Automated AI-Driven Measurement of Kidney Length in Ultrasound Images: A Validation Study in Adult and Pediatric Groups

Accurate kidney size assessment is critical for managing renal diseases, but manual ultrasound measurements suffer from inter-observer variability. This study evaluates an artificial intelligence (AI) system that automates kidney length measurement, comparing its performance against manual measurements by experienced radiologists. We collected 1200 sagittal ultrasound images from two academic referral centers. A subgroup analysis was also performed on 73 images from pediatric subjects. Kidney length was measured by both the AI system and two radiologists. Performance was evaluated using the intraclass correlation coefficient (ICC) and Bland-Altman analysis. The AI system showed excellent agreement with radiologists, with ICCs of 0.94–0.98 for right kidneys and 0.96–0.98 for left kidneys. This high level of agreement was maintained in the pediatric subgroup (ICC > 0.97). Additionally, the AI system achieved a mean absolute error (MAE) of 0.32 cm, root mean square error (RMSE) of 0.5 cm, coefficient of determination (R²) of 0.86, and concordance correlation coefficient (CCC) of 0.91 for those samples that used computed tomography (CT) as the gold standard. This AI system offers superior performance to expert radiologists across both adult and pediatric populations. Its integration into clinical workflows could reduce measurement variability and enhance diagnostic accuracy in renal assessments.