BackgroundThe General Movements Assessment (GMA) is a strong, non-invasive predictor of neurodevelopmental risk in early infancy, but automation studies often rely on small, heterogeneous public datasets, fostering overfitting and limited generalization.MethodsWe propose a unified preprocessing pipeline that standardizes 2D pose time-series from three public cohorts (MINI-RGBD, RVI-38, PMI-GMA; total n=1170) into histogram-encoded joint features. Sequences are represented as skeleton graphs and classified with an adapted multi-stage spatio-temporal graph convolutional network (MS-STGCN). Key hyperparameters (learning rate, weight decay, feature maps, bins, pooling, threshold) are tuned via systematic random search in single-dataset and pooled-dataset experiments.ResultsIn single-dataset evaluations, sensitivity reached up to 96% (with reduced specificity). On the combined cohort, the model achieved 62% accuracy, 83% sensitivity, and 44% specificity, indicating viable cross-dataset generalization with a sensitivity/specificity trade-off.ConclusionStandardizing preprocessing and rigorously tuning a graph-based model across multiple public datasets advances scalable, automated GMA while underscoring persistent specificity challenges in small, heterogeneous infant-movement data.

Malocclusion, a prevalent dental condition worldwide, necessitates orthodontic intervention to correct tooth misalignment and improve oral health. Treatment can involve extraction of permanent teeth, depending on dental crowding, jaw relationships, and facial aesthetics. Today, clinical decision support systems have introduced machine learning (ML) to assist orthodontists in determining optimal treatment plans. This study explores the development of a novel, fully automated method for extracting dentoalveolar features from 3D intraoral scans (IOS), aiming to enhance orthodontic decision-making. Using deep learning-based IOS segmentation as basis, dental measurements were developed and utilized to train supervised ML classifiers, including support vector machines (SVM), logistic regression, decision trees, and random forests. An ensemble of SVM models demonstrated the highest accuracy (73%) in predicting extraction decisions, with these novel domain-specific features proving more informative than traditional dental arch measurements. While we can make further improvements not only in the automated segmentation but also by applying feature selection, the results highlight the potential of AI-driven analysis to streamline orthodontic workflows, reduce manual intervention and improve clinical efficiency.

This study aimed to assess the synchronization between a new respiratory video-based motion monitor and the spontaneous breathing of premature infants on noninvasive mechanical ventilation. The experimental setup involved placing a tag on the infant’s chest and abdomen, capturing their breathing with a camera, and comparing the video data with the ventilator's readings. The video monitor proved more accurate than the respirator, detecting 93.3% of respiratory movements versus 73.2% by the respirator. The video also detected the onset of breathing earlier than the respirator. The video monitor showed good precision and potential clinical benefits for premature infants, offering a noninvasive, cost-effective alternative for respiratory monitoring. Further studies are needed to refine its performance in various settings.

Orthodontic treatment planning often involves de-ciding whether to extract teeth, a critical and irreversible decision. Integrating machine learning (ML) can enhance decision-making. This study proposes using Intraoral Scans (IOS) 3D models to predict extraction/non-extraction binary decisions with ML models. We leverage a multiview approach, using images taken from multiple points of view of the 3D model. The methodology involved a dataset composed of preprocessed IOS from 181 subjects and an experimental procedure that evaluated multiple ML models in their ability to classify subjects using either grayscale pixel intensities or radiomic features. The results indicated that a logistic model applied to the radiomic features from the back and frontal views of the 3D models was one of the best model candidates, achieving a test accuracy of 70 % and F1 score of. 73 and. 65 for non-extraction and extraction cases, respectively. Overall, these findings indicate that a multiview approach to IOS 3D models can be used to predict extraction/non-extraction decisions. In addition, the results suggest that radiomic features provide useful information in the analysis of IOS data.

Smart cities can generate a lot of useful data for analyzing behavior and infrastructures, helping with urban planning. However, the applications found are often limited to desktop viewing only, which can hinder or restrict data analysis. This work proposes an immersive data visualization of the population of medical centers to assist in the analysis and development of new strategies to deal with overcrowding situations. The prototype allows the visualization of crowd data located in places of interest on top of a 3D map. The user can view the number of people present at the location at the current stage of the simulation or over time. To test the prototype, data from an infectious disease simulator was used, called LODUS. The results showed that the application has potential for visualizing this type of data.

This study investigates the utilization of low-cost three-dimensional visualization devices., specifically virtual reality applications on smartphones., to enhance radiological diagnosis processes. Radiology., a critical field in medicine., often faces challenges such as external illumination and poor ergonomic conditions during diagnostic procedures. Virtual reality technology has emerged as a potential solution to address these issues. The research conducted user studies with radiologists and medical physicists to assess the effectiveness and usability of the virtual reality application. Feedback from participants indicated positive perceptions of the virtual environment., 3D model quality., and interaction with the application. The study also evaluated geometric transformations on the 3D model, highlighting the importance of user-friendly controls. Overall, the findings suggest that virtual reality technology on smartphones holds promise in supporting radiological diagnosis by providing an immersive and efficient tool for medical professionals.

This article describes a framework for modeling and executing training in augmented and virtual reality environments. The framework was designed based on characteristics observed in existing training applications for complex tasks, such as the use of tools, control panels and the need for step-by-step instructions. Unlike other frameworks, in the proposed system, it is possible to create the training entirely within the virtual/augmented environment, avoiding constant switching between 2D and 3D environments. The framework allows for the definition of steps in a training program, each of which includes textual instructions, videos, and 3D objects, static or animated, anchored in the real world. To demonstrate the capabilities of the framework, a training program for operating a Universal Testing Machine was created as a case study. Overall, the proposed framework allows for the creation of effective and efficient AR training programs for a variety of tasks and industries.

Measuring involuntary eye movement under specific stimuli is an important way to identify diseases such as balance disorders. Exams based on video-oculography (VOG) equipment are able to detect horizontal and vertical displacements of the pupil. However, detecting torsional movements is still a challenge. Although conventional methods have good accuracy, their results can be influenced by artifacts, such as a torsion center displacement, interference by illumination, reflections, and changes in the pupil dilation. We propose a novel method which improves the robustness of this measurement by applying the Lucas-Kanade Pyrm (LKP) optical flow technique to the captured image, directly over the iris, rather than making a polar transformation. Retaining this additional information allows multiple features over the iris to be analyzed individually and as a group, providing correction of the torsion center displacements, filtering features with reflections and adapting to different pupil dilations before the torsion angle is calculated. The accuracy and performance of this method were evaluated by comparing it against a conventional method when detecting torsional movements on videos with a known ground truth. Moreover, a simplified version of the proposed method is also evaluated, in order to analyze the impacts of a torsion center displacement. Results show that the proposed method has higher accuracy and equivalent performance to the conventional method.

The analysis of Confocal Laser Endomicroscopy (CLE) is one of the techniques used for diagnosing gastroenterological diseases. However, the manual analysis of such images requires training and experience and will often lead to wrong diagnostics. This work explores the use of attributes taken from classic texture description techniques, gray level co-occurrence matrices (GLCM) and local binary patterns (LBP), as inputs for classifiers to separate images from 3 common gastroenterological diseases, with 262 images. A baseline classifier was trained for the 10 smaller groups and two others were trained using GLCM and LBP attributes. Overall, the benefits of using texture analysis techniques and attributes can be observed as an increase in accuracy and consistency of the results.

This paper aims to study the process of converting the depth information of a real scene captured in a real environment into a tactile representation through a haptic device. We developed a belt-shaped interface with a matrix of 35 (7x5) vibrotactile actuators attached to the users' abdomen. Tests demonstrated that the device can help users to perceive the movement of objects and people, as well as allow them to move in environments containing obstacles without the usage of the vision. The system was tested both with users who are blind and with blindfolded participants. The stages of building and testing the interface as well as the tests applied in this research are described.

This paper presents a new methodology for quality assessment of interaction techniques in immersive virtual environments, based on the study of the relationships between physiological measures and usability metrics using multivariate data analysis. Our methodology defines a testing protocol, a normalization procedure and statistical techniques, considering the use of physiological measures during the evaluation process. A case study comparison between two 3D interaction techniques (ray-casting and HOMER) shows promising results, pointing to heart rate variability, as measured by the NN50 parameter, as a potential index of task performance. Besides, this study also shows that heart rate (HR) and skin conductance (SC) measures reflect the user's task performance during the interaction process. Despite these results, this work reveals that physiological measures still cannot be considered as substitutes of evaluation metrics for 3D interfaces, but may be useful in the interpretation and understanding process of them. Discussions also indicate the further studies are needed to establish guidelines for evaluation processes based on well-defined associations between human behaviors and human actions realized in 3D user interfaces.

This work consists of the study of techniques of robotics and virtual reality to develop a simulator that can be used in robotics schools, having an adequate visualization and a simple and intuitive way of interaction. For this, a 3D virtual environment for robotics was developed. Virtual reality resources have been incorporated to improve the visualization and to facilitate the user interaction with the environment. In order to evaluate the effectivity of the environment, user experiments were carried out on four different hardware configurations. During the simulations, the users had to create trajectories while implicitly defining reference points. From these experiments, automatic reports for the quantitative questions were generated, and questionnaires were filled for the qualitative questions. The results have shown that the use of virtual reality do helps the users in task execution, improving the visualization, reducing the time spent for the tasks and increasing the precision.

This study evaluates the usage of tactile feedback to aid cooperative object manipulation using the SkeweR technique. This technique is based on the use of crushing points, where the users grab the object for the first time, to simultaneously move/rotate an object. Once the user keeps his hand positioned on the crushing point, during the object manipulation, the interaction becomes more natural, in the sense that it is more similar to the real process. However, due to the lack of any physical constraint to the users' movements, it is often noticed that the user's hand moves apart from the crushing point during the interaction. To solve this problem, this work proposes the usage of tactile feedback to inform the user about the distance of his hand from the crushing point. The tactile feedback is provided by a vibration micromotor attached to the users' thumb. To validate our method, we ran a user study based on the 3D manipulation of a virtual object, which has to be translated and rotated through a virtual path along a virtual wire, from the beginning to the end of it. During the interaction, users manipulate a 3DOF position tracker and should keep this tracker at the same position of the crushing point. During the trials, the participants used three modalities of interaction: without any feedback, with a visual feedback and with tactile feedback. Results showed that the users kept the tracker closer to the crushing point when using tactile feedback.

Whether it in the military, law enforcement or private security, dismounted operators tend to deal with a large amount of volatile information that may or may not be relevant according to a variety of factors. In this paper we draft some ideas on the building blocks of an augmented reality system aimed to improve the situational awareness of dismounted operators by filtering, organizing, and displaying this information in a way that reduces the strain over the operator.

This paper describes a Virtual Reality simulator to perform dentist training in tasks such as the access to the pulp chamber and the preparation of root canal. The simulator consists of a haptic device adapted to simulate the instruments used in dental procedures, a virtual environment simulating the interactions between the haptic device and the tooth being treated and a visualization tool of the simulation.

This paper presents the project and the development of a prototype for a truck simulator aimed at assessing the use of Economic Driving Techniques. We describe the techniques for economic driving and the way they are monitored, as well as the process of modeling and creating a virtual environment and the interaction devices used in the simulator. The simulation is accomplished by using a virtual desktop environment with hardware configuration similar to that in a real truck, in order to provide the user with a higher level of immersion. By testing drivers and economic driving instructors, it was possible to observe that the prototype can be used as a tool for assessing drivers and that the system is able to perceive a great part of the violations in the use of Economic Driving Techniques during the simulation similarly to a human expert.

This work presents a methodology to formally model and to build collaborative three dimensional interaction tasks in virtual environments using three different tools: Petri Nets, interaction technique decomposition taxonomy and object-oriented concepts. The user operations in the virtual environment are represented as Petri Net nodes and these nodes, when linked, represent the interaction process stages. The integration of these approaches results in a modular application, based on the Petri Nets formalism that allows for specification of collaborative interaction tasks, and also the reuse of developed blocks in new virtual environment projects.