In this paper, a comparative experimental assessment of computer vision-based methods for sign language recognition is conducted. By implementing the most recent deep neural network methods in this field, a thorough evaluation on multiple publicly available datasets is performed. The aim of the present study is to provide insights on sign language recognition, focusing on mapping non-segmented video streams to glosses. For this task, two new sequence training criteria, known from the fields of speech and scene text recognition, are introduced. Furthermore, a plethora of pretraining schemes is thoroughly discussed. Finally, a new RGB+D dataset for the Greek sign language is created. To the best of our knowledge, this is the first sign language dataset where three annotation levels are provided (individual gloss, sentence and spoken language) for the same set of video captures.

The field of 3D hand pose estimation has been gaining a lot of attention recently, due to its significance in several applications that require human-computer interaction (HCI). The utilization of technological advances, such as cost-efficient depth cameras coupled with the explosive progress of Deep Neural Networks (DNNs), has led to a significant boost in the development of robust markerless 3D hand pose estimation methods. Nonetheless, finger occlusions and rapid motions still pose significant challenges to the accuracy of such methods. In this survey, we provide a comprehensive study of the most representative deep learning-based methods in literature and propose a new taxonomy heavily based on the input data modality, being RGB, depth, or multimodal information. Finally, we demonstrate results on the most popular RGB and depth-based datasets and discuss potential research directions in this rapidly growing field.

Parkinson’s disease (PD), the second most prevalent neurodegenerative condition, lacks a cure, but its symptoms can be managed. Its complex diagnosis and assessment need ongoing monitoring, highlighting the potential use of digital assessment tools for enhancing patient management, even outside the clinical settings. In this vein, this paper proposes a smartphone-based video analysis approach for assessing motor skills, particularly balance and posture, in individuals diagnosed with PD. In particular, the Movement Disorder Society Unified Parkinson’s Disease Rating Scale (MDS-UPDRS) ratings for items “3.8” (leg agility), “3.9” (arising from chair),“3.13” (posture) and “3.10” (gait) are estimated by capturing and analysing video from PD patients, while performing a Comprehensive Motor Function Test. Specifically, a 3D pose landmark detection (skeleton extraction) model based on the the MediaPipe Machine Learning Platform is used and different motion features are estimated from the captured videos that may correlate with the MDS-UPDRS assessments provided by clinicians. A machine learning pipeline (evaluating five different ML classifiers) is then proposed to examine the feasibility of using these features for monitoring the balance and posture of PD patients. Experimental results, obtained using a cohort of 17 Greek PD patients, voluntarily participating in this study, demonstrate that certain features have significant correlation with the clinical MDS-UPDRS ratings. These promising results showcase the potentiality of digital assessment to provide objective representation of the PD patient’s motor skills, supporting both PD clinical assessment and self-management. Ongoing work within the AI-PROGNOSIS project will further validate these findings within a larger cohort and from additional countries.

The growing emphasis on trustworthy artificial intelligence (AI) in health care reflects a shift away from models optimized for predictive performance toward governable and auditable systems that can be adopted and sustained in clinical practice. Nonetheless, many clinical AI applications continue to privilege technical performance while underaddressing ethical, regulatory, and societal considerations, leading to concerns around robustness, transparency, and clinical adoption. To address this, governance frameworks such as the Assessment List for Trustworthy Artificial Intelligence (ALTAI) have been proposed to operationalize trust-related requirements across the AI lifecycle. However, evidence on the practical use of these frameworks remains limited. In this Viewpoint, we describe the application of ALTAI as a procedural governance framework within the Horizon Europe AI-PROGNOSIS project, which aims to support Parkinson disease diagnosis and care through predictive models and digital biomarkers derived from everyday devices. The seven ALTAI requirements (ie, human agency and oversight; technical robustness and safety; privacy and data governance; transparency; diversity, nondiscrimination, and fairness; societal and environmental well-being; and accountability) were mapped to key stages of the AI lifecycle within the project, including design and specification, data preparation, model development and validation, user interface and user experience deployment, external prospective validation, and overarching management and workflow. To examine how these requirements were perceived in practice, we conducted a structured internal survey among AI developers and data scientists involved in the AI-PROGNOSIS project (n=10). Participants rated the relevance of the 17 ALTAI subdomains using a three-point prioritization scale. Technical accuracy, data governance, and privacy were consistently rated as highly relevant, whereas societal impact received the lowest prioritization. This pattern reflects a documented tension in AI development, where technical teams tend to deprioritize broader societal concerns under delivery and performance constraints. Nonetheless, this work should be interpreted as a context-specific case study rather than a validation of ALTAI. The small sample size and project-specific setting limit generalizability, and these findings should not be considered as representative of broader clinical AI development. Overall, by making prioritization gaps explicit and embedding multidisciplinary review across lifecycle checkpoints, this case study illustrates how structured governance frameworks can surface implementation tensions and support accountable AI development. While these approaches do not resolve all of the aforementioned challenges, they provide practical guidance for integrating trust-related considerations into clinical AI projects.