Automated segmentation of 3D echocardiographic images in patients with congenital heart disease is challenging, because the boundary between blood and cardiac tissue is poorly defined in some regions. Cardiologists mentally incorporate movement of the heart, using temporal coherence of structures to resolve ambiguities. Therefore, we investigated the merit of temporal cross-correlation for automated segmentation over the entire cardiac cycle. Optimal settings for maximum cross-correlation (MCC) calculation, based on a 3D cross-correlation based displacement estimation algorithm, were determined to obtain the best contrast between blood and myocardial tissue over the entire cardiac cycle. Resulting envelope-based as well as RF-based MCC values were used as additional external force in a deformable model approach, to segment the left-ventricular cavity in entire systolic phase. MCC values were tested against, and combined with, adaptive filtered, demodulated RF-data. Segmentation results were compared with manually segmented volumes using a 3D Dice Similarity Index (3DSI). Results in 3D pediatric echocardiographic images sequences (n = 4) demonstrate that incorporation of temporal information improves segmentation. The use of MCC values, either alone or in combination with adaptive filtered, demodulated RF-data, resulted in an increase of the 3DSI in 75% of the cases (average 3DSI increase: 0.71 to 0.82). Results might be further improved by optimizing MCC-contrast locally, in regions with low blood-tissue contrast. Reducing underestimation of the endocardial volume due to MCC processing scheme (choice of window size) and consequential border-misalignment, could also lead to more accurate segmentations. Furthermore, increasing the frame rate will also increase MCC-contrast and thus improve segmentation.
Segmentation of the heart muscle in 3D echocardiographic images provides a tool for visualization of cardiac anatomy and assessment of heart function, and serves as an important pre-processing step for cardiac strain imaging. By incorporating spatial and temporal information of 3D ultrasound image sequences (4D), a fully automated method using image statistics was developed to perform 3D segmentation of the heart muscle. 3D rf-data were acquired with a Philips SONOS 7500 live 3D ultrasound system, and an X4 matrix array transducer (2-4 MHz). Left ventricular images of five healthy children were taken in transthoracial short/long axis view. As a first step, image statistics of blood and heart muscle were investigated. Next, based on these statistics, an adaptive mean squares filter was selected and applied to the images. Window size was related to speckle size (5x2 speckles). The degree of adaptive filtering was automatically steered by the local homogeneity of tissue. As a result, discrimination of heart muscle and blood was optimized, while sharpness of edges was preserved. After this pre-processing stage, homomorphic filtering and automatic thresholding were performed to obtain the inner borders of the heart muscle. Finally, a deformable contour algorithm was used to yield a closed contour of the left ventricular cavity in each elevational plane. Each contour was optimized using contours of the surrounding planes (spatial and temporal) as limiting condition to ensure spatial and temporal continuity.
Better segmentation of the ventricle was obtained using 4D information than using information of each plane separately.
The feasibility of echographic imaging of the tissues in healthy lip and in reconstructed cleft lip and estimating the dimensions and the normalized echo level of these tissues is investigated. Echographic images of the upper lip were made with commercial medical ultrasound equipment, using a linear array transducer (7-11 MHz bandwidth) and a non-contact gel coupling. Tissue dimensions were measured by means of software calipers. Echo levels were calibrated and corrected for beam characteristics, gel path and tissue attenuation by using a tissue-mimicking phantom. At central position of philtrum, mean thickness (and standard deviation) of lip loose connective tissue layer, orbicularis oris muscle and dense connective layer was 4.0 (sd 0.1) mm, 2.3 (sd 0.7) mm, 2.2 (sd 0.7) mm, respectively, in healthy lip at rest. Mean (sd) echo level of muscle and dense connective tissue layer with respect to echo level of lip loose connective tissue layer was in relaxed condition: - 19.3 (sd 0.6) dB and - 10.7 (sd 4.0) dB, respectively. Echo level of loose connective tissue layer was +25.6 (sd 4.2) dB relative to phantom echo level obtained in the focus of the transducer. Color mode echo images were calculated, after adaptive filtering of the images, which show the tissues in separate colors and highlight the details of healthy lip and reconstructed cleft lip. Quantitative assessment of thickness and echo level of various lip tissues is feasible after proper calibration of the echographic equipment. Diagnostic potentials of the developed quantitative echographic techniques for non-invasive evaluation of the outcome of cleft lip reconstruction are promising.