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


The Latest on X

Kidney Week

Please note that you are viewing an archived section from 2023 and some content may be unavailable. To unlock all content for 2023, please visit the archives.

Abstract: TH-PO031

Ultrasound-Based Imaging Methods for the Assessment of Kidney Fibrosis on Artificial Intelligence

Session Information

Category: Augmented Intelligence, Digital Health, and Data Science

  • 300 Augmented Intelligence, Digital Health, and Data Science


  • Zhao, Dan, Shanghai Tongji Hospital, Shanghai, Shanghai, China

Renal fibrosis is the common pathway leading to end stage kidney disease, but the assessment of renal fibrosis remains limitations.
Artificial intelligence (AI)-based ultrasound imaging has showed high specificity and sensibility in disease diagnosis via image recognition. Hereby, we intend to use artificial intelligence technology to explore a new method for early non-invasive assessment of renal fibrosis through ultrasound images.


A retrospective cohort of patients receiving kidney biopsy was developped from our hospital between January 2018 and July 2022. All participant underwent kidney ultrasound within one week. The magnitude of renal intersitial fibrosis (RIF) was evaluated by two experienced pathologists according to Banff criteria. All ultrasound images were preprocessed, then randomly divided into training and test sets in a 7 to 3 ratio. Radiomic features extracted from selected regions of interest (ROI), including firstorder, texture, and wavelet-transformed features. Two-sample t-test, LASSO, and Spearman correlation analysis were used to reduce feature dimension. Three machine learning models and one deep learning model were established to distinguish mild and moderate-severe fibrosis. The area under the curve (AUC) of the receiver operating characteristic curve was employed to assess the models' performance.


A total of 193 paients were enrolled in this study, including 100 with mild fibrosis and 93 with moderate-severe fibrosis. 837 radiomic features were extracted from gray scale ultrasound images. After feature dimension reduction, 10 features were retained, of which 9 were wavelet-transformed features. Based on 10 selected radiomic features, three machine learning models were established: LR, SVM and RF. In test sets, the AUCs were 0.75 ,0.80, and 0.89, respectively. The accuracy were 60%, 63.7%, and 65% respectively. RF model achieved the best classification performance. For further visualization, LR prediction model were visualized as nomogram. In addition, based on full kidney ultrasound images, VGG16 network model were also establisehed, with AUC of 0.76 (95% CI, 0.63-0.90) and the accuracy of 77%.


AI-assisted imaging analysis endows traditional ultrasound techniques with the promising potential for early diagnosis of renal fibrosis. Our study would provide new insights for the future exploration of noninvasive assessment methods.