Improved biomedical data harmonisation, the cornerstone of trustworthy and responsible AI in Healthcare

阻碍人工智能进入医疗领域的一个关键问题是其可信度。人工智能有可能彻底改变医疗保健和医学，实现更快、更准确的诊断、个性化的治疗计划和更高效的医疗保健系统。然而，在医疗保健和医学领域采用人工智能也面临着挑战，特别是在确保这些系统的可信度方面。值得信赖的人工智能对于将人工智能成功整合到卫生环境中至关重要，因为它确保了人工智能系统的可靠性、透明度和问责制。数据对于人工智能的发展至关重要，因为它是机器学习算法的动力。训练数据集的偏差和缺乏可解释性将限制这些技术的有效性。偏差是通过各种因素引入的，包括研究设置、数据收集方法和用于准备数据的算法。所有这些都会导致不准确的预测和诊断，给患者带来潜在的有害后果。因此，减少训练数据集的偏差对于在医疗保健领域建立值得信赖的人工智能系统至关重要。生物医学部门正试图解决这些问题，但通常情况下，协调来自不同来源的核心生物医学数据阻碍了项目的推进。产品开发和机器学习团队经常无法使用更大的统一数据集，因为通常不容易将不同来源的生物医学数据整合在一起。该项目将使新成立的公司能够弥合关键的商业化差距，并获得一个工作示范，用于协调和评估应用程序开发人员感兴趣的关键数据资源的协调步骤，作为市场切入点。该项目的起源是人工智能模型对特定人群的不公平和不准确。我们的创新将有助于确保对全球人工智能/机器学习市场的公平影响，无论地理、历史、文化和收入如何。公共资金将用于弥合商业化差距，并使英国新企业的外来投资能够利用新兴的高增长行业。

A key problem holding back AI in healthcare is its trustworthiness. AI has the potential to revolutionise healthcare and medicine, enabling faster and more accurate diagnoses, personalised treatment plans, and more efficient healthcare systems. However, the adoption of AI in healthcare and medicine comes with challenges, particularly in ensuring the trustworthiness of these systems. Trustworthy AI is essential for the successful integration of AI into health settings, as it ensures the reliability, transparency, and accountability of AI systems.Data is essential for AI development as it is the fuel powering the ML algorithms. Bias in the training datasets and the lack of interpretability will limit the effectiveness of such technologies. Biases are introduced through a variety of factors, including the study setup, the data collection methods, and the algorithms used to prepare the data. All these result in inaccurate predictions and diagnoses, leading to potentially harmful consequences for patients. Reducing the bias of training datasets is therefore critical to building trustworthy AI systems in healthcare.The biomedical sector is trying to solve these issues, but more often than not, harmonising core biomedical data from different sources blocks projects moving forward. Product development and ML teams are often held back from using larger harmonised datasets because it is often not easy to bring different sources of biomedical data together.The project will enable the newly formed company to bridge a key commercialisation gap and obtain a working demonstrator for hamarising and evaluating the harmonisation steps for key data resources of interest to application developers, as a beachhead market entry point.The project's genesis is the inequity and inaccuracy of AI models to specific population groups. Our innovation will help to ensure an equitable impact on the global AI/ML market, regardless of geography, history, culture and income.The public funding will be used to bridge the commercialisation gap and enable inward investment in a new UK venture to capitalise on an emerging high-growth sector.