What happened

The NHS Highland team built a two-step AI tool that reads MRI referral notes to group scans and then identifies the exact type of scan. In early results, AI performed well for data-rich scans, but brain and spine accuracy lagged behind staff. Any AI suggestion still required a radiographer to approve before scanning.

Why it matters

The project shows AI can extend radiology work by saving time and matching human decision quality on some tasks, though data limits and safety checks keep human review essential.

The NHS Highland project trained an AI with 7,275 MRI referrals and built a simple two-step tool. First, it reads notes to group scans by area; second, it chooses the exact scan. Early results are promising for data-rich types, but brain and spine scans still lag behind staff performance. The AI matched or approached human decision levels in some cases, and two radiographers often disagreed when choosing brain/spine protocols, suggesting the AI can be a helpful companion for clinicians.

Safety and next steps were kept at the forefront. Any AI suggestion needed a radiographer’s check before scanning, which protects patients while lowering workload over time. The team plans ongoing training with more data to refine capabilities.

What this does not tell us

Data availability and varying accuracy across scan types limit generalization beyond the NHS Highland setting.

FOR PEOPLE

Benefits reported

The tool aims to save time while keeping safety checks.

FOR AI AND ITS OPERATORS

Benefits reported

AI may support staff, but not yet replace human checks in scans.

These are two separate readings of what the sources describe. Reported claims and risks do not by themselves establish a real-world effect.

Original sources · 1
  1. Using AI to Speed Up MRI Scans in NHS Highland ↗NHS Highland · 2026-10-27

Reporting discovered in United Kingdom. Discovery market does not mean the event happened there.