The potential role of artificial intelligence in the diagnosis of Acanthamoeba keratitis-a scoping review.

BACKGROUND

Acanthamoeba keratitis (AK) is a severe, sight-threatening infectious disease with a rising global incidence, linked to the increasing use of contact lenses. Clinical diagnosis is complicated by the nonspecific nature of early symptoms and limitations in current diagnostic methods. Recently, Artificial Intelligence (AI) and Deep Learning (DL) have been applied to ocular imaging to aid in diagnosis.

OBJECTIVE

The primary objective of this scoping review is to evaluate the diagnostic performance of AI and DL algorithms for AK across different imaging modalities and to identify the methodological gaps necessary to apply these methods in a clinical setting.

METHODS

This review followed the PRISMA for Scoping Reviews guidelines. A systematic search was conducted on PubMed, Embase, Web of Science, IEEE Xplore and the Cochrane Library, to identify original research applying AI/DL for AK diagnosis in humans using ocular imaging, published up to October 28, 2025.

RESULTS

Nine studies using DL algorithms were included. Imaging modalities included slit-lamp photography (n = 4), in vivo confocal microscopy (IVCM) (n = 4), and smartphone-captured images (n = 1). In internal validation, models demonstrated high diagnostic capability, with Area Under the Curve (AUC) values frequently exceeding 0.90. Slit-lamp models generally achieved high specificity (>94%) with variable sensitivity (68.3%-89.2%). IVCM models reported more varying sensitivities ranging from 55.0% to 96.0%. The smartphone study reported a strong AUC of 0.99. Only three studies performed external validation.

CONCLUSION

The use of AI and deep learning algorithms show promise in diagnosing AK. Internal performance metrics vary but are sometimes comparable to current diagnostic methods. However, real-world applicability is currently limited by small sample sizes, heterogeneity in available data, and scarcity of external validation. Future research must prioritize multi-center validation and standardized methodologies to ensure clinical utility.