Data augmentation for medical imaging: A systematic literature review.

Recent advances in Deep Learning have largely benefited from larger and more diverse training sets. However, collecting large datasets for medical imaging is still a challenge due to privacy concerns and labeling costs. Data augmentation makes it possible to greatly expand the amount and variety of data available for training without actually collecting new samples. Data augmentation techniques range from simple yet surprisingly effective transformations such as cropping, padding, and flipping, to complex generative models. Depending on the nature of the input and the visual task, different data augmentation strategies are likely to perform differently. For this reason, it is conceivable that medical imaging requires specific augmentation strategies that generate plausible data samples and enable effective regularization of deep neural networks. Data augmentation can also be used to augment specific classes that are underrepresented in the training set, e.g., to generate artificial lesions. The goal of this systematic literature review is to investigate which data augmentation strategies are used in the medical domain and how they affect the performance of clinical tasks such as classification, segmentation, and lesion detection. To this end, a comprehensive analysis of more than 300 articles published in recent years (2018-2022) was conducted. The results highlight the effectiveness of data augmentation across organs, modalities, tasks, and dataset sizes, and suggest potential avenues for future research.

Self-supervised and semi-supervised learning for road condition estimation from distributed road-side cameras.

Monitoring road conditions, e.g., water build-up due to intense rainfall, plays a fundamental role in ensuring road safety while increasing resilience to the effects of climate change. Distributed cameras provide an easy and affordable alternative to instrumented weather stations, enabling diffused and capillary road monitoring. Here, we propose a deep learning-based solution to automatically detect wet road events in continuous video streams acquired by road-side surveillance cameras. Our contribution is two-fold: first, we employ a convolutional Long Short-Term Memory model (convLSTM) to detect subtle changes in the road appearance, introducing a novel temporally consistent data augmentation to increase robustness to outdoor illumination conditions. Second, we present a contrastive self-supervised framework that is uniquely tailored to surveillance camera networks. The proposed technique was validated on a large-scale dataset comprising roughly 2000 full day sequences (roughly 400K video frames, of which 300K unlabelled), acquired from several road-side cameras over a span of two years. Experimental results show the effectiveness of self-supervised and semi-supervised learning, increasing the frame classification performance (measured by the Area under the ROC curve) from 0.86 to 0.92. From the standpoint of event detection, we show that incorporating temporal features through a convLSTM model both improves the detection rate of wet road events (+ 10%) and reduces false positive alarms ([Formula: see text] 45%). The proposed techniques could benefit also other tasks related to weather analysis from road-side and vehicle-mounted cameras.

An Evaluation Testbed for Locomotion in Virtual Reality.

A common operation performed in Virtual Reality (VR) environments is locomotion. Although real walking can represent a natural and intuitive way to manage displacements in such environments, its use is generally limited by the size of the area tracked by the VR system (typically, the size of a room) or requires expensive technologies to cover particularly extended settings. A number of approaches have been proposed to enable effective explorations in VR, each characterized by different hardware requirements and costs, and capable to provide different levels of usability and performance. However, the lack of a well-defined methodology for assessing and comparing available approaches makes it difficult to identify, among the various alternatives, the best solutions for selected application domains. To deal with this issue, this article introduces a novel evaluation testbed which, by building on the outcomes of many separate works reported in the literature, aims to support a comprehensive analysis of the considered design space. An experimental protocol for collecting objective and subjective measures is proposed, together with a scoring system able to rank locomotion approaches based on a weighted set of requirements. Testbed usage is illustrated in a use case requesting to select the technique to adopt in a given application scenario.

Augmented Reality Learning Environment for Basic Life Support and Defibrillation Training: Usability Study.

BACKGROUND: Basic life support (BLS) is crucial in the emergency response system, as sudden cardiac arrest is still a major cause of death worldwide. Unfortunately, only a minority of victims receive cardiopulmonary resuscitation (CPR) from bystanders. In this context, training could be helpful to save more lives, and technology-enhanced BLS simulation is one possible solution. OBJECTIVE: The aim of this study is to assess the feasibility and acceptability of our augmented reality (AR) prototype as a tool for BLS training. METHODS: Holo-BLSD is an AR self-instruction training system, in which a standard CPR manikin is "augmented" with an interactive virtual environment that reproduces realistic scenarios. Learners can use natural gestures, body movements, and spoken commands to perform their tasks, with virtual 3D objects anchored to the manikin and the environment. During the experience, users were trained to use the device while being guided through an emergency simulation and, at the end, were asked to complete a survey to assess the feasibility and acceptability of the proposed tool (5-point Likert scale; 1=Strongly Disagree, 5=Strongly Agree). RESULTS: The system was rated easy to use (mean 4.00, SD 0.94), and the trainees stated that most people would learn to use it very quickly (mean 4.00, SD 0.89). Voice (mean 4.48, SD 0.87), gaze (mean 4.12, SD 0.97), and gesture interaction (mean 3.84, SD 1.14) were judged positively, although some hand gesture recognition errors reduced the feeling of having the right level of control over the system (mean 3.40, SD 1.04). CONCLUSIONS: We found the Holo-BLSD system to be a feasible and acceptable tool for AR BLS training.

Virtual Character Animation Based on Affordable Motion Capture and Reconfigurable Tangible Interfaces.

Software for computer animation is generally characterized by a steep learning curve, due to the entanglement of both sophisticated techniques and interaction methods required to control 3D geometries. This paper proposes a tool designed to support computer animation production processes by leveraging the affordances offered by articulated tangible user interfaces and motion capture retargeting solutions. To this aim, orientations of an instrumented prop are recorded together with animator's motion in the 3D space and used to quickly pose characters in the virtual environment. High-level functionalities of the animation software are made accessible via a speech interface, thus letting the user control the animation pipeline via voice commands while focusing on his or her hands and body motion. The proposed solution exploits both off-the-shelf hardware components (like the Lego Mindstorms EV3 bricks and the Microsoft Kinect, used for building the tangible device and tracking animator's skeleton) and free open-source software (like the Blender animation tool), thus representing an interesting solution also for beginners approaching the world of digital animation for the first time. Experimental results in different usage scenarios show the benefits offered by the designed interaction strategy with respect to a mouse & keyboard-based interface both for expert and non-expert users.

Sensors for Entertainment.

Sensors are becoming ubiquitous in all areas of science, technology, and society. In this Special Issue on "Sensors for Entertainment", developments in progress and the current state of application scenarios for sensors in the field of entertainment is explored.

Adding pluggable and personalized natural control capabilities to existing applications.

Advancements in input device and sensor technologies led to the evolution of the traditional human-machine interaction paradigm based on the mouse and keyboard. Touch-, gesture- and voice-based interfaces are integrated today in a variety of applications running on consumer devices (e.g., gaming consoles and smartphones). However, to allow existing applications running on desktop computers to utilize natural interaction, significant re-design and re-coding efforts may be required. In this paper, a framework designed to transparently add multi-modal interaction capabilities to applications to which users are accustomed is presented. Experimental observations confirmed the effectiveness of the proposed framework and led to a classification of those applications that could benefit more from the availability of natural interaction modalities.

Advances in target detection and tracking in Forward-Looking InfraRed (FLIR) imagery.

Here we give context to the Special Issue on "Detection and Tracking of Targets in Forward-Looking InfraRed (FLIR) Imagery" in Sensors. We start with an introduction to the role of infrared images in today's vision-based applications, by outlining the specific challenges that characterize detection and tracking in FLIR images. We then illustrate why selected papers have been chosen to represent the domain of interest, by summarizing their main contributions to the state-of-the-art. Lastly, we sum up the main evidence found, and we underline some of the aspects that are worthy of further investigation in future research activities.

Extensible GUIs for remote application control on mobile devices.

This framework lets users easily and effectively use desktop applications on mobile devices without any redesign. An automatically generated description of the original interface lets the original application be recreated and migrated to nomadic scenarios. Users can also customize interfaces to suit their needs and match a device's capabilities.

A streaming-based solution for remote visualization of 3D graphics on mobile devices.

Mobile devices such as Personal Digital Assistants, Tablet PCs, and cellular phones have greatly enhanced user capability to connect to remote resources. Although a large set of applications are now available bridging the gap between desktop and mobile devices, visualization of complex 3D models is still a task hard to accomplish without specialized hardware. This paper proposes a system where a cluster of PCs, equipped with accelerated graphics cards managed by the Chromium software, is able to handle remote visualization sessions based on MPEG video streaming involving complex 3D models. The proposed framework allows mobile devices such as smart phones, Personal Digital Assistants (PDAs), and Tablet PCs to visualize objects consisting of millions of textured polygons and voxels at a frame rate of 30 fps or more depending on hardware resources at the server side and on multimedia capabilities at the client side. The server is able to concurrently manage multiple clients computing a video stream for each one; resolution and quality of each stream is tailored according to screen resolution and bandwidth of the client. The paper investigates in depth issues related to latency time, bit rate and quality of the generated stream, screen resolutions, as well as frames per second displayed.

Tracking endothelial cells using multiframe point correspondence.

This paper investigates the problem of tracking individual circular cell nuclei of endothelial cells during their migration. The purpose is to extract statistical information to support research activities devoted to reaching a better understanding of the underlying biological phenomena. This goal is achieved by exploiting a general tracking framework based on a multiframe point correspondence approach and capable of translating the matching problem into a maximum path cover problem over a graph based representation of moving objects. The application to phase contrast microscopy images containing hundreds of cells aggregating to form capillary tubes required to design an ad-hoc weight function capable of properly handling cells occlusion and aggregation.