ASDC-NET: Anterior Segment Disease Classification Network Based on Slit-Lamp Images

Anterior segment diseases are common in clinical ophthalmology, but most studies focus on individual conditions, with relatively few addressing the broader classification of these diseases. In this paper, we propose an anterior segment disease classification network (ASDC-NET) to classify 5 types of diseases based on slit-lamp images. The proposed method designs a wavelet feature enhancement (WFE) module to extract frequency domain information, a multi-scale feature fusion (MSFF) module to combine features at different scales, and an adaptive channel aggregation (ACA) module to adapt to lesion variability. Evaluation results on the dataset including 1,250 slit-lamp images demonstrate the excellent classification performance of the proposed method, outperforming other state-of-the-art methods.