ASN's Mission

To create a world without kidney diseases, the ASN Alliance for Kidney Health elevates care by educating and informing, driving breakthroughs and innovation, and advocating for policies that create transformative changes in kidney medicine throughout the world.

learn more

Contact ASN

1401 H St, NW, Ste 900, Washington, DC 20005

email@asn-online.org

202-640-4660

The Latest on Twitter

Kidney Week

Abstract: SA-PO915

CT-Based Radiomic Feature Analysis for Identifying Baseline Kidney Function in Patients With Underlying CKD

Session Information

Category: CKD (Non-Dialysis)

  • 2202 CKD (Non-Dialysis): Clinical‚ Outcomes‚ and Trials

Authors

  • Jo, Seongho, Department of Internal Medicine, Inha University Hospital, Incheon, Korea (the Republic of)
  • Kim, Kipyo, Department of Internal Medicine, Inha University College of Medicine, Incheon, Korea (the Republic of)
Background

Decreased kidney size and cortical thickness were known as indicators for chronic changes in kidney, but the diagnostic accuracy was relatively low. Radiomics is a promising approach for quantitative analysis of various medical images. In this study, we investigated CT-based radiomic feature analysis for identifying baseline kidney function in patients with underlying chronic kidney disease (CKD).

Methods

A total of 487 patients who underwent a non-enhanced CT scan of the abdomen were included in the main analysis. Three-dimensional kidney segmentation was performed using semi-automated tools. A total of 1218 radiomic features were extracted using the PyRadiomics package, including shape features (n=14), original first order features (n=108), texture features (n=408), and wavelet features (n=688). We perform feature selection (n=20) using maximum-relevance minimum-redundancy methods. A prediction model for baseline kidney function was developed and validated using the XGBoost algorithm. For kidney biopsy findings within 3 months before and after CT scan, the association between selected radiomic features and chronic pathological findings was also examined.

Results

Conventional markers of CKD showed relatively low diagnostic accuracy (major axis length, AUC 0.75; kidney volume, AUC 0.82). The wavelet-HHH gldm dependence variance was the most predictive radiomic feature for eGFR <45ml/min/1.73m2 (AUC 0.85). A multivariable predictive model only with radiomic features revealed an improved performance (AUC 0.86). Finally, a model combined with clinical and radiomic features showed the best performance (AUC 0.87). Most selected radiomic features were highly correlated with baseline kidney function, and several wavelet transform features for texture showed a high AUC value for the presence of severe interstitial fibrosis and tubular atrophy.

Conclusion

Our findings indicate that CT-based radiomics feature analysis can provide a more accurate predictive model for CKD than traditional morphological markers.