iToBoS

Introduction

As rates of melanoma – one of the most aggressive skin cancers – rise, the labour-intensive screening process which manually analyses each mole on a patient’s body stands as a potential barrier to early diagnosis and intervention. The iToBoS project sought to remedy this by developing a bespoke AI-powered total body scanner, which captures images of a patient’s entire body and applies machine learning to identify potentially cancerous moles. The scanner is accompanied by a Computer Aided Diagnostics (CAD) tool, which integrates a range of data sources to analyse the images, and an AI cognitive assistant to support healthcare professionals by providing a risk assessment for each mole. Alongside these tools, the project pushed forward explainability techniques, providing visualisations and interpretations of the AI health monitoring tools’ outputs to support the production of transparent, trustworthy results.

Trilateral’s role

Trilateral Research’s role in the project included:

Assessing the ethical and social impacts associated with AI-driven public health strategies, including the impact of medical AI on trust between patients and clinicians.
Enabling the deployment of custom AI healthcare solutions by monitoring privacy and data protection issues, producing guidebooks for partners explaining regulations such as the GDPR, ethical and social issues such as dignity, informed consent, and equity, as well as by conducting privacy and data impact assessments of the developed tools.
Supporting the development and integration of explainability features into the AI tools to enable quality control, build trust, decrease uncertainty, and improve the transparency of “black box” algorithms.
Drafting socio-cultural and ethical guidelines for the development of future AI solutions for public health management, drawing on lessons learned from the project to assist future users of new technologies and encourage trust.

Project impact

The iToBoS project generated the following impacts:

Improved the efficiency of public health and diagnostics systems by developing AI solutions to deliver personalised healthcare and improve diagnostic accuracy, thus reducing the disease burden of melanoma
Supported the digital transformation of public health services by addressing and issuing forward-looking guidance about managing ethical, legal, data security, and privacy concerns related to the use of AI in medicine
Fostered the uptake of custom AI solutions by medical professionals by incorporating explainability features, visualisations, and novel insights into a user-friendly dashboard for clinicians
Employed interdisciplinary techniques – considering technical, scientific, and social science insights – to build trust in medical AI among professionals, patients, and other stakeholders, addressing a barrier to the adoption of bespoke AI in the medical field
Experimented with and deployed novel explainable AI techniques, pushing forward innovation in this area