30793 research outputs found
Sort by
Integrating Explainable AI in Medical Devices: Technical, Clinical and Regulatory Insights and Recommendations
Get PDFA preprint version of the article is available at arXiv:2505.06620v1 [cs.HC], https://arxiv.org/abs/2505.06620, [v1] Sat, 10 May 2025 12:09:19 UTC (1,260 KB), under a CC BY license. It has not been certified by peer review.Availability of data and materials: CPRD cardiovascular disease synthetic dataset used in this paper can be requested from CPRD (https://cprd.com/cprdcardiovascular-disease-synthetic-dataset)There is a growing demand for the use of Artificial Intelligence (AI) and Machine Learning (ML) in healthcare, particularly as clinical decision support systems to assist medical professionals. However, the complexity of many of these models, often referred to as black box models, raises concerns about their safe integration into clinical settings as it is difficult to understand how they arrived at their predictions. This paper discusses insights and recommendations derived from an expert working group convened by the UK Medicine and Healthcare products Regulatory Agency (MHRA). The group consisted of healthcare professionals, regulators, and data scientists, with a primary focus on evaluating the outputs from different AI algorithms in clinical decision-making contexts. Additionally, the group evaluated findings from a pilot study investigating clinicians' behaviour and interaction with AI methods during clinical diagnosis. Incorporating AI methods is crucial for ensuring the safety and trustworthiness of medical AI devices in clinical settings. Adequate training for stakeholders is essential to address potential issues, and further insights and recommendations for safely adopting AI systems in healthcare settings are provided.This work was funded by the Regulators Pioneer Fund 3, Department for Science, Innovation and Technology. The RPF is a grant-based fund to enable UK regulators and local authorities to help create a UK regulatory environment that encourages business innovation and growth. The current £12m round is being delivered by DSIT. This work was also supported by the UK Regulatory Science and Innovation Networks– Implementation Phase: Human Health CERSIs programme through the project RADIANT: Regulatory Science Empowering Innovation in Transformative Digital Health and AI (Grant Ref: MCPC24031), funded by the Medical Research Council (MRC) and Innovate UK
Memorandum submitted to the House of Commons Energy Security and Net-zero Select Committee for their inquiry on: Workforce planning to deliver clean, secure energy
Get PDFWritten evidence submitted to the House of Commons Energy Security and Net-zero Select Committee for the inquiry on: Workforce planning to deliver clean, secure energy, 13 January 2025
Is videoconference pulmonary rehabilitation associated with improvements in knowledge in people living with COPD? A propensity-matched service-evaluation
Get PDFIntroduction: Pulmonary rehabilitation (PR) services are increasingly using alternative programme delivery modes, for example telerehabilitation strategies including videoconferencing, to improve patient choice and accessibility. Although telerehabilitation results in improvements in core outcomes, the effect on knowledge attainment is not known. Aim: To observe the real-world responses of patients choosing to undergo videoconference PR to a matched control group choosing to undergo in-person PR, in terms of knowledge attainment. Methods: Using propensity score matching, 25 people with COPD who completed videoconference PR were matched 1:1 with a control group of 25 people with COPD who completed in-person PR. Knowledge attainment was measured using the Lung Information Needs Questionnaire (LINQ). Results: There was a statistically and clinically significant improvement in LINQ score in both groups (mean (95%CI): videoconference −3.2 (−4.7 to −1.6); in-person −3.0 (−4.5 to −1.4)), with no significant between-group difference (mean (95%CI): 0.2 (−2.0 to −2.4)). 76% and 80% of participants achieved the minimal important difference of the LINQ in the videoconference and in-person PR groups respectively. Conclusion: In conclusion, this real-world service evaluation indicates that videoconference PR may be associated with similar improvements in knowledge attainment as in-person PR, but this requires corroboration due to the small sample size.The author(s) received no financial support for the research, authorship, and/or publication of this article
Education-focused academics in pursuit of third-space visibility: a five-step process
Get PDFSpecial Edition, Third Space In HEEducation-focused academics were introduced to enhance the quality of education and assist universities in addressing sector metrics that would place them favourably within an increasingly competitive environment. We argue that they form part of third-space staff in universities as their work remains mostly hidden as part of a ‘taken-for-granted’ infrastructure. Following the infrastructure metaphor, we propose a five-step process in the career development of education-focused academics, from early career to educational leaders, that would bring visibility to their work and that of other third-space occupants. Making third-space work visible on the personal, organisational, and sectoral levels will ensure real impact and fulfilment of its disruptive and transformative potential for Higher Education Institutions (HEIs)
Formar a los Formadores: Uso de un Profesor de IA para Capacitar a Docentes en la India
Get PDFThis study explores the effectiveness of utilising an autonomous AI teacher, represented through an avatar, in enhancing the skills and upskilling existing teachers in India. Focused on nursery-level teaching methods and curriculum development, the research employed a specialised generative AI language model, OIMISA7B, with a 7-billion parameter setup designed specifically for educational delivery. The AI avatar, designed to resemble a female trainer, delivered a series of four lessons, each approximately one hour in length. The study assessed the efficacy of this innovative teaching approach through monitoring programme completion rates, and tracking inquiries made to the AI for further clarifications and doubts. The core focus of the paper is to go beyond directly AI-enabled teaching for students which the researchers have conducted several times before, and instead look at empowering the ecosystem, in this case training the teachers.Resumen: Este estudio explora la eficacia de utilizar un profesor autónomo basado en inteligencia artificial (IA), representado a través de un avatar, para mejorar las habilidades y capacitar a docentes en la India. Centrado en métodos de enseñanza para nivel preescolar y desarrollo curricular, la investigación empleó un modelo de lenguaje generativo de IA especializado, OIMISA7B, con una configuración de 7 mil millones de parámetros diseñada específicamente para la enseñanza. El avatar de IA, diseñado para parecerse a una instructora femenina, impartió una serie de cuatro lecciones, cada una de aproximadamente una hora de duración. El estudio evaluó la eficacia de este enfoque innovador de enseñanza al monitorear las tasas de finalización del programa y rastrear las consultas realizadas a la IA para aclaraciones y dudas adicionales. El enfoque principal del artículo es ir más allá de la enseñanza habilitada directamente por IA para los estudiantes, la cual los investigadores han llevado a cabo en varias ocasiones anteriormente, y centrarse en fortalecer el ecosistema, en este caso, capacitando a los docentes
Conic relaxations for conic minimax convex polynomial programs with extensions and applications
Get PDFIn this paper, we analyze conic minimax convex polynomial optimization problems. Under a suitable regularity condition, an exact conic programming relaxation is established based on a positivity characterization of a max function over a conic convex system. Further, we consider a general conic minimax ρ-convex polynomial optimization problem, which is defined by appropriately extending the notion of conic convexity of a vector-valued mapping. For this problem, it is shown that a Karush-Kuhn-Tucker condition at a global minimizer is necessary and sufficient for ensuring an exact relaxation with attainment of the conic programming relaxation. The exact conic programming relaxations are applied to SOS-convex polynomial programs, where appropriate choices of the data allow the associated conic programming relaxation to be reformulated as a semidefinite programming problem. In this way, we can further elaborate the obtained results for other special settings including conic robust SOS-convex polynomial problems and difference of SOS-convex polynomial programs.Research of J. Vicente-Pérez was partially supported by Grant PID2022-136399NB-C21 funded by MICIU/AEI/10.13039/501100011033 and by ERDF/EU, and by Grant AICO/2021/165 from the Generalitat Valenciana
A Novel Pairwise Domain-Adaptation-Assisted Dual-Task Learning Approach to Coprediction of Robotic Machining Efficiency and Quality in New Parameter Spaces
Get PDFAccurate prediction of material removal depth and averaged surface roughness is crucial for evaluating the performance of robotic belt grinding (RBG). Nevertheless, the machining parameters of RBG across different spaces exhibit various data distributions, which often results in prediction shifts on unseen machining parameters when using conventional approaches. In this article, we introduce a pairwise domain adaptation-assisted dual-task learning (PW-DA-DTL) method for copredicting material removal depth and averaged surface roughness with regard to new RBG machining parameter spaces. The multigate mixture-of-experts method is employed as the foundational framework for dual-task learning, effectively capturing and modeling the relationships between material removal depth and average surface roughness by leveraging their inherent task interdependencies. The pairwise domain adaptation strategy is put forward to simultaneously enhance sample diversity and mitigate cross-domain data distribution discrepancy between the existing and new RBG machining parameter spaces. Comparative experiments are presented to demonstrate the effectiveness and superiority of the proposed PW-DA-DTL method.10.13039/501100001809-National Natural Science Foundation of China (Grant Number: 52188102);
10.13039/501100002858-China Postdoctoral Science Foundation (Grant Number: 2024M750991);
Postdoctor Project of Hubei Province of China (Grant Number: 2024HBBHCXA010);
Open Project Fund of Key Laboratory of Image Processing and Intelligent Control;
10.13039/501100002338-Ministry of Education of the People's Republic of China
Performance Analysis of Terahertz Communication Systems With RSMA and Hardware Impairments
Get PDFTerahertz (THz) communication has received much attention recently for its large bandwidth availability, high data-rate transmission and alleviating the spectrum shortage, and it can meet the requirements of Internet of Things with large system capacity and networking capability. In this paper, the performance of multi-antenna THz communication systems with rate-splitting multiple access (RSMA) under the hardware impairments and imperfect successive interference cancelation (SIC) are investigated, where the THz channel is modeled as a composite fading channel including the molecular absorption effects, misalignment fading and small-scale α-μ fading. Taking the hardware impairments and imperfect SIC into account, the probability density function and cumulative distribution function of the effective channel gain are derived. A joint zero-forcing and maximum ratio transmission beamforming design is employed to eliminate the interference among devices. Then, with the performance analysis, the closed-form outage probability (OP) and diversity gain of the system are respectively deduced. By minimizing the OP, a closed-form power allocation (PA) scheme is proposed to adjust the PA coefficients between the common stream and private streams, and resultant lower OP is attained. Moreover, the closed-form expression of the ergodic sum rate (ESR) is derived by means of Fox-H function and the Meijer-G function. With this ESR expression, the asymptotic ESR at high signal to noise ratio (SNR) is also provided to gain further insights. Furthermore, a simple upper bound of the ESR is derived for performance evaluation based on the Jensen’s inequality. Simulation results show that the theoretical analysis is effective, and the proposed PA scheme can obtain lower OP. Besides, the impact of different system and fading parameters on the performance are also analyzed.10.13039/501100001809-National Natural Science Foundation of China (Grant Number: 62471228
A complexity leadership theory framework for engineering management
Get PDFThis thesis was submitted for the award of Doctor of Philosophy and was awarded by Brunel University LondonComplexity Leadership Theory (CLT) views organizations as complex systems and proposes important implications for leadership theory. However, current research on CLT has focused mainly on its theoretical development, providing limited practical application examples for complex industries, such as that found in manufacturing engineering. To address this research gap, a complexity leadership framework was developed in collaboration with practitioners through an action research program in an engineering organization headquartered in the United Kingdom that specializes in aerospace and defence. The proposed framework builds on previous research that identified three leadership modes (i.e., Administrative, Enabling, and Adaptive) for complex engineering environments. Specifically, data were collected from 852 leaders, with action research involving 37 participants from the UK, USA, Canada, Australia, and Ireland. The proposed complexity leadership framework, titled “the 5-point framework”, reflects the fundamentals of leading within the organization’s complex engineering environment. The framework was validated through practitioner review and five independent elite interviews with leaders from diverse organizations. The study’s results contribute to current literature by expanding the previously proposed three leadership modes to five modes (i.e., Direction, Strategic Administrative, Enabling, Tactical Administrative and Adaptive). The findings contribute to the current discourse by providing engineering leaders with a refined framework that addresses the multifaceted leadership requirements critical for navigating complex environments
Data sources:
Access Repository Dashboard
Do you manage Open Research Online? Become a CORE Member to access insider analytics, issue reports and manage access to outputs from your repository in the CORE Repository Dashboard! 👇