AI in vasculitis and kidney care
Is any of this in clinics yet? Short answer: these are research tools, mostly retrospective and not externally validated. None is meant for patients to use on their own.
Every medical claim on this page links to its source. Last updated 15 November 2026. Educational information, not medical advice.
This site has no AI symptom checker or chatbot, on purpose.
Deep learning to classify kidney biopsies in ANCA glomerulonephritis
Retrospective study
What it is: An AI model that detects and classifies glomerular lesions on biopsy slides, with heatmaps showing what it looked at.
Status: Biopsy slides from 80 patients at 3 European centers (Kidney Int Rep 2025). 1
What this means for you now: A possible decision-support tool for pathologists. Not in clinics.
Hope vs hype: Small, retrospective; needs external validation.
Last checked 15 Nov 2026
KidneyVision pathomics pipeline
Retrospective study
What it is: Software that measures kidney biopsy features across several glomerular diseases, including AAV.
Status: Conference abstract (Lab Invest 2026). 2
What this means for you now: Research measurement tool, not diagnosis.
Hope vs hype: Abstract only.
Last checked 15 Nov 2026
Relapse risk model from PEXIVAS
Retrospective study
What it is: A statistical model estimating 2-year risk of first relapse after remission. (A statistical model, not "AI" in the popular sense.)
Status: 649 trial participants; optimism-corrected area under the curve about 0.71 (Ann Rheum Dis 2026). Authors say it needs external validation. 3
What this means for you now: Not ready for personal use.
Hope vs hype: Moderate discrimination; internal validation only.
Last checked 15 Nov 2026
Patient-reported outcomes plus a relapse risk score
Retrospective study
What it is: Adding quality-of-life questionnaire results to an existing clinical relapse score.
Status: 98 people with MPA or GPA; combined area under the curve 0.727 (Clin Exp Rheumatol 2026). 4
What this means for you now: Interesting idea: how you feel may add information.
Hope vs hype: Small, single cohort.
Last checked 15 Nov 2026
Kidney outcome prediction (AKRiS, CT radiomics)
Retrospective study
What it is: Scores and image-analysis models predicting kidney failure risk in ANCA kidney disease.
Status: ANCA Kidney Risk Score validated in a German cohort (C-index about 0.84, Kidney360 2024); a single-center CT radiomics model (Front Digit Health 2026). 5,6
What this means for you now: Research tools your team may know about; not for self-use.
Hope vs hype: Promising; needs prospective validation.
Last checked 15 Nov 2026
AI trial matching (TrialGPT, NIH)
Research tool
What it is: A large-language-model system that matches patient descriptions to trial eligibility criteria.
Status: Tested on synthetic patients: 87.3% criterion-level accuracy, and screening time cut by 42.6% in a user study (Nat Commun 2024). 7,8
What this means for you now: Clinician-facing research tool; not vasculitis-specific.
Hope vs hype: General purpose, not tested in vasculitis.
Last checked 15 Nov 2026
Computational drug repurposing
Preclinical / lab / computational
What it is: Using genetics and immune profiling to find existing drugs that might work in vasculitis.
Status: Shared vasculitis risk genes pointing to repurposing targets (Ann Rheum Dis 2023); immune-profiling-guided ustekinumab (Nat Commun 2024). 9,10
What this means for you now: Hypotheses only.
Hope vs hype: Ideas to test, not treatments.
Last checked 15 Nov 2026
Questions to ask about any AI health claim
- Was it tested on people like me, at a different hospital from where it was built?
- Was it tested forward in time (prospectively)?
- Did it change decisions or outcomes, or only predict?
- Who is responsible if it is wrong?
Sources
- Kidney International Reports (2025): Explainability of a deep learning-based classification model for ANCA-associated glomerulonephritis
- Laboratory Investigation (2026, abstract): Pathomics pipeline for medical kidney biopsy with KidneyVision
- Annals of the Rheumatic Diseases (2026): Estimating 2-year risk of first relapse in AAV: risk-prediction model from PEXIVAS participants
- Clinical and Experimental Rheumatology (2026): Patient-reported outcomes in a clinical risk score to improve relapse prediction in AAV
- Kidney360 (2024): ANCA Kidney Risk Score performance in a German cohort
- Frontiers in Digital Health (2026): Radiomics-based prediction of renal prognosis in AAV
- Nature Communications (2024): Matching patients to clinical trials with large language models (TrialGPT)
- NIH (news release): NIH-developed AI algorithm matches potential volunteers to clinical trials
- Annals of the Rheumatic Diseases (2023): New risk loci shared across systemic vasculitides point to potential drug-repurposing targets
- Nature Communications (2024): Immune profiling-based targeting of pathogenic T cells with ustekinumab in ANCA-associated glomerulonephritis
Links checked 15 November 2026. Full list: all sources.