Table 3.

Performance of the Traditional Machine-Learning Model on The Ohio State University Wexner Medical Center Data Set

Description	Minimal	Mild	Moderate	Severe
Precision	0.12 ± 0.15	0.25 ± 0.13	0.22 ± 0.12	0.29 ± 0.18
Recall/sensitivity	0.10 ± 0.13	0.40 ± 0.17	0.27 ± 0.13	0.23 ± 0.17
Specificity	0.89 ± 0.09	0.59 ± 0.14	0.67 ± 0.12	0.86 ± 0.07

A machine-learning model based on weighted neighbor distance using compound hierarchy of algorithms representing morphology was constructed by deriving approximately 3000 features from the whole-slide image data obtained from The Ohio State University Wexner Medical Center to predict interstitial fibrosis and tubular atrophy grade. The trained model was then used to predict on the data obtained from the Kidney Precision Medicine Project. Performance of the model after fivefold cross validation on The Ohio State University Wexner Medical Center data set is shown. Data are expressed as means SD.