Acanthamoeba and Artificial Intelligence

Acanthamoeba Keratitis, Artificial Intelligence

Acanthamoeba, Keratitis, Artificial intelligence, Corneal scraping, Confocal images, Acanthamoeba pattern, Machine learning

Acanthamoeba keratitis, caused by the pathogen Acanthamoeba spp, is recognized worldwide as a severe ocular infection that can pose potential risks to vision. This observational retrospective and single-center study, of exploratory nature, aims to determine the possibility of identifying patterns that may be useful for future rapid diagnosis of Acanthamoeba keratitis from confocal images, leveraging the normality of corneal examination and the high specificity and sensitivity of computational models. The data will be based on patients who have been confirmed positive through laboratory tests with proven effectiveness in detecting the infection. The laboratory tests considered for the division of patients into their respective groups are bacterial examination, PCR examination, and culture examination. Patients were divided into two groups, the first comprising patients positive for Acanthamoeba infection, while the second comprised patients negative for Acanthamoeba but positive for other pathogens. The study will last for 18 months. The cohort under study includes 151 patients from the IRCCS San Raffaele Hospital who underwent the aforementioned examinations, of which 76 cases will be included in the group of patients positive for Acanthamoeba and 75 in the group of controls positive for other pathogens. The confocal images of this cohort will be fed into artificial intelligence software. To evaluate the model, the test set will be used, and the AI model's ability will be assessed using the most commonly used metrics in the field of computer vision such as accuracy, specificity, sensitivity, and f1-score; culminating in a comprehensive evaluation of the model.

No related papers found yet.

None listed