In recent years, novel ultrasound functional imaging (UFI) techniques have been introduced to assess cardiac function by measuring, e.g. cardiac output (CO) and/or myocardial strain. Verification and reproducibility assessment in a realistic setting remain major issues. Simulations and phantoms are often unrealistic, whereas in vivo measurements often lack crucial hemodynamic parameters or ground truth data, or suffer from the large physiological and clinical variation between patients when attempting clinical validation. Controlled validation in certain pathologies is cumbersome and often requires the use of lab animals. In this study, an isolated beating pig heart setup was adapted and used for performance assessment of UFI techniques such as volume assessment and ultrasound strain imaging. The potential of performing verification and reproducibility studies was demonstrated. For proof-of-principle, validation of UFI in pathological hearts was examined. Ex vivo porcine hearts (n  =  6, slaughterhouse waste) were resuscitated and attached to a mock circulatory system. Radio frequency ultrasound data of the left ventricle were acquired in five short axis views and one long axis view. Based on these slices, the CO was measured, where verification was performed using flow sensor measurements in the aorta. Strain imaging was performed providing radial, circumferential and longitudinal strain to assess reproducibility and inter-subject variability under steady conditions. Finally, strains in healthy hearts were compared to a heart with an implanted left ventricular assist device, simulating a failing, supported heart. Good agreement between ultrasound and flow sensor based CO measurements was found. Strains were highly reproducible (intraclass correlation coefficients >0.8). Differences were found due to biological variation and condition of the hearts. Strain magnitude and patterns in the assisted heart were available for different pump action, revealing large changes compared to the normal condition. The setup provides a valuable benchmarking platform for UFI techniques. Future studies will include work on different pathologies and other means of measurement verification.

No abstract

Recent studies have shown the efficacy of myocardial strain estimated using speckle tracking echocardiography (STE) in predicting response to cardiac resynchronisation therapy. This study focuses on circumferential strain patterns, comparing STE-acquired strains to tagged-magnetic resonance imaging (MRI-T). Second, the effect of regularisation was examined. Two-dimensional parasternal ultrasound (US) and MRI-T data were acquired in the left ventricular short-axis view of canines before (n = 8) and after (n = 9) left bunch branch block (LBBB) induction. US-based strain analysis was performed on Digital Imaging and Communications in Medicine data at the mid-level using three overall methods ("Commercial software," "Basic block-matching," "regularised block-matching"). Moreover, three regularisation approaches were implemented and compared. MRI-T analysis was performed using SinMod. Normalised regional circumferential strain curves, based on standard six or septal/lateral segments, were analysed and cross-correlated with MRI-T data. Systolic strain (SS) and septal rebound stretch (SRS) were calculated and compared. Overall agreement of normalised circumferential strain was good between all methods on a global and regional level. All STE methods showed a bias (4% strain) toward higher SS estimates. Pre-LBBB, septal and lateral segment correlation was excellent between the Basic (mean r = 0.96) and regularised (mean r = 0.97) methods and MRI-T. The Commercial method showed a significant discrepancy between the two walls (septal r = 0.94, lateral r = 0.68). Correlation with MRI-T reduced between pre-and post-LBBB (Commercial r = 0.79, Basic r = 0.82, mean regularised r = 0.86). Septal strain patterns and SRS varied with the STE software and type of regularisation, with all STE methods estimating nonzero SRS values pre-LBBB. Absolute values showed moderate agreement, with a bias for higher strain from STE. SRS varied with the type of software and extra regularisation applied. Open efforts are needed to understand the underlying causes of differences between STE methods before standardisation can be achieved. This is particularly important given the apparent clinical value of strain-based parameters such as SRS.

Background: Exercise stress echocardiography is clinically used to assess cardiovascular diseases. For accurate cardiac evaluation, a stable field-of-view is required. However, transducer orientation and position are difficult to preserve. Hands-free acquisitions might provide more consistent and reproducible results. In this study, the field-of-view stability and variability of hands-free acquisitions are objectively quantified in a comparison with manually obtained images, based on image structural and feature similarities. In addition, the feasibility and consistency of hands-free strain imaging is assessed. Methods: In twelve healthy males, apical and parasternal images were acquired hands-free, using a fixation device, and manually, during semi-supine exercise sessions. In the final ten seconds of every exercise period, the image structural similarity and cardiac feature consistency were computed using a steerable pyramid employing complex, oriented wavelets. An algorithm discarding images displaying lung artifacts was created. Hands-free strain consistency was analyzed. Results: Hands-free acquisitions were possible in 9 of the 12 subjects, whereas manually 10 out of 12 could be imaged. The image structural similarity was significantly improved in the hands-free apical window acquisitions (0.91 versus 0.82), and at least equally good in the parasternal window (0.90 versus 0.82). The change in curvature and orientation of the interventricular septum also appeared to be lower in the hands-free acquisitions. The variability in field-of-view was similar in both acquisitions. Longitudinal, septal strain was shown to be at least as consistent when obtained hands-free compared to manual acquisitions. Conclusions: The field-of-view was shown to be more or equally stable and consistent in the hands-free data in comparison to manually obtained images. The variability was similar, thus respiration-and exercise-induced motions were comparable for manual and hands-free acquisitions. Additionally, the feasibility of hands-free strain has been demonstrated. Furthermore, the results suggest the hands-free measurements to be more reproducible, though further analysis is required.

Lightweight segmentation models are becoming more popular for fast diagnosis on small and low cost medical imaging devices. This study focuses on the segmentation of the left ventricle (LV) in cardiac ultrasound (US) images. A new lightweight model [LV network (LVNet)] is proposed for segmentation, which gives the benefits of requiring fewer parameters but with improved segmentation performance in terms of Dice score (DS). The proposed model is compared with state-of-the-art methods, such as UNet, MiniNetV2, and fully convolutional dense dilated network (FCdDN). The model proposed comes with a postprocessing pipeline that further enhances the segmentation results. In general, the training is done directly using the segmentation mask as the output and the US image as the input of the model. A new strategy for segmentation is also introduced in addition to the direct training method used. Compared with the UNet model, an improvement in DS performance as high as 5% for segmentation with papillary (WP) muscles was found, while showcasing an improvement of 18.5% when the papillary muscles are excluded. The model proposed requires only 5% of the memory required by a UNet model. LVNet achieves a better trade-off between the number of parameters and its segmentation performance as compared with other conventional models. The developed codes are available at ht. tps://github.com/navchetanawasthi/Left_Ventricle_Segmentation.