Key result
The EGPA-ML tool, an SVM-based machine learning model, achieved high diagnostic performance for eosinophilic granulomatosis with polyangiitis, demonstrating a recall of 0.992 and an F1-score of 0.926.
Why the study?
EGPA is often diagnosed late due to heterogeneous presentation, leading to severe complications including cardiac involvement, prompting the development of a machine learning tool to support early and accurate diagnosis.
Does EGPA-ML accurately diagnose eosinophilic granulomatosis with polyangiitis in patients with suspected vasculitis?
Cohort
No
Does EGPA-ML accurately diagnose eosinophilic granulomatosis with polyangiitis in patients with suspected vasculitis?
A novel machine learning tool (EGPA-ML) demonstrated high diagnostic performance for eosinophilic granulomatosis with polyangiitis, potentially aiding early diagnosis in non-specialized settings.
May aid early EGPA diagnosis outside specialty centers; leaves open prospective validation before clinical use.
Eosinophilic granulomatosis with polyangiitis (EGPA), formerly Churg-Strauss syndrome, is a rare systemic vasculitis often diagnosed late due to its heterogeneous presentation, leading to severe complications—particularly cardiac involvement, a major cause of morbidity and mortality. We developed EGPA-ML, an artificial intelligence (AI)-based tool using supervised machine learning (ML), to support early and accurate EGPA diagnosis, especially in non-specialized settings. A retrospective cohort of patients evaluated for suspected vasculitis at Hedi Chaker Hospital, Sfax, Tunisia, from 1997 to 2023 (nearly three decades), provided 1,904 clinical, biological, and histological features. After data cleaning, standardization, and feature selection, 56 key features were retained. Patients were classified as {EGPA} or {NOT_EGPA} per the 2022 ACR/EULAR criteria, with expert consensus (κ = 0.85). Multiple supervised ML algorithms were evaluated via 10-fold cross-validation. The best model was integrated into EGPA-ML, a Java-based clinical decision support system. Performance was assessed on an independent dataset of n = 280 key features, with reference classification {EGPA}/{NOT_EGPA} validated by experts (κ = 0.89). On the test and evaluation dataset, EGPA-ML achieved a recall of 0.992, precision of 0.869, and F1-score of 0.926. Feature importance analysis identified asthma and eosinophil count as top predictors (36.5% each), followed by ANCA status, vascular purpura, and histological vasculitis. EGPA-ML is a high-performance, interpretable, and adaptive tool based on supervised ML, supporting timely EGPA diagnosis. It represents a practical advancement for clinical decision-making in rare diseases, particularly in internal medicine, pulmonology, and cardiology. • A novel SVM-based tool (EGPA-ML) supports early diagnosis of eosinophilic granulomatosis with polyangiitis. • Trained on 1,904 features from a 26-year retrospective cohort (1997–2023) at Hedi Chaker Hospital, Sfax. • Validated on an independent test set of 280 clinical, biological, and histological features. • Achieved high performance: recall = 0.992, precision = 0.869, F1-score = 0.926. • Top predictors align with clinical criteria: asthma, eosinophilia, ANCA status, purpura, vasculitis.
No takes yet. Share an insight, caveat, or question.
Bouattour et al. (2025) conducted a cohort in Eosinophilic granulomatosis with polyangiitis (EGPA). EGPA-ML vs. Expert consensus was evaluated on Diagnostic performance (recall, precision, F1-score) (Recall 0.992, Precision 0.869, F1-score 0.926). The EGPA-ML tool, an SVM-based machine learning model, achieved high diagnostic performance for eosinophilic granulomatosis with polyangiitis, demonstrating a recall of 0.992 and an F1-score of 0.926.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: