A combination of unenhanced CT-derived features predicts renal parenchymal involvement of immunoglobulin G4-related...

To predict renal parenchymal involvement in patients with immunoglobulin G4-related disease (IgG4-RD) using texture- and volume-based analysis of unenhanced computed tomography (CT) images.

Fifty-nine patients with IgG4-RD, diagnosed according to American College of Rheumatology/ European Alliance of Associations for Rheumatology classification criteria, who underwent both unenhanced CT and contrast-enhanced (CE)-CT) covering the kidneys before treatment (May 2007-July 2023), were enrolled. After excluding 15 cases, 44 patients remained for analysis. Clinical variables (sex, age, body weight, estimated glomerular filtration rate) and kidney volume-related and texture metrics extracted from unenhanced CT were analyzed for inter-reader repeatability and correlation with renal parenchymal involvement.

Kidney involvement of IgG4-RD was determined on CE-CT by a rheumatologist and a radiologist in consensus, resulting in 23 cases with and 21 cases without renal parenchymal involvement. One radiologist and a radiology resident independently extracted metrics from unenhanced CT. Two metrics, kidney volume per body weight and gray-level co-occurrence matrix (GLCM) cluster shade, showed high repeatability (intraclass correlation coefficients 0.997 and 0.762, respectively) and significant association with renal parenchymal involvement (area under the receiver operating characteristic curve [AUROC] > 0.75). Binary logistic regression combining these two metrics predicted renal parenchymal involvement with AUROCs of 0.872 and 0.896 for the two readers, respectively.

A combination of kidney volume per body weight and GLCM cluster shade derived from unenhanced CT can predict renal parenchymal involvement in IgG4-RD.

The combination of two unenhanced CT-related metrics-kidney volume per body weight and GLCM cluster shade-can predict renal parenchymal involvement of immunoglobulin G4-related disease, with the AUROCs for the two readers being 0.872 and 0.896, respectively.

Unenhanced CT has been considered not to detect renal parenchymal involvement of IgG4-RD. By using image segmentation and texture analysis techniques, renal parenchymal involvement could be predicted in IgG4-RD. A combination of kidney volume per body weight and GLCM cluster shade could predict renal parenchymal involvement with high AUROC.