Research Design
🔹 Integrated dermoscopy and clinical image data from public databases, published literature, and local hospitals in Fuzhou, Fujian Province.
🔹 Trained 18 deep learning models on public datasets and constructed a model ensemble.
🔹 Evaluated the diagnostic performance of the models in comparison with a panel of dermatologists.
🔹 Proposed a novel model, DCSNeXt, and validated its performance and cross-domain adaptability.
🔹 Designed a collaborative diagnostic paradigm combining multi-view explainability analysis, prediction results, saliency heatmaps, and dermatologist secondary assessment.
🔹 Developed an intelligent diagnostic platform for preliminary deployment.