Chronic Changes on Kidney Histology by a Multiclass Artificial Intelligence Model

Abstract

Introduction: Chronic changes in kidney histology are often approximated by using human vision but with limited accuracy.

Methods: An interactive annotation tool trained an artificial intelligence (AI) model for segmenting structures on whole slide images (WSIs) of kidney tissue. A total of 20,509 annotations trained the AI model with 20 classes of structures, including separate detection of cortex from medulla. We compared the AI model detections with human-based annotations in an independent validation set. The AI model was then applied to 1426 donors and 1699 patients with renal tumor to calculate chronic changes as defined by measures of nephron size (glomerular volume, cortex volume per glomerulus, and mean tubular areas) and nephrosclerosis (globally sclerotic glomeruli, increased interstitium, increased tubular atrophy (TA), arteriolar hyalinosis (AH), and artery luminal stenosis from intimal thickening). We then assessed whether chronic kidney disease (CKD) outcomes were associated with these chronic changes.

Results: During the AI model validation step, the agreement between the AI detections and human annotations was similar to the agreement between human pairs, except that the AI model showed less agreement with AH. Chronic changes calculated solely from AI-based detections associated with low glomerular filtration rate (GFR) during follow-up after kidney donation and with kidney failure after a radical nephrectomy for tumor. A chronicity score based on AI detections was calculated from cortex per glomerulus, percent glomerulosclerosis, TA foci density, and mean area of AH lesions and showed good prognostic discrimination for kidney failure (cross-validation C-statistic = 0.819).

Conclusion: A multiclass AI model can help automate quantification of chronic changes on WSIs of kidney histology.

Keywords: artificial intelligence; chronic kidney disease; digital pathology; kidney histology; nephron size; nephrosclerosis.

Graphical abstract

Figure 1

Flowchart shows the selection of model development and outcome sets for living kidney donors and patients with a renal tumor. eGFR, estimated glomerular filtration rate; QC, quality check

Figure 2

The multilayer 20 class AI model with examples of detections for each class. AI, artificial intelligence.

Figure 3

An example of image analysis and a pathology report of chronic changes in a 71-year-old patient with a renal tumor. (a) The original PAS-stained section was (b) overlaid with AI-detected cortex, medulla, and 3 glomerular classes, and tubular atrophy clusters (TA) (light green) and (c) all arteries and arterioles (red). (d) An example of a nonsclerosed glomerulus (cyan), globally sclerosed glomerulus (dark blue) and empty capsule (dark green). (e) An enlarged example of TA (light green), interstitium (light blue), and tubule area (brown). (f) An example of an artery (red) with intimal thickening (green) and lumen (yellow). (g) An example of an arteriole (red) with arteriolar hyalinosis (AH) lesion (dark green). (h) A pathology report of AI-calculated measures of nephron size and nephrosclerosis that includes upper reference limits (based on the living kidney donors) and patient’s value. AI, artificial intelligence; PAS, Periodic acid Schiff.

Figure 4

An example of the whole and magnified views of the same PAS-stained wedge section with AI detections used to calculate measures of tubular atrophy. (a) A clean image shown without AI detections. (b) AI-derived detections of tubular atrophy clusters (TA) (green) and all tubules (brown) that were used to calculate %TA per tubular area. (c) AI-derived detections of interstitium and TA cluster (ITA) (blue), and all tubules (brown) that were used to calculate %ITA per tubulointerstitial area. AI, artificial intelligence; PAS, Periodic acid Schiff.

Figure 5

Risk of progressive CKD and kidney failure from 4 months postnephrectomy based on 2 chronicity scores. Risk of progressive CKD and kidney failure based on severity of Nephrosclerosis Chronicity Score (a and b), Nephron Hypertrophy and Nephrosclerosis Chronicity Score (c and d). Nephrosclerosis Chronicity Score was obtained by summing the %GSG score (0–3), %ITA score (0–3), TA density score (0–3), and AH area score (0–3). Nephron Hypertrophy and Nephrosclerosis Chronicity Score was obtained by summing the cortex per glomerulus score (0–3), %GSG score (0–3), TA density score (0–3), and AH area score (0–3). Each of the 4 variables (0–3) was summed, but the highest chronicity score among these patients was 9 (not 12). The chronicity scores (0–9) were displayed across three levels (0–3, 4–6, and 7–9). %GSG, percent globally sclerosed glomeruli; AH, arteriolar hyalinosis; CKD, chronic kidney disease; %ITA, percent interstitium and TA cluster; TA, tubular atrophy.