NICE is consulting on a proposal to let five AI tools help radiologists read prostate MRI scans while further NHS evidence is collected. The proposal is still draft guidance. It does not establish that every tool improves the whole diagnostic pathway, or that an AI result replaces the radiologist’s assessment. NICE consultation
The distinction matters for services considering adoption: finding a suspicious area on an image, reaching a diagnosis sooner and improving a patient’s outcome are connected questions, but they need different evidence.
What the five tools would do
NICE’s 23 October announcement names AI-Rad Companion Prostate MR, DeepHealth Prostate, ProstatID, Prostate Intelligence and QP-Prostate. They are intended to help identify potentially concerning areas on prostate MRI and support interpretation. A radiologist would continue to review the scan and the AI output. NICE announcement
The draft’s proposed route is use alongside evidence generation, with conditions attached. It is not a comparison naming a clinical winner among the five products. Nor does being named establish that a particular hospital has bought, integrated or started using one. Draft recommendation
That conditional route also differs from routine adoption supported by a settled evidence assessment. NICE’s proposed plan says companies would submit evidence after a two-year generation period, allowing the technologies to be assessed again. The plan itself is not a study protocol. Purpose of the evidence plan
A more accurate scan report is one part of the pathway
NICE identifies limited and inconsistent evidence comparing AI-assisted reporting with current NHS care. Its questions extend beyond whether the software marks a lesion correctly.
| Evidence question | What it helps establish |
|---|---|
| How does assisted reporting compare with current care? | Diagnostic performance in the intended service setting |
| What happens to biopsies and diagnosis timing? | Whether changes in reporting alter the next stages of care |
| How are clinically significant and less significant cancers detected? | The consequences of changing who receives further investigation |
| What changes in resource use and clinical decisions? | Effects on the pathway and inputs to cost-effectiveness analysis |
These are questions in the draft evidence-generation plan, rather than outcomes already demonstrated for all five tools.
For example, prioritising an MRI for review could shorten one queue while leaving a later biopsy wait unchanged. That is a pathway distinction, not a finding that this proposal has already reduced either wait. Our chest X-ray AI evaluation explainer examines this issue in a different imaging service; its results cannot be transferred to prostate MRI.
Whose scans are represented also matters
The plan asks for evidence across ethnic groups and in people whose anatomy, previous pelvic treatment, metalwork or other circumstances can affect MRI interpretation. It also identifies less common prostate cancers and treatments that alter the prostate’s appearance as relevant to subgroup assessment. Subgroup evidence gaps
An average performance estimate can therefore leave an implementation question unanswered. A service needs to know whether the evidence includes the people and scan circumstances it expects to encounter. This article does not assess any product’s compliance or suitability for a local deployment.
The next decision is still being consulted on
The consultation closes at 5pm on 12 November 2026. NICE explicitly says the recommendations may change before final guidance. Its project record lists the expected final publication date as to be confirmed. Consultation status
For now, the useful reading of the proposal is its evidence agenda: human-assisted interpretation, a defined period for collecting further data and specific gaps about patients, workflow and costs. The final wording and any subsequent local service availability remain separate developments to check.