AI Ultrasound for Kidney Fibrosis Detection - EMJ

This site is intended for healthcare professionals

AI-Assisted Ultrasonography Advances Non-invasive Kidney Assessment

Author:
AI

Key Summary:

  • AI-assisted ultrasonography was evaluated as a non-invasive method for detecting kidney fibrosis.
  • Analysis found 82.1% sensitivity, 82.8% specificity and 0.880 pooled AUROC.
  • Findings supported further prospective validation of AI-based kidney imaging.

AI-assisted ultrasonography demonstrated promising diagnostic performance in detecting moderate-to-severe interstitial fibrosis and tubular atrophy (IFTA) in patients with chronic kidney disease (CKD), potentially offering a non-invasive alternative to renal biopsy, according to findings from a systematic review and meta-analysis. 

The research evaluated the diagnostic accuracy of AI-supported ultrasound techniques, highlighting their potential to improve the assessment and monitoring of irreversible kidney damage. 

Evaluating AI-Assisted Ultrasonography 

IFTA is a common pathway of irreversible kidney injury and an important consideration in assessing CKD progression. Although renal biopsy remains the reference standard for evaluating IFTA, its invasive nature presents challenges, particularly when repeated assessments are required. 

To investigate alternative approaches, researchers searched five databases, including PubMed, Scopus, the Cochrane Library, Wanfang Database, and China National Knowledge Infrastructure, covering publications up to October 2025. 

Studies assessing AI-assisted B-mode ultrasonography, shear wave elastography (SWE), or combined ultrasound and SWE were included. All studies used renal biopsy as the reference standard and investigated moderate-to-severe IFTA, defined as involvement of at least 25%. 

The review included 14 studies involving 3,776 participants, with 1,244 contributing to the pooled diagnostic performance estimates. 

Diagnostic Accuracy Demonstrates Promise 

The meta-analysis found that AI-assisted ultrasonography achieved a summary sensitivity of 82.1% and specificity of 82.8%, indicating balanced performance in identifying moderate-to-severe IFTA. 

Furthermore, the pooled area under the receiver operating characteristic curve (AUROC) reached 0.880 (95% confidence interval: 0.815–0.924), demonstrating strong overall diagnostic discrimination. 

Subgroup analyses also identified numerically higher diagnostic performance with SWE and deep learning models. However, the researchers emphasised that these findings were exploratory, reflecting the limited number of studies and wide confidence intervals. 

These results suggest that AI-assisted ultrasonography could offer clinicians a valuable non-invasive approach to identifying significant kidney fibrosis, potentially supporting assessment without relying exclusively on invasive tissue sampling. 

Further Validation Needed 

Despite the encouraging findings, the researchers highlighted the need for further investigation before these technologies can be established in routine clinical practice. 

In particular, larger prospective studies will be necessary to validate diagnostic performance and clarify the relative contributions of different ultrasound techniques and AI architectures. 

The findings support the potential role of AI-assisted ultrasonography in the non-invasive assessment of kidney damage, particularly where repeated monitoring may be clinically valuable. Further validation could help establish its place alongside existing diagnostic approaches for patients with CKD. 

Reference 

Yin WJ et al. AI-assisted ultrasonography for noninvasive detection of moderate-to-severe interstitial fibrosis and tubular atrophy in chronic kidney disease: a systematic review and meta-analysis. Sci Rep. 2026;DOI:10.1038/s41598-026-80053-8.

Featured Image: Андрей Жерновой on Adobe Stock

Each article is made available under the terms of the Creative Commons Attribution-Non Commercial 4.0 License.

Rate this content's potential impact on patient outcomes

Average rating 0 / 5. Vote count: 0

No votes so far! Be the first to rate this content.