With the advances of data-driven machine learning research, a wide variety of prediction problems have been tackled. It has become critical to explore how machine learning and specifically deep learning methods can be exploited to analyse healthcare data. A major limitation of existing methods has been the focus on grid-like data; however, the structure of physiological recordings are often irregular and unordered, which makes it difficult to conceptualise them as a matrix. As such, graph neural networks have attracted significant attention by exploiting implicit information that resides in a biological system, with interacting nodes connected by edges whose weights can be determined by either temporal associations or anatomical junctions. In this survey, we thoroughly review the different types of graph architectures and their applications in healthcare. We provide an overview of these methods in a systematic manner, organized by their domain of application including functional connectivity, anatomical structure, and electrical-based analysis. We also outline the limitations of existing techniques and discuss potential directions for future research.

Machine learning (also called data-driven) methods have become popular in modeling flood inundations across river basins. Among data-driven methods, traditional machine learning (ML) approaches are widely used to model flood events, and recently deep learning (DL) approaches have gained more attention across the world. In this paper, we reviewed recently published literature on ML and DL applications for flood modeling for various hydrologic and catchment characteristics. Our extensive literature review shows that DL models produce better accuracy compared to traditional approaches. Unlike physically based models, ML/DL models suffer from the lack of using expert knowledge in modeling flood events. Apart from challenges in implementing a uniform modeling approach across river basins, the lack of benchmark data to evaluate model performance is a limiting factor for developing efficient ML/DL models for flood inundation modeling.

Underwater image restoration is of significant importance in unveiling the underwater world. Numerous techniques and algorithms have been developed in recent decades. However, due to fundamental difficulties associated with imaging/sensing, lighting, and refractive geometric distortions in capturing clear underwater images, no comprehensive evaluations have been conducted with regard to underwater image restoration. To address this gap, we constructed a large-scale real underwater image dataset, dubbed Heron Island Coral Reef Dataset (‘HICRD’), for the purpose of benchmarking existing methods and supporting the development of new deep-learning based methods. We employed an accurate water parameter (diffuse attenuation coefficient) to generate the reference images. There are 2000 reference restored images and 6003 original underwater images in the unpaired training set. Furthermore, we present a novel method for underwater image restoration based on an unsupervised image-to-image translation framework. Our proposed method leveraged contrastive learning and generative adversarial networks to maximize the mutual information between raw and restored images. Extensive experiments with comparisons to recent approaches further demonstrate the superiority of our proposed method. Our code and dataset are both publicly available.

Clinical gait analysis (CGA) using computer vision is an emerging field in artificial intelligence facing great challenges in obtaining accessible and annotated real-world data and clear task objectives. This paper lays the groundwork for advancing CGA, presenting vision-based methods and datasets tailored for gait analysis. We introduce the Gait Abnormality in Video Dataset (GAVD) in response to a review of over 150 existing gait-related datasets, designed to address the critical need for a large, clinically annotated gait dataset. GAVD contains 1,874 video sequences, capturing normal, abnormal, and pathological gait patterns. It features RGB data, clinically annotated from publicly available online sources, and encompassing over 400 subjects who have undergone clinical-grade visual screening. The dataset spans diverse environments, such as hospital clinics and uncontrolled urban settings. We validated GAVD through action recognition models, achieving abnormality detection rates of 94% and 92% with Temporal Segment Networks (TSN) and SlowFast networks, respectively. We further demonstrate GAVD’s clinical utility by testing on the Clinical Abnormality Simulated Dataset (CASD) achieving abnormality detection accuracy of 73% and 87%, respectively, with TSN and SlowFast networks. To facilitate further research, we provide a GitHub repository GAVD with URL links and clinically relevant annotation for over 450 videos featuring a variety of gait patterns and unique subjects.