AI-Assisted Fungal Detection in Dermatopathology: A Study on Accuracy, Efficiency, and Usability

AI-based image analysis is increasingly applied in pathology. Excluding fungal elements in PAS-stained skin sections is labor-intensive and well suited for AI assistance. While fungal detection in nails has been studied, skin-biopsy applications and real-world benefit remain limited. This study aimed to develop an AI algorithm for fungal detection in skin histology, create intuitive visualization software, and assess its diagnostic utility for pathologists.

A total of 466 PAS-stained cases from four institutions were digitised using three scanners. A lightweight UNet-based model was trained with semi-supervised learning and hard negative mining. Six dermatopathologists reviewed 204 independent cases in two phases: first without then with AI after a 4-week washout. Inter-observer agreement, diagnostic time, and AI-human concordance were recorded; ROC analysis assessed algorithm performance.

The AI demonstrated strong correlation with human assessment (ROC-AUC up to 0.949). At least one pathologist agreed with AI in 91.2% of cases. AI assistance increased Fleiss' kappa from 0.67 to 0.82, raised full agreement from 64% to 78%, and halved diagnostic time, especially in negative cases. Software-usability scored 9.5/10 and integration willingness 8.3/10.

AI assistance improved accuracy, reduced reading time, and offered user-friendly visualization, supporting integration into diagnostic workflows.