Ultrasound (US) is generally preferred because it is of low-cost, safe, and non-invasive. US image segmentation is crucial in image analysis. Recently, deep learning-based methods are increasingly being used to segment US images. This survey systematically summarizes and highlights crucial aspects of the deep learning techniques developed in the last five years for US segmentation of various body regions. We investigate and analyze the most popular loss functions and metrics for training and evaluating the neural network for US segmentation. Furthermore, we study the patterns in neural network architectures proposed for the segmentation of various regions of interest. We present neural network modules and priors that address the anatomical challenges associated with different body organs in US images. We have found that variants of U-Net that have dedicated modules to overcome the low-contrast and blurry nature of images are suitable for US image segmentation. Finally, we also discuss the advantages and challenges associated with deep learning methods in the context of US image segmentation.
18 December 2024 in IEEE Transactions on Biomedical Engineering
Objective: Despite advancements in artificial intelligence (AI) for predicting cardiac arrest (CA) with multivariate time-series vital signs data, existing models continue to face significant problems, particularly concerning balance, efficiency, accuracy, and explainability. While neural networks have been proposed to extract multiscale features from raw data in various applications, to our knowledge, no work has utilized multiscale feature extraction, specifically for diagnostic CA prediction. This paper presents a new framework that tackles these difficulties by utilizing multiscale feature aggregation via Independent Component Analysis (ICA). Methods: We present the Pareto optimal StrataChron Pyramid Fusion Framework (SPFF) that improves temporal vital signs statistics by capturing long-term dependencies using multi-scale temporal feature aggregation. SPFF integrates with ICA to eliminate information redundancy and enhance model efficiency. We have developed and validated the approach using the public MIMIC IV dataset. Results: The proposed model demonstrates resilience to data imbalance and enhances explainability through SHapley Additive exPlanations (SHAP), Partial Dependence Plots (PDP), and Individual Conditional Expectation (ICE) across varying time windows. The multilayer perceptron (MLP) using SPFF and ICA achieves an accuracy of 0.982, precision of 0.969, recall of 0.989, F1-score of 0.979, and AUROC of 0.998. Conclusion: The proposed method effectively predicts CA across varying time windows, offering a robust solution to the challenges of efficiency, accuracy, and explainability in current models. Significance: This method is significant to biomedical research as it provides superior performance in CA prediction while capturing both short- and long-term dependencies in patient data, potentially improving patient outcomes and guiding clinical decision-making.