ACCORDING TO new research presented recently at the European Congress of Radiology (ECR) 2026, AI-assisted CT could streamline cancer follow-up imaging by reducing reading time while improving consistency in tumour measurement. These new findings suggest that AI-assisted CT may help standardise radiology assessments and reduce disagreements in treatment response evaluation.
AI-Assisted CT and the Challenge of Consistent Tumour Measurement
Accurate tumour measurement is critical when evaluating treatment response in oncology imaging. Radiologists typically rely on RECIST 1.1 criteria to measure lesions on follow-up CT scans, but variability between readers can affect clinical decisions. AI-assisted CT tools are increasingly being explored to improve efficiency and consistency during image interpretation.
By automatically identifying and measuring target lesions, AI-assisted CT may help reduce workload and minimise subjective differences between readers. Researchers therefore investigated whether AI-assisted CT could shorten reading time while improving agreement between radiologists assessing tumour response in follow-up imaging studies.
Multi-Centre Evaluation of AI-Assisted CT Performance
In this retrospective multi-centre reader study, 23 readers (15 radiologists, 8 residents) evaluated follow-up chest-abdomen-pelvis CT scans from 212 oncology patients, with 539 lesions in total. Each scan was assessed under three conditions: unassisted, AI-assisted, and expert-assisted, with readers informed that all support originated from AI to prevent bias.
AI-assisted CT significantly reduced reading time per patient compared with unassisted reading (–35.96 s; 95% CI: –52.95, –22.07). At the lesion level, inter-reader variability relative to the consensus standard increased slightly with AI assistance (1.32 mm; 95% CI: 0.83, 1.91).
However, at the clinically relevant patient level, variability decreased substantially. Differences in the Sum of Longest Diameters of more than 20% between radiologists occurred in 43.4% of unassisted cases, 28.3% of AI-assisted cases, and 17.4% of expert-assisted cases. These findings were consistent across experience levels.
Clinical Implications for AI-Assisted CT in Oncology Imaging
The results suggest that AI-assisted CT may reduce clinically meaningful disagreement between radiologists when assessing treatment response in cancer follow-up scans. Although lesion-level variability increased slightly, overall patient-level consistency improved.
Combined with the substantial reduction in reading time, AI-assisted CT could help standardise follow-up assessments and improve efficiency in radiology workflows. More consistent measurements may also reduce discordant treatment decisions in oncology practice.
Researchers note that further validation across imaging systems and international patient populations will be important. Continued development of AI-assisted CT tools could strengthen decision support in radiology and support more reliable monitoring of cancer therapy outcomes.
Reference
Hering A et al. Validation of AI-Assisted RECIST lesion measurements in follow-up CT: a multi-center reader study. ECR, 4-8 March, 2026.
Featured image: Maryna on Adobe Stock
