Moving towards vertically integrated artificial intelligence development

Abstract

Substantial interest and investment in clinical artificial intelligence (AI) research has not resulted in widespread translation to deployed AI solutions. Current attention has focused on bias and explainability in AI algorithm development, external validity and model generalisability, and lack of equity and representation in existing data. While of great importance, these considerations also reflect a model-centric approach seen in published clinical AI research, which focuses on optimising architecture and performance of an AI model on best available datasets. However, even robustly built models using state-of-the-art algorithms may fail once tested in realistic environments due to unpredictability of real-world conditions, out-of-dataset scenarios, characteristics of deployment infrastructure, and lack of added value to clinical workflows relative to cost and potential clinical risks. In this perspective, we define a vertically integrated approach to AI development that incorporates early, cross-disciplinary, consideration of impact evaluation, data lifecycles, and AI production, and explore its implementation in two contrasting AI development pipelines: a scalable “AI factory” (Mayo Clinic, Rochester, United States), and an end-to-end cervical cancer screening platform for resource poor settings (Paps AI, Mbarara, Uganda). We provide practical recommendations for implementers, and discuss future challenges and novel approaches (including a decentralised federated architecture being developed in the NHS (AI4VBH, London, UK)). Growth in global clinical AI research continues unabated, and introduction of vertically integrated teams and development practices can increase the translational potential of future clinical AI projects.

Insights and trends review: artificial intelligence in hand surgery

“…Bias and discrimination should be considered during model development with an awareness that machines are ultimately reliant on the data they are given. There should be an equality impact assessment (NHS England, 2019b), as highlighted by the Mayo Clinic AI factory where there is an automated assessment of model bias across population subgroups (Mayo Clinic Platform, 2021; Zhang et al., 2022) and the concept of ‘Fairness through awareness’ (Dwork et al., 2012) whereby knowledge of protected attributes is used to explicitly mitigate (unconscious) bias (Leslie, 2019).…”

“…Artificial intelligence has captured the imagination of clinicians and researchers, funders, and commercial investors; but without widespread translation to clinical impact, we risk disillusionment and a collapse in willingness to invest resources. (Zhang et al., 2022)…”

“…However, if must be remembered that the AI will only be as good as the data made available to it. To achieve this, there must be appropriate infrastructure in place (Mayo Clinic Platform, 2021; Zhang et al., 2022) with a well-designed AI supply chain (Zhang et al., 2022), which includes many aspects outside of the main model-building phase (which currently holds most of the attention). This is well reported by Zhang et al.…”

“…To achieve this, there must be appropriate infrastructure in place (Mayo Clinic Platform, 2021; Zhang et al., 2022) with a well-designed AI supply chain (Zhang et al., 2022), which includes many aspects outside of the main model-building phase (which currently holds most of the attention). This is well reported by Zhang et al. (2022) and is needed to ensure both clinical impact and return on investment.…”

Digital public health technologies: rationale for a multidisciplinary research laboratory

“…The World Health Organization mentioned at least 12 applications of digital technologies in healthcare: "client education and behaviour change communication, sensors and point-ofcare diagnostics, registries and tracking of births and deaths, data collection and reporting, electronic health records, electronic decision support, provider-to-provider communication, provider work planning and scheduling, provider training and education, human resource approaches. The intense interest in AI applications in the healthcare sector is evidenced by the plethora of relevant studies and publications and the growth of investment and regulatory approvals (Zhang et al, 2022). Here we discuss the prospects for AI in public health and factors limiting its widespread and immediate adoption.…”