Limits...
Validation and Development of a New Automatic Algorithm for Time-Resolved Segmentation of the Left Ventricle in Magnetic Resonance Imaging.

Tufvesson J, Hedström E, Steding-Ehrenborg K, Carlsson M, Arheden H, Heiberg E - Biomed Res Int (2015)

Bottom Line: Ninety subjects imaged with a cine balanced steady state free precession sequence were included in the study (training set n = 40, test set n = 50).Manual delineation was reference standard and second observer analysis was performed in a subset (n = 25).The mean differences between automatic segmentation and manual delineation were EDV -11 mL, ESV 1 mL, EF -3%, and LVM 4 g in the test set.

View Article: PubMed Central - PubMed

Affiliation: Department of Clinical Physiology, Lund University Hospital, Lund University, 221 85 Lund, Sweden ; Department of Numerical Analysis, Centre for Mathematical Sciences, Faculty of Engineering, Lund University, 221 00 Lund, Sweden.

ABSTRACT

Introduction: Manual delineation of the left ventricle is clinical standard for quantification of cardiovascular magnetic resonance images despite being time consuming and observer dependent. Previous automatic methods generally do not account for one major contributor to stroke volume, the long-axis motion. Therefore, the aim of this study was to develop and validate an automatic algorithm for time-resolved segmentation covering the whole left ventricle, including basal slices affected by long-axis motion.

Methods: Ninety subjects imaged with a cine balanced steady state free precession sequence were included in the study (training set n = 40, test set n = 50). Manual delineation was reference standard and second observer analysis was performed in a subset (n = 25). The automatic algorithm uses deformable model with expectation-maximization, followed by automatic removal of papillary muscles and detection of the outflow tract.

Results: The mean differences between automatic segmentation and manual delineation were EDV -11 mL, ESV 1 mL, EF -3%, and LVM 4 g in the test set.

Conclusions: The automatic LV segmentation algorithm reached accuracy comparable to interobserver for manual delineation, thereby bringing automatic segmentation one step closer to clinical routine. The algorithm and all images with manual delineations are available for benchmarking.

No MeSH data available.


Related in: MedlinePlus

Example of segmentation in end-diastole and end-systole. An example of automatic segmentation is shown in end-diastole (a) and end-systole (b). Each panel shows the short axis stack covering the left ventricle from base to apex with endocardial (red) and epicardial (green) segmentations. Note how the outflow tract has moved out of the two most basal slices in end-systole (b, images marked ∗), compared to end-diastole (a) and that the algorithm has automatically corrected for this long-axis motion.
© Copyright Policy - open-access
Related In: Results  -  Collection


getmorefigures.php?uid=PMC4491381&req=5

fig4: Example of segmentation in end-diastole and end-systole. An example of automatic segmentation is shown in end-diastole (a) and end-systole (b). Each panel shows the short axis stack covering the left ventricle from base to apex with endocardial (red) and epicardial (green) segmentations. Note how the outflow tract has moved out of the two most basal slices in end-systole (b, images marked ∗), compared to end-diastole (a) and that the algorithm has automatically corrected for this long-axis motion.

Mentions: Automatic segmentation was performed and compared to manual delineation in the test and compared to interobserver variability in a second observer subset. In one patient the automatic segmentation failed due to a severe bright fold-in artifact connecting the right and left ventricle. This patient was excluded from further analysis resulting in a test set of 49 patients and a second observer subset of 24 patients. Figure 4 shows an example of automatic segmentation in all slices in end-diastole and end-systole. A comparison between automatic segmentation and manual delineation can be seen in Figure 5 for a basal, midventricular, and apical slice in end-diastole and end-systole. In the additional file a time-resolved 3D-rendering of left ventricle shows the long-axis motion of the epicardial surface resulting from the automatic segmentation algorithm. The differences between automatic segmentation and manual delineation for clinical parameters were EDV −11 ± 11 mL (R = 0.96), ESV 1 ± 10 mL (R = 0.95), EF −3 ± 4% (R = 0.86), LVM 4 ± 15 g (R = 0.87), SV −12 ± 8 mL (R = 0.92), and CO −0.7 ± 0.5 L/min (R = 0.94) (Table 1, Figures 6 and 7). The image processing error measurements were for endocardial segmentation DSC = 0.91 ± 0.03 and P2C = 2.1 ± 0.5 mm and for epicardial segmentation DSC = 0.93 ± 0.02 and P2C = 2.1 ± 0.5 mm as mean over all slices and both end-diastole and end-systole (Table 2). End-diastolic image processing error measurements performed better than end-systolic (Table 2). Midventricular slices performed better than basal and apical slices (Table 3).


Validation and Development of a New Automatic Algorithm for Time-Resolved Segmentation of the Left Ventricle in Magnetic Resonance Imaging.

Tufvesson J, Hedström E, Steding-Ehrenborg K, Carlsson M, Arheden H, Heiberg E - Biomed Res Int (2015)

Example of segmentation in end-diastole and end-systole. An example of automatic segmentation is shown in end-diastole (a) and end-systole (b). Each panel shows the short axis stack covering the left ventricle from base to apex with endocardial (red) and epicardial (green) segmentations. Note how the outflow tract has moved out of the two most basal slices in end-systole (b, images marked ∗), compared to end-diastole (a) and that the algorithm has automatically corrected for this long-axis motion.
© Copyright Policy - open-access
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC4491381&req=5

fig4: Example of segmentation in end-diastole and end-systole. An example of automatic segmentation is shown in end-diastole (a) and end-systole (b). Each panel shows the short axis stack covering the left ventricle from base to apex with endocardial (red) and epicardial (green) segmentations. Note how the outflow tract has moved out of the two most basal slices in end-systole (b, images marked ∗), compared to end-diastole (a) and that the algorithm has automatically corrected for this long-axis motion.
Mentions: Automatic segmentation was performed and compared to manual delineation in the test and compared to interobserver variability in a second observer subset. In one patient the automatic segmentation failed due to a severe bright fold-in artifact connecting the right and left ventricle. This patient was excluded from further analysis resulting in a test set of 49 patients and a second observer subset of 24 patients. Figure 4 shows an example of automatic segmentation in all slices in end-diastole and end-systole. A comparison between automatic segmentation and manual delineation can be seen in Figure 5 for a basal, midventricular, and apical slice in end-diastole and end-systole. In the additional file a time-resolved 3D-rendering of left ventricle shows the long-axis motion of the epicardial surface resulting from the automatic segmentation algorithm. The differences between automatic segmentation and manual delineation for clinical parameters were EDV −11 ± 11 mL (R = 0.96), ESV 1 ± 10 mL (R = 0.95), EF −3 ± 4% (R = 0.86), LVM 4 ± 15 g (R = 0.87), SV −12 ± 8 mL (R = 0.92), and CO −0.7 ± 0.5 L/min (R = 0.94) (Table 1, Figures 6 and 7). The image processing error measurements were for endocardial segmentation DSC = 0.91 ± 0.03 and P2C = 2.1 ± 0.5 mm and for epicardial segmentation DSC = 0.93 ± 0.02 and P2C = 2.1 ± 0.5 mm as mean over all slices and both end-diastole and end-systole (Table 2). End-diastolic image processing error measurements performed better than end-systolic (Table 2). Midventricular slices performed better than basal and apical slices (Table 3).

Bottom Line: Ninety subjects imaged with a cine balanced steady state free precession sequence were included in the study (training set n = 40, test set n = 50).Manual delineation was reference standard and second observer analysis was performed in a subset (n = 25).The mean differences between automatic segmentation and manual delineation were EDV -11 mL, ESV 1 mL, EF -3%, and LVM 4 g in the test set.

View Article: PubMed Central - PubMed

Affiliation: Department of Clinical Physiology, Lund University Hospital, Lund University, 221 85 Lund, Sweden ; Department of Numerical Analysis, Centre for Mathematical Sciences, Faculty of Engineering, Lund University, 221 00 Lund, Sweden.

ABSTRACT

Introduction: Manual delineation of the left ventricle is clinical standard for quantification of cardiovascular magnetic resonance images despite being time consuming and observer dependent. Previous automatic methods generally do not account for one major contributor to stroke volume, the long-axis motion. Therefore, the aim of this study was to develop and validate an automatic algorithm for time-resolved segmentation covering the whole left ventricle, including basal slices affected by long-axis motion.

Methods: Ninety subjects imaged with a cine balanced steady state free precession sequence were included in the study (training set n = 40, test set n = 50). Manual delineation was reference standard and second observer analysis was performed in a subset (n = 25). The automatic algorithm uses deformable model with expectation-maximization, followed by automatic removal of papillary muscles and detection of the outflow tract.

Results: The mean differences between automatic segmentation and manual delineation were EDV -11 mL, ESV 1 mL, EF -3%, and LVM 4 g in the test set.

Conclusions: The automatic LV segmentation algorithm reached accuracy comparable to interobserver for manual delineation, thereby bringing automatic segmentation one step closer to clinical routine. The algorithm and all images with manual delineations are available for benchmarking.

No MeSH data available.


Related in: MedlinePlus