A. Stork

Automatic Image-Driven Segmentation of the Ventricles in Cardiac Cine MRI

To propose and to evaluate a novel method for the automatic segmentation of the hearts two ventricles from dynamic (“cine”) short‐axis “steady state free precession” (SSFP) MR images. This segmentation task is of significant clinical importance. Previously published automated methods have various disadvantages for routine clinical use.

Multicontrast-weighted magnetic resonance imaging of atherosclerotic plaques at 3.0 and 1.5 Tesla: ex-vivo comparison with histopathologic correlation

The purpose was to analyze magnetic resonance (MR) plaque imaging at 3.0 Tesla and 1.5 Tesla in correlation with histopathology. MR imaging (MRI) of the abdominal aorta and femoral artery was performed on seven corpses using T1-weighted, T2-weighted, and PD-weighted sequences at 3.0 and 1.5 Tesla. Cross-sectional images at the branching of the inferior mesenteric artery and the profunda femoris were rated with respect to image quality. Corresponding cross sections of the imaged vessels were obtained at autopsy. The atherosclerotic plaques in the histological slides and MR images were classified according to the American Heart Association (AHA) and analyzed for differences. MRI at 3.0 Tesla offered superior depiction of arterial wall composition in all contrast weightings, rated best for T2-weighted images. Comparing for field strength, the highest differences were observed in T1-weighted and T2-weighted techniques (both P≤0.001), with still significant differences in PD-weighted sequence (P≤0.005). The majority of plaques were histologically classified as calcified plaques. In up to 21% of the cases, MRI at both field strengths detected signal loss characteristic of calcification although calcified plaque was absent in histology. MRI at 3.0 Tesla offers superior plaque imaging quality compared with 1.5 Tesla, but further work is necessary to determine whether this translates in superior diagnostic accuracy.

Assessment of functional anatomy of the mitral valve in patients with mitral regurgitation with cine magnetic resonance imaging: comparison with transesophageal echocardiography and surgical results

The ability of magnetic resonance imaging (MRI) to accurately define the functional anatomy of mitral regurgitation was assessed. Transesophageal echocardiography (TEE) and cine MRI were performed on 43 patients with mitral regurgitation and were compared for the jet number, location, direction and presence of a prolapse (atrial displacement, malapposition or a flail). In 36 patients, diagnostic accuracy in reference to surgery was assessed. Comparing TEE and MRI the jet number and location were judged in concordance in 86% of patients. Jet location did not show a significant difference (Wilcoxon: P = 0.66) and both modalities correlated strongly (Spearman: r = 0.68, P<0.0001). Jet direction was judged with high concordance (kappa=0.63). Additionally, prolapse evaluation showed high concordance (kappa: valve, 0.63; anterior mitral leaflet, 0.70; posterior mitral leaflet, 0.73). Compared with surgery, the sensitivity for the detection of malapposition of any leaflet or one of both leaflets ranged between 75% and 93% for TEE and 71% and 89% for MRI. Specificities ranged between 88 and 96% for TEE and 88 and 100% for MRI. TEE detected torn chordae in all ten patients, six of which were missed by MRI. MRI is comparable with TEE in prolapse and jet evaluation. MRI is inferior to TEE in depicting anatomical details such as torn chordae.

Magnetic resonance imaging of the coronary vessel wall at 3 T using an obliquely oriented reinversion slab with adiabatic pulses

Three‐dimensional methods offer volumetric coverage in coronary vessel wall imaging, in addition to high signal‐to‐noise ratios (SNR). To increase SNR further, it is desirable to implement such 3D methods at 3 T. At this field strength, the pulse sequence must be robust to main field and RF inhomogeneities. To achieve this, the double inversion‐recovery (DIR) preparation was adapted to use adiabatic pulses, with a slab‐selective reinversion replacing the previously used 2D pencil‐beam. The slab was oriented obliquely, in order to avoid upstream blood (e.g., left ventricle) or the navigator beam. Phantom experiments suggest that at 3 T, this approach improves both the net profile of the DIR pulse pair and the restoration of magnetization in the navigator region. Using this method, the feasibility of 3D coronary vessel wall imaging was demonstrated at 3 T. Fourteen healthy subjects were scanned using a segmented gradient‐echo sequence with prospective navigator gating. Good‐quality images of left and right coronary arteries were obtained, with SNR values of 29.7 ± 7.5 (vessel wall); 10.5 ± 4.4 (blood); 14.3 ± 5.2 (fat); and 45.6 ± 18.0 (myocardium). No problems occurred with ECG‐gating or power deposition (SAR) limits. Magn Reson Med, 2005.

Segmentation of Medical Images with a Shape and Motion Model: A Bayesian Perspective

This paper describes a Bayesian framework for the segmentation of a temporal sequence of medical images, where both shape and motion prior information are integrated into a stochastic model. With this approach, we aim to take into account all the information available to compute an optimum solution, thus increasing the robustness and accuracy of the shape and motion reconstruction. The segmentation algorithm we develop is based on sequential Monte Carlo sampling methods previously applied in tracking applications. Moreover, we show how stochastic shape models can be constructed using a global shape description based on orthonormal functions. This makes our approach independent of the dimension of the object (2D or 3D) and on the particular shape parameterization used. Results of the segmentation method applied to cardiac cine MR images are presented.

Analysis of the Left Ventricle After Myocardial Infarction Combining 4D Cine-MR and 3D DE-MR Image Sequences

Spatial-temporal MR image sequences of the heart contain information about shape and motion changes and pathological structures after myocardial infarction. In this paper a software system called HeAT for the quantitative analysis of 4D MR image sequences of infarct patients is presented. HeAT supports interactive segmentation of anatomical and pathological structures. Registration of Cine- and DE-MR image data is applied to enable their combined evaluation during the analysis process. Partitioning of the myocardium in segments enables the analysis with high local resolution. Corresponding segments are generated and used for inter/intra patient comparison. Quantitative parameters were extracted and visualized. Parameters like endocard movement in the infarcted area of 6 infarct patients were computed in HeAT. Parameters in the infarct area show the expected dysfunctional characteristics. Based on theses parameters passive endocardial movement and myocardial areas with decreased contraction were identified.

Assessment of left ventricular function parameters with a new three-dimensional shape model.

PURPOSE To evaluate a 3D model of the left ventricle (LV) which allows calculation of LV function parameters on the basis of both short axis (SA) and long axis (LA) cine acquisitions. Comparison with the conventional Simpsons rule method in a volunteer and patient collective. MATERIALS AND METHODS Cine imaging was performed with a prospectively triggered SSFP sequence: trueFISP: TR 3.6 msec, TE 1.8 msec, bFFE: TR 3.0 msec, TE 1.4 msec, flip angle 60 degrees , resolution 1.37 x 1.37 mm, slice thickness 8 mm, gap 2 mm in SA orientation from apex to basis and in radial LA orientation (spacing 15 degrees) in 11 volunteers and 27 patients with mitral valve insufficiency. Five different volume computations were compared: Simpsons rule based on all SA slices (M0), 3D shape model based on all SA slices (M1a), 3D shape model based on 3 SA slices (M1b), 3D shape model based on all SA and LA slices (M2a), and 3D shape model based on 3 SA slices and 1 LA slice (M2b). RESULTS M 0 and M 1a give similar results (r: 0.99, b: 0.98). M 2a produces larger volumes than M 0 (b: 0.85) due to the inclusion of the LA contours. M 1b effectively reproduces the volumes computed with M 0 (r: 0.99, b: 1.02). M 2b effectively reproduces the volumes computed with M 2a (r: 0.99, b: 0.94). M 2b and M 0 give similar results in the patient collective (r: 0.99, b: 0.97). CONCLUSION The proposed 3D shape model allows merging of information acquired in different orientations and thus the combination of SA and LA contours with better coverage of the left ventricle. It provides a suitable fit with a reduced number of segmented contours.

1120 Detection of mechanical ventricular asynchrony by cine-MRI

Methods 32 patients with (n = 17) and without (n = 15) LBBB in surface ECG underwent HTRC-MRI at 1.5 Tesla. For a temporal resolution of 6.3 ms, a shared-phases trueFISP sequence and parallel imaging was used. Time to peak contraction (TPC) was defined as the interval between Rwave and the peak of systolic contraction. Interand intraventricular delays in HTRC-MRI were calculated as the difference of TPC of lateral LVto lateral RV-wall and lateral to septal LV-wall, respectively. 1) Analysis of interand intraobserver agreement for the assessment of HTRC-MRI delays was performed. 2) HTRC-MRI delays were compared between patients with and without LBBB. 3) HTRCMRI delays were correlated with standard echocardiographic parameters in patients with LBBB.

Verwendung eines neuen 3-D-Oberflächenmodells zur linksventrikulären Volumenbestimmung mittels Kurzachsen-Cine-MRT und rotierter Längsachse-Cine-MRT

Ziele: Berechnung linksventrikularer Volumina unter Berucksichtigung der kurzen Achsen (KA) und langen Achsen (LA) mit einem neuen 3-D-Oberflachenmodell. Vergleich mit konventioneller Berechnung durch Simpsons Regel bei gesunden Probanden und Patienten mit Mitralklappeninsuffizienz. Methode: Elf gesunden Probanden (mittleres Alter: 27,5 Jahre SD ±3,6) und 27 Patienten (54,4 Jahre, SD ±12.6) mit echokardiographisch gesicherter Mitralklappeninsuffizienz wurden mittels CineTrueFisp Sequenzen (TR: 3,6, TE: 1,8, Flipwinkel: 60°, FOV: 350×306mm, matrix: 256×139, Schichtdicke: 8mm, Schichtabstand 2mm) wurden in der KA von Apex bis Basis sowie in der langen Achse (15° Rotationsintervall) bei 1,5 Tesla (Achieva, Philips, Niederlande; Symphony, Siemens, Deutschland) untersucht. Epi- und Endokard wurden in den LA und KA manuell in der enddiastolischen und -systolischen Phase konturiert. Anhand der gespeicherten Konturen wurden die epi- und endokardialen Oberflachen mit dem vorgestellten 3D-Modell angenahert. Vier unterschiedliche Volumenberechnungen wurden verglichen: Simpsons Regel auf der Basis aller SA (M0), 3D-Modell auf der Basis aller SA (M1), auf der Basis aller SA und LA (M2) und auf der Basis von 3 SA and 1 LA (M3). Ergebnis: M0 und M1 ergeben vergleichbare Ergebnisse (r: 0,98, b: 0,96). M2 resultiert in groseren gemessenen Volumina als M0 (r: 0,98, b: 0,85) aufgrund der Berucksichtigung der LA. M3 resultiert trotz reduzierter Schichtanzahl in vergleichbaren Ergebnissen wie M0 bei Probanden (r: 0,99, b: 0,94) und Patienten (r: 0,98, b: 0,95). Schlussfolgerung: Das vorgestellte 3D-Oberflachenmodell ermoglicht die Zusammenfassung von Bildinformation aus kurzen und langen Achsen und somit verbesserter Abbildung des linken Ventrikels. Selbst mit einer reduzierten Zahl von Schichten wird so eine genaue linksventrikulare Volumenbestimmung ermoglicht. Korrespondierender Autor: Bansmann PM Universitatsklinikum Hamburg-Eppendorf, Klinik fur Diagnostische und Interventionelle Radiologie, Martinistrasse 52, 20246 Hamburg E-Mail: pbansmann@uke.uni-hamburg.de