Pediatric CAKUT Diagnosis

This study evaluated whether deep transfer learning could help distinguish ultrasound images of kidneys from children with congenital abnormalities of the kidney and urinary tract (CAKUT) and healthy controls. Using images from 100 children, researchers built support vector machine classifiers based on transfer learning features, conventional imaging features, and their combination, finding that the combined approach performed best with AUCs up to 0.92 and strong accuracy, sensitivity, and specificity. These results suggest that pairing deep learning–derived features with traditional imaging measures can meaningfully improve automated classification of abnormal pediatric kidneys, though larger datasets are needed to confirm the findings.