IFTA Quantification

This single-center diagnostic study developed and validated a deep learning algorithm to noninvasively predict interstitial fibrosis and tubular atrophy (IFTA) grade from kidney ultrasound images, using biopsy-confirmed pathology as the reference standard. Trained on over 6,000 ultrasound images from 352 patients, the model achieved high segmentation accuracy (91%) and strong classification performance, with approximately 90% patient-level accuracy in the independent test set. The algorithm accurately predicted IFTA severity independent of clinical variables, suggesting ultrasound-based AI could provide a reliable, noninvasive alternative to biopsy for assessing chronic kidney damage.