Abstract:The precise identification of end-systolic (ES) and end-diastolic (ED) frames from ultrasound films is crucial for automated biometric assessment and diagnosis of fetal echocardiograms. Unlike adult cardiac imaging, fetal echocardiograms present unique challenges due to cardiac images variations across different gestational weeks, rapid heart rates, and changes in cardiac shape influenced by fetal position and sound-beam angle. Addressing these challenges, we propose an Optimized Pyramidal Convolution Shuffle Attention Neural Network for Fetal Cardiac Cycle Detection (OPCSANN-FCCD). Utilizing data from the Shenzhen Maternal and Child Health Hospital dataset, the input is first preprocessed with an Anisotropic Diffusion Filter (ADF) to eliminate noise and enhance image quality. Statistical features are then extracted using a window-adaptive gray level co-occurrence matrix (GLCM). These features are classified into ES and ED frames by a Pyramidal Convolution Shuffle Attention Neural Network (PCSAN). Implemented in Python and evaluated with performance metrics such as accuracy, precision, F1-score, sensitivity, specificity, error rate, and computational time, the OPCSANN-FCCD method demonstrates significant improvements. When compared to the existing methods, it shows an accuracy increase of 25.0%, 27.6%, and 22.4%, precision enhancement of 26.4%, 25.2%, and 23.0%, and a computational load reduction of 31.2%, 29.8%, and 26.6% against existing methods such as, Fetal cardiac cycle detection in multi-resource echocardiograms utilizing hybrid categorization framework (FCCD-ME-HCF), Fetal cardiac ultrasound standard section detection model depending on multitask learning with mixed attention mechanism (FCUSD-ML-MAM) and HFSCCD: A Hybrid Neural Network for Fetal Standard Cardiac Cycle Detection in Ultrasound Videos (HFSCCD-UV) respectively.

Optical Flow Networks for Heartbeat Estimation in 4D Ultrasound Images

Unsupervised 4D Cardiac Motion Tracking with Spatiotemporal Optical Flow Networks

LAFlowNet: A Dynamic Graph Method for the Prediction of Velocity and Pressure Fields in Left Atrium and Left Atrial Appendage

A Dense Optical Flow Registration Algorithm Based on Deep Learning

Deep Learning Assessment of Myocardial Infarction From MR Image Sequences.

HFSCCD: A Hybrid Neural Network for Fetal Standard Cardiac Cycle Detection in Ultrasound Videos

A Unified Framework Integrating Recurrent Fully-Convolutional Networks and Optical Flow for Segmentation of the Left Ventricle in Echocardiography Data

Automatic Detection of Cardiac Chambers Using an Attention-based YOLOv4 Framework from Four-chamber View of Fetal Echocardiography

Multiview and Multiclass Image Segmentation using Deep Learning in Fetal Echocardiography

A deep learning framework for identifying and segmenting three vessels in fetal heart ultrasound images

Improving the Quality of Fetal Heart Ultrasound Imaging with Multihead Enhanced Self-Attention and Contrastive Learning

Optical flow estimation of coronary angiography sequences based on semi-supervised learning

4DFlowNet: Super-Resolution 4D Flow MRI Using Deep Learning and Computational Fluid Dynamics

Respiration Monitoring for Premature Neonates in NICU

Automatic Recognition of Fetal Heart Standard Section Based on Fast-RCNN

Segmentation of four-chamber view images in fetal ultrasound exams using a novel deep learning model ensemble method

A Coarse-Fine Collaborative Learning Model for Three Vessel Segmentation in Fetal Cardiac Ultrasound Images

Automated Detection of Congenital Heart Disease in Fetal Ultrasound Screening

Active Cardiac Model and Its Application on Structure Detection from Early Fetal Ultrasound Sequences

Optimized Pyramidal Convolution Shuffle attention neural network based fetal cardiac cycle detection from echocardiograms

Application of computer-aided diagnosis of congenital heart disease in four-chamber view of fetal heart basic screening