A rapid review exploring the effectiveness of artificial intelligence for cancer diagnosis

Background and Context

In Wales, and the UK generally, there has been a growing problem with NHS waiting lists, exacerbated by the COVID pandemic. A number of ways are being looked at to reduce these waiting times and facilitate an improvement in the successful delivery of cancer care.

Artificial Intelligence (AI) or ‘Machine Learning’ is being assessed and used to reduce waiting lists in many global healthcare systems, including eye care, cancer screening and medical diagnostics. The Welsh Government is assessing AI use in the Welsh NHS, with the intention to improve health outcomes and reduce waiting times and treatment backlogs. Initial efforts are being focused on cancer screening and diagnosis (especially radiology), as this will provide a quick return on the initial investment in the technology and have an immediate impact on cancer treatment pathways overall, as quicker diagnosis is connected to more positive treatment outcomes.

Aims

The Health and Care Research Wales Evidence Centre was asked to assess the evidence for the use of AI in cancer diagnostics, especially radiology. The Centre conducted a Rapid Review to give Welsh Government the ‘State of Play’, so that future policy and research in this area could be developed.

Strategy

A literature search was performed across a number of Clinical Trials and other databases using ‘AI in radiology for Cancer Diagnosis’.

The search period was from January 2019 to June 2023.

The strategy was split into two phases:

1. Initial Mapping
2. Selection of Relevant Studies                                   

Outcomes

After reviewing available literature, 28 studies were found to be relevant. 21 of these were ongoing trails.

Deeper investigation of these studies found that most had a low number of participants (less than 1000) and/or were conducted on experimental AI programs. 5 studies looked at the impact of AI on diagnostic time and clinical outcomes. These found that patients receiving a cancer diagnosis had, overall, a longer diagnostic timeline with AI methods than with non-AI methods. However, there is no existing ‘comparison study available to assess the impact of different AI programs.

Impact

The study has shown that there is an ongoing need for more research into the use of AI in cancer diagnosis, especially radiology. Further research on this needs to be specific to Wales to reflect the needs of Welsh cancer patients and screening participants.

The research will help the Welsh Government to understand the current usage of AI in cancer research and the surrounding evidence base to inform future Clinical Trials  and research efforts, in this area. This has the potential to benefit cancer patients and screening participants across the UK.

Author: Anthony Cope, PPG Member

Read the full report.