Deep Learning for the Discrimination Among Different Types of Keratits: a Nationwide Study

Ningbo Eye Hospital

Status: Completed

Conditions: Automatic Judgement; Image; Keratitis

Study type: Observational

Funder types: Other

Identifiers

NCT05538793

NEH2022091015

Details and patient eligibility

About

Detecting the cause of keratitis fast is the premise of providing targeted therapy for reducing vision loss and preventing severe complications. Due to overlapping inflammatory features, even expert cornea specialists have relatively poor performance in the identification of causative pathogen of infectious keraitis. In this project, the investigators aim to develop an automated and accurate deep learning system to discriminate among bacterial, fungal, viral, amebic and noninfectious keratitis based on slit-lamp images and evaluated this system using the datasets obtained from mutiple independent clinical centers across China.

Enrollment: 10,369 patients

Sex: All

Ages: 1 week to 100 years old

Volunteers: No Healthy Volunteers

Inclusion criteria

Slit-lamp images with sufficient diagnostic certainty and showing keratitis at the active phase.

Exclusion criteria

• Poor-quality images
• Images presenting mixed infections (i.e., cornea infected by two or more causative pathogens)