Comparison of Manual, Semi- and Fully Automated Heart Segmentation for Assessing Global Left Ventricular Function in Multidetector Computed Tomography

Purpose:To evaluate the reliability of global left ventricular (LV) function and mass measurements with the aid of a semi-automated (Circulation; Siemens, Forchheim, Germany) and a new fully automated software (Philips Research Europe, Aachen, Germany) versus an established manual segmentation method (Argus; Siemens). Material and Methods:Forty-one patients (31 men, 10 women; mean age: 62 ± 5 years) with known or suspected coronary heart disease underwent contrast-enhanced Dual-Source computed tomography of the heart (120 kV, 410 mAs/rotation, collimation 2 × 32 × 0.6 mm, gantry rotation time 0.33 milliseconds). Global LV function measurements of end-diastolic volume (EDV), end-systolic volume (ESV), stroke volume, ejection fraction (EF), and LV mass were each assessed with a manual, a semi- and fully automated method. The latter were compared with the manual contour tracing method, which was considered as standard of reference. Postprocessing time for each method was recorded. For statistical analysis, repeated-measures analysis of variance, post hoc t test, and concordance correlation coefficients were calculated. Bland-Altman plots were generated. Results:In general, ESV and EF assessed with the semi-automated and with the fully automated prototype version agreed well with the manual contour tracing method. The mean ESV (±SD) calculated from the manual, the semi-automated, and the fully automated method was 67 ± 43 mL, 74 ± 54 mL, and 75 ± 48 mL, respectively. No statistically significant differences between the methods were found for ESV and EF. In contrast, significant variations (P < 0.05) among the different segmentation methods were shown for EDV, stroke volume, and LV mass. This variation was predominantly due to variation in endocardial delineations among the different techniques. Concordance correlation coefficients demonstrated a better accuracy for the fully automated method than for the semi-automated technique when compared with the manual drawing method. Furthermore, fully automated postprocessing heart segmentation yielded time savings of approximately 80% compared with the manual segmentation tool and 63% compared with the semi-automated technique. Mean postprocessing time (±SD) for the manual, the semi-automated, and the fully automated method was 345 ± 75 seconds, 192 ± 58 seconds, and 72 ± 58 seconds, respectively. Conclusion:LV function and mass analyses using semi- or fully automated segmentation algorithms are feasible even if significant differences in EDV assessment are observed. The fully automated method results in better accuracy and time savings when compared with manual and semi-automated data analysis.

[1]  Georg Mühlenbruch,et al.  Automated vs. manual assessment of left ventricular function in cardiac multidetector row computed tomography: comparison with magnetic resonance imaging , 2006, European Radiology.

[2]  Olivier Ecabert,et al.  Segmentation of the heart and major vascular structures in cardiovascular CT images , 2008, SPIE Medical Imaging.

[3]  R M Whitlock,et al.  Left ventricular end-systolic volume as the major determinant of survival after recovery from myocardial infarction. , 1987, Circulation.

[4]  Marco Das,et al.  Global left ventricular function in cardiac CT. Evaluation of an automated 3D region-growing segmentation algorithm , 2006, European Radiology.

[5]  B. Marinček,et al.  Einsatz der Mehrschicht-Spiral-CT beim traumatologischen Notfall und ihre Auswirkungen auf den Untersuchungs- und Behandlungsalgorithmus , 2004 .

[6]  P. Wielopolski,et al.  Addition of the Long-Axis Information to Short-Axis Contours Reduces Interstudy Variability of Left-Ventricular Analysis in Cardiac Magnetic Resonance Studies , 2008, Investigative radiology.

[7]  Olivier Ecabert,et al.  Automatic Whole Heart Segmentation in Static Magnetic Resonance Image Volumes , 2007, MICCAI.

[8]  J G Goldin,et al.  Left ventricular ejection fraction using 64-slice CT coronary angiography and new evaluation software: initial experience. , 2008, The British journal of radiology.

[9]  Gabriel P. Krestin,et al.  High-Resolution Spiral Computed Tomography Coronary Angiography in Patients Referred for Diagnostic Conventional Coronary Angiography , 2005, Circulation.

[10]  Gorka Bastarrika,et al.  Quantification of left ventricular function and mass in cardiac Dual-Source CT (DSCT) exams: comparison of manual and semiautomatic segmentation algorithms , 2008, European Radiology.

[11]  S Wildermuth,et al.  Time-effectiveness, Observer-dependence, and Accuracy of Measurements of Left Ventricular Ejection Fraction Using 4-channel MDCT , 2004, RoFo : Fortschritte auf dem Gebiete der Rontgenstrahlen und der Nuklearmedizin.

[12]  Marco Das,et al.  Evaluation of aortocoronary bypass stents with cardiac MDCT compared with conventional catheter angiography. , 2007, AJR. American journal of roentgenology.

[13]  Victor Mor-Avi,et al.  Multidetector computed tomography evaluation of left ventricular volumes: sources of error and guidelines for their minimization. , 2008, Journal of cardiovascular computed tomography.

[14]  Herbert Bruder,et al.  Dual-Source Computed Tomography for Assessing Cardiac Function: A Phantom Study , 2007, Investigative radiology.

[15]  Jonathan S. Lewin,et al.  Functional Cardiac CT and MR: Effects of Heart Rate and Software Applications on Measurement Validity , 2005, Journal of thoracic imaging.

[16]  Roman Fischbach,et al.  Left ventricular function studied with MDCT , 2006, European Radiology.

[17]  Borut Marincek,et al.  Left Ventricular and Left Atrial Dimensions and Volumes: Comparison Between Dual-Source CT and Echocardiography , 2008, Investigative radiology.

[18]  M. Reiser,et al.  Clinical value of MDCT in the diagnosis of coronary artery disease in patients with a low pretest likelihood of significant disease. , 2006, AJR. American journal of roentgenology.

[19]  Thomas Schmitz-Rode,et al.  Assessment of Global Left and Right Ventricular Function Using Dual-Source Computed Tomography (DSCT) in Comparison to MRI: An Experimental Study in a Porcine Model , 2007, Investigative radiology.

[20]  Matthijs Oudkerk,et al.  Evaluation of global left ventricular function assessment by dual-source computed tomography compared with MRI , 2009, European Radiology.

[21]  Stephen Schroeder,et al.  Diagnostic accuracy of noninvasive coronary imaging using 16-detector slice spiral computed tomography with 188 ms temporal resolution. , 2005, Journal of the American College of Cardiology.

[22]  K. Stierstorfer,et al.  First performance evaluation of a dual-source CT (DSCT) system , 2006, European Radiology.

[23]  Dana H. Ballard,et al.  Generalizing the Hough transform to detect arbitrary shapes , 1981, Pattern Recognit..

[24]  M. F. Reiser,et al.  Quantitative assessment of left ventricular function with dual-source CT in comparison to cardiac magnetic resonance imaging: initial findings , 2008, European Radiology.