White paper reveals algorithm matches radiologists in 93.5 per cent of acute CT head scans

Medica and Qure.ai have published a white paper showing a 93.5 per cent overall agreement between radiologists and Qure.ai’s qER algorithm in analysing 1,315 non-contrast CT head scans with acute presentation histories. Medica says this highlights AI’s reliability for bleed detection and prioritisation, although areas such as subtle subarachnoid haemorrhage detection need improvement. Medica hosted an event presenting these findings, chaired by chief medical officer Dr Robert Lavis, which marked its 20th anniversary.

Consultant interventional radiologist Dr Neelan Das from East Kent Hospitals University NHS Foundation Trust shared five key steps for AI deployment success: identifying needs, engaging stakeholders, setting goals, pre-deployment validation and ongoing surveillance. He emphasised understanding algorithm performance in specific populations.

Radiology consultant and AI lead Dr Sarim Ather from Oxford University Hospitals NHS Foundation Trust discussed the importance of post- deployment monitoring and AI’s potential in screening for conditions like osteoporosis and coronary artery calcification, and highlighted the need for rigorous real-world assessments.

Picture: Medica hosted an event presenting the findings of a white paper.

See the full report on page 19 of the August 2024 issue of RAD Magazine.