Image quality assessment (IQA) algorithms have significantly advanced over the past two decades, primarily focusing on natural images. However, applying these methods directly to medical imaging often yields suboptimal performance due to inherent differences such as the structural complexity of medical images and the limited availability of annotated databases. In this study, we conduct a comprehensive evaluation of state-of-the-art IQA methods, including 29 traditional full-reference (FR), 4 traditional no-reference (NR), and 9 deep learning-based approaches, to assess their effectiveness in the context of medical imaging. Our evaluation is performed on a recently developed MRI image quality assessment benchmark, revealing critical performance gaps in existing methods. Building on these findings, we propose a novel dual-branch deep learning framework specifically designed for medical IQA (MIQANet). The proposed approach effectively combines global contextual information with local structural details, enhancing the model’s ability to capture subtle degradations and structural inconsistencies in MRI scans. Experiential results demonstrate the superiority of our approach over existing methods, providing valuable theoretical and practical insights for enhancing quality assessment of medical images.

Radiologists’ search strategies and decision-making processes during chest X-ray diagnosis vary with their levels of expertise. Understanding these differences can inform training programmes and support the development of tools to enhance diagnostic accuracy. We hypothesize that eye-tracking data can reveal variations in expertise levels that serve as a predictor of radiologist expertise. To investigate this, we develop a database of 191 chest X-ray images with ground-truth annotations, including diagnostic decisions and eye movement patterns from 13 radiologists of varying levels of expertise. Statistical analyses reveal distinct diagnostic search patterns associated with different expertise levels. In addition, we propose a predictive framework to estimate expertise levels based on eye-tracking data. This study advances the understanding of expertise-driven differences in diagnostic search strategies and demonstrates the potential of eye-tracking data in enhancing training in clinical radiology.

Payment Channel Networks (PCNs) provide a means to improve the scaling of cryptocurrency payments by allowing peers to make payments between themselves in an efficient manner. To make a payment between two peers, the task of path planning must first be performed to determine a path in the PCN connecting the peers before the payment is performed using this path. To date, existing research has focused on the problem of performing path planning in PCNs that contain two-party channels. It has been hypothesised that the scaling of PCNs could be further improved by considering the inclusion of multi-party channels that contain more than two peers. However, the problem of performing path planning in PCNs that contain multi-party channels has not yet been considered. In this article, we address this gap in the research literature and propose a novel path planning method for PCNs containing multi-party channels. This method involves modelling the PCN with multi-party channels as a hypergraph, a type of graph where edges can contain two or more vertices, and using this model to solve the path planning problem in question. We prove that the proposed method is correct and computationally efficient. Furthermore, assuming path planning is performed using this method, we also present theoretical and experimental analyses that demonstrate the scaling benefits of using multi-party channels.

Long video action quality assessment (AQA) aims to evaluate the performance of long-term actions depicted in a video and produce an overall assessment for action quality. A video of long-term actions often contains more complicated temporal and spatial information than that of short-term actions. However, existing approaches that segment a video into individual clips for independent analysis potentially disrupt the narrative flow and diminish contextual details within and across clips, impeding comprehensive video understanding. To address this challenge, we propose an adaptive spatiotemporal graph transformer network (ASGTN) that combines multiple graph structures and transformer attention mechanisms to capture both local and global contextual information within and across clips in a long video. Specifically, the adaptive spatiotemporal graph (ASG) combines a spatial graph branch, designed to enrich the local nuanced spatiotemporal relations within an individual clip, and a temporal graph branch, tailored to dynamically learn the semantic context across different clips. Furthermore, a transformer encoder is integrated to amplify the global dependencies across clips in the entire video. This structure is designed to preserve narrative coherence and maintain essential contextual details in video-level features. Finally, we employ a level-focused decoder to predict the action quality score distribution. Experiments demonstrate that our model achieves state-of-the-art results on popular AQA datasets. Our code is available athttps://github.com/jiangliu5/ASGTN_AQA.

Radiologists' eye movements during medical image interpretation reflect their perceptual-cognitive processes of diagnostic decisions. The eye movement data can be modeled to represent clinically relevant regions in a medical image and potentially integrated into an artificial intelligence (AI) system for automatic diagnosis in medical imaging. In this article, we first conduct a large-scale eye-tracking study involving 13 radiologists interpreting 191 chest X-ray (CXR) images, establishing a best-of-its-kind CXR visual saliency benchmark. We then perform analysis to quantify the reliability and clinical relevance of saliency maps (SMs) generated for CXR images. We develop CXR image saliency prediction method (CXRSalNet), a novel saliency prediction model that leverages radiologists' gaze information to optimize the use of unlabeled CXR images, enhancing training and mitigating data scarcity. We also demonstrate the application of our CXR saliency model in enhancing the performance of AI-powered diagnostic imaging systems.

Mean opinion score (MOS) has been used as the benchmark to measure the perceived quality of digital images. However, the usefulness of MOS diminishes when a substantial variation between individual opinions occurs. It is critical to measure the stimulus-driven variance of opinion scores (VOS) and scrutinise images that evoke a large VOS, and consequently, use VOS to inform our interpretation of MOS. In this paper, we create a VOS benchmark for individual differences in image quality assessment and analyse the importance of VOS classification as a function of distortion intensity, distortion type and scene content. In addition, a simple yet effective deep learning-based model is built, aiming to identify images with a large variation in viewers’ quality judgements.

Predicting the saliency of images affected by distortion is a challenging but emerging research problem. Given a distorted image, we wish to accurately predict saliency as perceived by humans. A recent distortion-aware saliency benchmark – the CUDAS database – reveals the inadequacy of existing saliency models in handling distorted images. In this paper, we devise a deep learning Distortion-Aware Saliency Module (DASM) that enables capturing saliency features related to image distortions, and integrates this module into a saliency prediction architecture. To achieve the high expressive capability of DASM using supervised learning, we create a dedicated dataset that draws upon a large-scale saliency dataset and machine-generated image quality assessments . Experimental results demonstrate the superior performance of the proposed model in predicting the saliency of distorted images.

Magnetic resonance imaging (MRI) is susceptible to visual artifacts that can degrade the perceptual image quality, potentially leading to inaccurate or inefficient diagnoses in clinical practice. It is critical to evaluate the perceptual image quality and build this technique into clinical solutions. In a previous study, an MRI database was created for image quality assessment (IQA), where various types of MRI artifacts with different degrees of degradation were simulated. Application specialists assessed the image quality; however, radiologists’ perception of MRI image quality remains unknown. To make IQA clinically relevant, in this paper we conduct a new subjective experiment where 13 radiologists rated the quality of images contained in the MRI database. Based on this subjective IQA benchmark named RAD-IQMRI, we evaluate the performance of state-of-the-art objective IQA models, providing insights into their application for MRI image quality assessment in clinical settings.

Low quality medical images can lead to inaccurate interpretation and diagnosis. Therefore, it is important to understand radiologists' perception of distortions in visual content. In this study, 12 radiologists with different degrees of experience scored MRI images of varying levels of quality. Statistical analyses were conducted to reveal the influence of the radiologists' experience on their perception of image quality. In scoring images of joints, brain and liver, the highly experienced radiologists gave significantly higher scores than less experienced radiologists. In scoring images of fetus and spine, there were no significant differences in scores between groups with different degrees of experience. No radiologists had expertise in breast images, and their experience in other anatomical areas did not significantly affect scoring of breast images. Overall, highly experienced radiologists gave higher scores for images with edge ghosting, plain ghosting or white noise than radiologists with less experience. The findings will provide a reference for determining or improving image quality standards in clinical practice.

Topological data analysis (TDA) is an emerging field of research, which considers the application of topology to data analysis. Recently, these methods have been successfully applied to research problems in the field of geographical information science (GIS) and there is much potential for future applications. In this article, we provide an introduction to the fundamentals of TDA for GIS researchers and practitioners and highlight specific benefits that TDA methods provide relative to some conventional methods. We focus on the method of persistent homology, which is the most commonly used TDA method. We describe how persistent homology can be applied to data types commonly encountered in the GIScience domain, namely sets of points, networks and sequences of images. We also describe the application of persistent homology to two specific GIS problems, which are the point pattern analysis of UK city pubs and the analysis of UK rainfall radar imagery. In each case we stress the specific benefits of TDA methods that include, for example, generating an output signature in a form that can be subject to subsequent analyses; identification of void regions in point patterns; and providing a relatively simple method to track objects in spatio-temporal images.

We propose a novel method for generalizing continuous valued raster data with respect to topological constraints whereby smaller scale connected components and holes in the data sublevel sets are removed. The proposed method formulates the problem of generalization as an optimization problem with respect to persistent homology. We prove the objective function to be locally continuous with analytical gradients which can be used to perform optimization using gradient descent. Furthermore, we prove the convergence of gradient descent to a global optimal solution. The proposed method is general in nature and can be applied to raster data of any dimension. The utility of the method is demonstrated with respect to generalizing two- and three-dimensional raster data corresponding to digital elevation models (DEM) and subsurface mineral interpolation respectively.

In many instances of the object tracking problem the topological properties of objects can change over time. Such changes include the splitting of an object into multiple objects or merging of multiple objects into a single object. We propose a novel tracking model which is robust to such changes. This model is formulated terms of homology theory whereby 0-dimensional homology classes, which correspond to path-connected components, are tracked. A generalisation of this model for tracking spatially close objects lying in an ambient metric space is also proposed. This generalisation is particularly suitable for tracking spatial-temporal phenomena such as weather phenomena. The utility of the proposed model is demonstrated with respect to tracking rain clouds in radar imagery.

The allocation of resources to challenge city centre violent crime traditionally relies on historical data to identify hot-spots. The usefulness of such data-driven approaches is limited when historical data is scarce or unavailable (e.g. planning of a new city) or insufficiently representative (e.g. does not account for novel events, such as Olympic Games). In some cities, crime data is not systematically accumulated at all. We present a graph-constrained agent based simulation model of alcohol-related violent crime that is capable of predicting areas of likely violent crime without requiring any historical data. The only inputs to our simulation are publicly available geographical data, which makes our method immediately applicable to a wide range of tasks, such as optimal city planning, police patrol optimisation, devising alcohol licensing policies. In experiments, we evaluate our model and demonstrate agreement of our model's predictions on where and when violence will occur with real-world violent crime data. Analyses indicate that our agent based model may be able to make a significant contribution to attempts to prevent violence through deterrence or by design.

We propose a method for modeling the topology of swarm behavior in a manner which facilitates the application of machine learning techniques such as clustering. This is achieved by modeling the persistence of topological features, such as connected components and holes, of the swarm with respect to time using zig-zag persistent homology. The output of this model is subsequently transformed into a representation known as a persistence landscape. This representation forms a vector space and therefore facilitates the application of machine learning techniques. The proposed model is validated using a real data set corresponding to a swarm of 300 fish. We demonstrate that it may be used to perform clustering of swarm behavior with respect to topological features.

We present a method for compressing trajectories in an unsupervised manner. Given a set of trajectories sampled from a space we construct a basis for compression whose elements correspond to paths in the space which are topologically distinct. This is achieved by computing a canonical representative for each element in a generating set for the first homology group and decomposing these representatives into a set of distinct paths. Trajectory compression is subsequently accomplished through representation in terms of this basis. Robustness with respect to outliers is achieved by only considering those elements of the first homology group which exist in the super-level sets of the Kernel Density Estimation (KDE) above a threshold. Robustness with respect to small scale topological artifacts is achieved by only considering those elements of the first homology group which exist for a sufficient range in the super-level sets. We demonstrate this approach to trajectory compression in the context of a large set of crowd-sourced GPS trajectories captured in the city of Chicago. On this set, the compression method achieves a mean geometrical accuracy of 108 meters with a compression ratio of over 12.

This paper proposes a method for automatically inferring semantic type information for a street network from its corresponding geometrical representation. Specifically, a street network is modelled as a probabilistic graphical model and semantic type information is inferred by performing learning and inference with respect to this model. Learning is performed using a maximum-margin approach while inference is performed using a fusion moves approach. The proposed model captures features relating to individual streets, such as linearity, as well as features relating to the relationships between streets such as the co-occurrence of semantic types. On a large street network containing 32,412 street segments, the proposed model achieves precision and recall values of 68% and 65% respectively. One application of this work is the automation of street network mapping.

The task of Simultaneous Localization and Mapping (SLAM) is regularly performed in network spaces consisting of a set of corridors connecting locations in the space. Empirical research has demonstrated that such spaces generally exhibit common structural properties relating to aspects such as corridor length. Consequently there exists potential to improve performance through the placement of priors over these properties. In this work we propose an appearance-based SLAM method which explicitly models the space as a network and in turn uses this model as a platform to place priors over its structure. Relative to existing works, which implicitly assume a network space and place priors over its structure, this approach allows a more formal placement of priors. In order to achieve robustness, the proposed method is implemented within a multi-hypothesis tracking framework. Results achieved on two publicly available datasets demonstrate the proposed method outperforms a current state-of-the-art appearance-based SLAM method.

OpenStreetMap (OSM) has been demonstrated to be a valuable source of spatial data in the context of many applications. However concerns still exist regarding the quality of such data and this has limited the proliferation of its use. Consequently much research has been invested in the development of methods for assessing and/or improving the quality of OSM data. However most of these methods require ground-truth data, which, in many cases, may not be available. In this paper we present a novel solution for OSM data quality assessment that does not require ground-truth data. We consider the semantic accuracy of OSM street network data, and in particular, the associated semantic class (road class) information. A machine learning model is proposed that learns the geometrical and topological characteristics of different semantic classes of streets. This model is subsequently used to accurately determine if a street has been assigned a correct/incorrect semantic class.

Convexity represents a fundamental descriptor of object shape. This paper presents a new convexity measure for both open and closed simple contours. Given such a contour this measure extracts two corresponding open convex hulls. The shape similarity between these two hulls and the original contour is then computed and normalized to give a measure of convexity. The time complexity of the proposed technique is O(n). The authors believe this technique represents the first measure of convexity which uses shape similarity and which can be applied to both open and closed contours. The proposed technique is shown to provide similar or greater performance relative to two other state of the art techniques.

The use of location-enabled mobile technology is ubiquitous. We outline the opportunities and risks involved in using user-generated spatial data and information. User generated spatial data is a very dynamic but has many inconsistencies. This could severely limit its use in many security and intelligence applications.

This article contains a mathematical analysis of strategies for determining topological consistency of vector map simplifications. Such techniques exploit assumptions that can be made regarding the similarity of corresponding objects in successive simplifications. We propose that all topological relationships may be classified as planar or non-planar. A formal analysis of techniques for determining topological consistency of a simplification in terms of such relationships is presented. For each technique we analyse any corresponding constraints that are imposed. This provides a unified understanding of the benefits and limitations of individual techniques and the relationships that exist between techniques. Subsequently, a new strategy for determining the topological consistency of a simplification is proposed. This technique integrates the benefits all methods studied to provide a solution which is subject to less constraints. The effectiveness of this approach is demonstrated through fusion with an existing simplification technique resulting in simplifications that have equal topology and similar shaped features to the original map.

To perform simultaneous localization and mapping (SLAM) in dynamic environments, static background objects must first be determined. This condition can be achieved using a priori information in the form of a map of background objects. Such an approach exhibits a causality dilemma, because such a priori information is the ultimate goal of SLAM. In this paper, we propose a background foreground segmentation method that overcomes this issue. Localization is achieved using a robust iterative closest point implementation and vehicle odometry. Background objects are modeled as objects that are consistently located at a given spatial location. To improve robustness, classification is performed at the object level through the integration of a new segmentation method that is robust to partial object occlusion.

Volunteered Geographic Information (VGI) is currently a "hot topic" in the GIS community. The OpenStreetMap (OSM) project is one of the most popular and well supported examples of VGL Traditional measures of spatial data quality are often not applicable to OSM as in many cases it is not possible to access ground-truth spatial data for all regions mapped by OSM. We investigate to develop measures of quality for OSM which operate in an unsupervised manner without reference to a "trusted" source of ground-truth data. We provide results of analysis of OSM data from several European countries. The results highlight specific quality issues in OSM. Results of comparing OSM with ground-truth data for Ireland are also presented.