding to end-diastole, A V (x,y) was the variation of signal intensity, g 2 (t) was the window function, which depended on the two transition times T ON (x,y) and T OFF (x,y), and e P (x,y,t) was the residual error. A fast

3 Wall Motion Analysis of Cine MR Images 1129 algorithm was developed to estimate the four parametric images A B, A V, T ON, and T OFF while minimizing e P (8). Interpretation of Parametric Images The image A B was close to the end-diastole image, the image A V was close to zero outside the heart region, where there was no contraction-relaxation movement, and its largest positive variations were between the endocardial end-diastolic and the end-systolic contours. Owing to the inflow effect, A V showed negative values inside the cavities. The transition times images T ON and T OFF described the chronology of wall motion. The T ON image showed an onion-ring pattern with increasing values from the endocardial end-diastolic contour toward the center of the cavity (Fig. 2). Conversely the T OFF image showed decreasing values when going from the endocardial end-diastolic contour toward the center of the cavity. The combination of T ON and T OFF images into one mean contraction time image: T M (T ON T OFF )/2 showed homogeneous values in normal subjects. The time values were converted in delays from the R-wave expressed in milliseconds (msec) and normalized using Bazett s formula. The time images were coded in warm colors for positive values of A V, and in cold colors for negative values. The signal inside the cavity (negative values of A V ) was discarded, the resulting maps being referred as T ON and T M. For qualitative analysis, these maps were superimposed on the end diastolic image, coded in gray levels. Quantitative Analysis of T ON and T M Maps (Fig. 2) The analysis was restricted to pixels located inside the region defined by the end-diastolic endocardial contour. For each slice, a distance map from the endocardial border was expressed in centimeters. The isolated points of the map T ON were removed by a 3 3 median filter. For each segment, connected pixels with the same transition time T ON defined isotime regions. These regions were associated if positioned at the same distance from the end-diastolic endocardial border, and regions with less than four pixels were excluded. A dot plot was drawn by associating the mean transition time of each region with its corresponding mean distance, computed from the distance map. Finally, a linear fit was applied, by weighting each point according to the number of pixels in the region that provided it. The mean radial velocity of the segment, V m, was defined as the slope, and the time to first contraction of the endocardium, T fc, was the shortest transition time in the dot plot. Similarly, for each segment, the mean contraction time corresponding to the isotime region from the T M map having the highest number of pixels, was defined as the regional mean contraction time, T mc. Statistical Analysis For control subjects, mean values standard deviation (SD) of T fc, V m, and T mc were computed as a function of slice position and the cardiac wall. For patients, the variations of these parameters on pathological segments were studied. For statistical comparison, the Tukey-Kramer test was used, with a P value less than 0.05 to indicate significance. RESULTS In the control group, T ON images showed uniform layers around the cavity of the left ventricle, and T M images showed a large continuous and homogenous red band (Fig. 3a). Overall, the thickness of the color bands was fairly uniform with the exception of a reduced thickness of the septal wall, particularly in the basal slices. In patients with MI, the T ON image showed a lower number of layers in pathological segments, while a reduced thickness of the color band and delays in segmental contraction were highlighted on T ON and T M images (Fig. 3b). For the patient with LBBB, the thickness of the color bands was normal, but the septal wall highlighted an important delay in wall contraction (Fig. 3c). In the control group, T mc and T fc were estimated for all the segments, and V m in 86% of segments. For the 18 remaining segments, the linear fit was visually unsatisfactory: it concerned 11 segments containing papillary muscles, 4 located in apical slices and 3 in the septal wall of basal slices. Figure 4 shows the extracted parameters averaged over all control subjects, in bull seye diagrams according to slice position and ventricular segment. Overall, the mean T fc msec, the mean V m cm second 1, and the mean T mc msec. An increasing delay in time to first contraction, T fc, was observed from base (61 23 msec) to apex (80 26 msec), the mean differences of T fc (22 12 msec) being statistically different from zero. Moreover, an intraventricular septal delay of T fc was observed: msec vs msec for the anterior wall, msec for the inferior wall, and msec for the lateral wall. The mean differences between the septal and the lateral wall (24 6 msec) was statistically different from zero. The mean contraction time T mc was almost homogeneous. No significant delay was found between the basal ( msec), the midventricular ( msec), and the apical ( msec) slices. However, T mc in the septal wall was delayed in comparison with the lateral wall, with a significant difference of 28 9 msec. The mean radial velocity, V m, showed no significant differences between slices: cm second 1 for the apical, cm second 1 for the midventricular, and cm second 1 for the basal slices. However, the differences between septal and lateral walls ( cm second 1 ) was statistically significant. For patients with MI, of the 208 segments visually interpreted on cine loops, 129 had normal contraction, 53 were hypokinetic, and 26 were akinetic. The parameters T mc and T fc were estimated in all segments and V m was estimated in 178 segments and could not be estimated in five segments because of akinesia and in 25 segments because of the papillary muscles or apical trabeculae. In pathological segments, radial velocities decreased and T fc and T mc increased (Fig. 5). Moreover,

4 1130 Kachenoura et al. Figure 2. Example of T ON map, T ON map, and distance map computed from the end-diastolic endocardial border, and standard segmentation of the left ventricle. Dot plot derived from the posterolateral segment (in abscissa: T ON expressed in msec; in ordinate: distance to the left ventricular border expressed in cm) and estimation of the time to first contraction, T fc, and the mean radial velocity, V m. when compared to visual scores, T fc was found to be the most discriminating quantitative parameter (Fig. 5). For the patient with LBBB, large delays (118 msec for T fc, 125 msec for T mc ) were found between the septal and the lateral walls. However, the difference of V m ( 2.4 cm second 1 ) was similar to that of control subjects. DISCUSSION Regional evaluation of left ventricular function is based on endocardial excursion and myocardial thickening. Since visual assessment of wall motion is experiencedependent and subjective, quantitative techniques Figure 3. Examples of parametric images, T ON maps (up) and T M maps (down) presented for three subjects: (a) control subject, (b) subject with inferior myocardial infarction, and (c) subject with left ventricular bundle branch block.

5 Wall Motion Analysis of Cine MR Images 1131 Figure 4. Bull s-eye diagram showing mean values SD of (a) time to first contraction T fc (msec), (b) mean contraction time T mc (msec), and (c) mean radial velocity, V m (cm second 1 ), estimated for the eight control subjects. have been developed using automated or manual tracing of the endocardial and epicardial borders. However, these methods are time-consuming and not practical for clinical use. The method that is proposed here computes parametric images automatically in few seconds. It has focused on chronological information of wall motion via T ON and T M images, which is not provided when working only on end-diastolic and end-systolic images. Parametric T ON images represent the inward timemotion history of the systolic endocardial excursion from end-diastole to end-systole. These images have some similarities with color kinesis images, which have proved their usefulness in the detection of wall motion abnormalities, particularly to distinguish normal contraction and hypokinesia (10). The T M maps include information for inward and outward motion and can illustrate easily contraction delays between walls. Quantitative parameters: times to first contraction, mean contraction time, and the radial component of the endocardial velocity can be estimated semiautomatically with the delineation of one single contour. If they are not sufficient to describe the cine examination exhaustively, their potential interest is demonstrated. Indeed, their values are consistent with physiological values assessed in echocardiography or in cardiac MRI. Thus, the value of radial velocities in the control group is comparable to values established in echocardiography (11). Moreover the lower values of V m in the septal wall when compared with the lateral wall are consistent with the literature (12), as well as the reduction of V m for segments with motion abnormalities (13). When comparing the V m provided here with the estimated values using three-dimensional velocity mapping (14), the same range of values is found, with smaller variation coefficients in the present study. This effect may be due to a shorter acquisition time (one apnea vs. a half-hour). The time to first contraction is conceptually close to the onset time of circumferential shortening in early systole (15); a delay in the septal walls was reported (15), which is in agreement with the present study. The earliest inward motion near the base that is found here was already described (16). The temporal parameters can both outline pathological delays, which are difficult to assess visually when they are below msec (17). Moreover, T fc is more discriminating than T mc for the description of wall motion abnormalities (Fig. 5). If the numerical significance of V m is low for akinetic segments (due to a reduced number of points on the dot plot), these segments can be easily visually detected by a thin color band on the T ON map and a delayed pattern on T M map; thus, they cannot be missed. Moreover, if it is possible to take the papillary muscles into account when establishing the visual interpretation, the estimation of V m on the concerned segments can be doubtful; the only solution would be to remove them before computing the dot plot. The proposed approach, as for all the methods that are based on time signal intensity curves, is sensitive to global heart motion. One solution would be to correct for it, prior to the analysis of the cine MR. Another limitation is due to the slice obliquity, but information provided by the long-axis slices could be Figure 5. Quantitative parameters: (a) V m,(b) T fc, and (c) T mc extracted from normal, hypokinetic, and akinetic segments of the 13 patients with myocardial infarction; (*) indicates a statistical significant difference between two groups.

6 1132 Kachenoura et al. used to overcome this geometrical problem, as it has been proposed with the two-dimensional centerline method (18). In conclusion, the proposed method can outline easily physiological nonhomogeneities in the normal left ventricle and parameters due to wall motion abnormalities. These findings show its promising role in the study of wall motion asynchrony and in characterizing wall motion abnormalities. With the advent of sequences with higher temporal resolution, it could be valuable in detecting contractile reserve and/or myocardial ischemia from stress protocols and in extracting diastolic information to detect early dysfunction. REFERENCES 1. Hammermeister KE, DeRouen TA, Dodge HT. Variables predictive of survival in patients with coronary disease. Selection by univariate and multivariate analyses from the clinical, electrocardiographic, exercise, arteriographic, and quantitative angiographic evaluations. Circulation 1979;59: Lieberman AN, Weiss JL, Jugdutt BI, et al. Two-dimensional echocardiography and infarct size: relationship of regional wall motion and thickening to the extent of myocardial infarction in the dog. Circulation 1981;63: Wu KC, Zerhouni EA, Judd RM, et al. Prognostic significance of microvascular obstruction by magnetic resonance imaging in patients with acute myocardial infarction. Circulation 1998;97: Sheehan FH, Stewart DK, Dodge HT, Mitten S, Bolson EL, Brown BG. Variability in the measurement of regional left ventricular wall motion from contrast angiograms. Circulation 1983;68: Holman ER, Vliegen HW, van der Geest RJ, et al. Quantitative analysis of regional left ventricular function after myocardial infarction in the pig assessed with cine magnetic resonance imaging. Magn Reson Med 1995;34: Caiani EG, Toledo E, MacEneaney P, Collins KA, Lang RM, Mor-Avi V. The role of still-frame parametric imaging in magnetic resonance assessment of left ventricular wall motion by non-cardiologists. J Cardiovasc Magn Reson 2004;6: Redheuil A, Lyon X, Azarine A, et al. Inter observer variability in the evaluation of left ventricular segmental function can be reduced by combining visual assessment of MRI cine sequences and analysis of corresponding parametric images of myocardial contraction. J Cardiovasc Magn Reson 2005;7: Ruiz Dominguez C, Kachenoura N, De Cesare A, et al. Assessment of left ventricular contraction by parametric analysis of main motion (PAMM): theory and application for echocardiography. Phys Med Biol 2005;50: Cerqueira MD, Weissman NJ, Dilsizian V, et al. Standardized myocardial segmentation and nomenclature for tomographic imaging of the heart: a statement for healthcare professionals from the Cardiac Imaging Committee of the Council on Clinical Cardiology of the American Heart Association. Circulation 2002;105: 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. 11. Miyatake K, Yamagishi M, Tanaka N, et al. New method for evaluating left ventricular wall motion by color-coded tissue Doppler imaging: in vitro and in vivo studies. J Am Coll Cardiol 1995;25: Garot J, Derumeaux GA, Monin JL, et al. Quantitative systolic and diastolic transmyocardial velocity gradients assessed by M-mode colour Doppler tissue imaging as reliable indicators of regional left ventricular function after acute myocardial infarction. Eur Heart J 1999;20: Krahwinkel W, Haltern G, Gulker H. Echocardiographic quantification of regional left ventricular wall motion with color kinesis. Am J Cardiol 2000;85: Kvitting JP, Ebbers T, Engvall J, Sutherland GR, Wranne B, Wigstrom L. Three-directional myocardial motion assessed using 3D phase contrast MRI. J Cardiovasc Magn Reson 2004;6: Zwanenburg JJ, Gotte MJ, Kuijer JP, Heethaar RM, van Rossum AC, Marcus JT. Timing of cardiac contraction in humans mapped by high-temporal-resolution MRI tagging: early onset and late peak of shortening in lateral wall. Am J Physiol Heart Circ Physiol 2004; 286:H1872 H Hammermeister KE, Gibson DG, Hughes D. Regional variation in the timing and extent of left ventricular wall motion in normal subjects. Br Heart J 1986;56: Sutherland GR, Kukulski T, Kvitting JE, et al. Quantitation of left-ventricular asynergy by cardiac ultrasound. Am J Cardiol 2000;86:4G 9G. 18. Beyar R, Shapiro EP, Graves WL, et al. Quantification and validation of left ventricular wall thickening by a three-dimensional volume element magnetic resonance imaging approach. Circulation 1990;81: