Routine clinical test markers with MRI anatomical anchoring for dysphagia risk identification in wilson disease: model development and external validation using a two-center database

Wilson disease (WD)-related dysphagia may lead to aspiration, pulmonary infection and malnutrition, but it is often recognized late in routine care. This study evaluated the value of routine clinical test markers, magnetic resonance imaging (MRI) topography and their integration for earlier identification of dysphagia risk in WD.

We retrospectively included 318 patients with WD from the First Affiliated Hospital of Anhui University of Chinese Medicine between September 2019 and January 2026, including 130 patients with dysphagia (40.9%), for model development and internal validation. An independent cohort of 111 patients from the Institute of Neurology, Anhui University of Chinese Medicine during the same period, including 46 patients with dysphagia (41.4%), was used for external validation. Three predefined feature layers were assessed: MRI topography, routine clinical tests and an integrated layer. Multiple candidate algorithms were benchmarked within each layer.

MRI topography alone showed limited predictive ability. The routine clinical-test Light Gradient Boosting Machine (LightGBM) model achieved favorable screening performance in external validation (area under the receiver operating characteristic curve [AUROC] = 0.873, area under the precision-recall curve [AUPRC] = 0.855, negative predictive value = 0.844). Integrated models further improved risk confirmation. The integrated random forest model had the highest external AUROC and AUPRC, whereas the integrated LightGBM model had a low Brier score and stable risk-stratification performance. In the external validation cohort, observed dysphagia rates increased across integrated LightGBM risk strata from 10.9% to 47.8% and 87.9%.

Routine clinical test markers provide the main early screening signal for WD-related dysphagia. MRI topography alone has limited screening value but supports anatomical anchoring and high-risk confirmation. Combining clinical-test and MRI information may improve early dysphagia assessment, aspiration prevention and rehabilitation planning in hospitalized patients with WD.