In the past few decades, great progress has been made in prenatal diagnosis of congenital heart disease (CHD). Fetal echocardiography is recognized as the main prenatal screening and diagnostic tool that can accurately detect approximately 85 % of fetal cardiac abnormalities. Evaluation of the fetal heart remains a major challenge in prenatal ultrasound screening and diagnosis due to fetal position, involuntary movement, small and complex fetal cardiac anatomy, maternal abdominal wall conditions, and lack of expertise in fetal echocardiography by some physicians engaged in obstetric ultrasound. Artificial intelligence (AI) can automate and standardize the display of each diagnostic section of the fetal heart and thus contribute to accurate diagnosis, which significantly optimizes the clinical application of fetal echocardiography. In this review, we not only clarify the role of AI but also highlight its significance and future solutions in the field of fetal echocardiography.
Background: Left atrial (LA) and left atrial appendage (LAA) dysfunction has been demonstrated to contribute to atrial fibrillation (AF)-related stroke. However, usefulness of LA and LAA mechanics has not been fully compared. We sought to investigate the association of LA and LAA mechanics with stroke and to compare their diagnostic values in the risk stratification of stroke in patients with nonvalvular AF. Methods: A total of 208 consecutive patients with AF (63.58 ± 10.37 years, 63.9% male,57.7% persistent AF) who underwent echocardiography before catheter ablation were prospectively enrolled. Speckle-tracking was used to measure LA and LAA global longitudinal strain (GLS). LA and LAA mechanical dispersions (MD) were defined as the standard deviation (SD) of time to peak positive strain corrected by the R-R interval. Results: Patients with prior stroke/ transient ischemic attack (TIA) (n = 31) had significantly higher LA and LAA MD than those without (n = 177) (11.56 ± 4.38% vs. 8.43 ± 3.44%, 15.15 ± 5.46% vs. 10.94 ± 4.40%, both P < 0.01). In multivariable analysis, LA and LAA MD were independently associated with stroke/TIA (odds ratio, 1.18–1.29, 1.19–1.22, respectively, both P < 0.01), providing incremental values over clinical and standard echocardiographic parameters. In a subgroup analysis, LA MD was more useful than LAA MD in patients with normal LA volumes, while LAA MD was superior to LA MD in patients with LA enlargement. Conclusions: Higher LA and LAA mechanical dispersion are independently associated with stroke/TIA in AF patients and had incremental values over clinical and conventional echocardiographic parameters. What’s more, priorities of dispersion assessment are different depending on patients’ LA size.
14 July 2022 in Computational Intelligence and Neuroscience
Objectives. Measuring anatomical parameters in fetal heart ultrasound images is crucial for the diagnosis of congenital heart disease (CHD), which is highly dependent on the clinical experience of the sonographer. To address this challenge, we propose an automated segmentation method using the channel-wise knowledge distillation technique. Methods. We design a teacher-student architecture to conduct channel-wise knowledge distillation. ROI-based cropped images and full-size images are used for the teacher and student models, respectively. It allows the student model to have both the fine-grained segmentation capability inherited from the teacher model and the ability to handle full-size test images. A total of 1,300 fetal heart ultrasound images of three-vessel view were collected and annotated by experienced doctors for training, validation, and testing. Results. We use three evaluation protocols to quantitatively evaluate the segmentation accuracy: Intersection over Union (IoU), Pixel Accuracy (PA), and Dice coefficient (Dice). We achieved better results than related methods on all evaluation metrics. In comparison with DeepLabv3+, the proposed method gets more accurate segmentation boundaries and has performance gains of 1.8 on mean IoU (66.8 to 68.6), 2.2 on mean PA (79.2 to 81.4), and 1.2 on mean Dice (80.1 to 81.3). Conclusions. Our segmentation method could identify the anatomical structure in three-vessel view of fetal heart ultrasound images. Both quantitative and visual analyses show that the proposed method significantly outperforms the related methods in terms of segmentation results.