Learning models such as DeepKeratitis may eventually reduce the need for cornea specialist consultations and promote telehealth opportunities. Photo: Suzanne Sherman, OD. Click image to enlarge.

Microbial keratitis is an urgent condition that requires prompt treatment. In an attempt to apply deep learning algorithms to help rapidly discriminate between fungal keratitis (FK) and bacterial keratitis (BK), researchers recently found that a convolutional neural network with ensemble learning called DeepKeratitis had the best results compared with single architecture models.

The team evaluated 2,167 anterior segment images retrospectively acquired from 194 patients, with 128 patients having BK (1,388 images, 64.1%) and 66 having FK (779 images, 35.9%), to develop their model. The images were split into training, validation and test sets. Three convolutional neural networks were trained to classify images. Model performance was evaluated using precision (positive predictive value), sensitivity (recall), F1 score (test accuracy) and area under the precision-recall curve (AUPRC). Ensemble learning was then applied to improve classification performance.

“For clinical applications, our convolutional neural network models, particularly ensemble learning, are potentially helpful for clinicians in discriminating between BK and FK,” the study authors noted. “They can be considered as an adjunctive tool that provides rapid provisional diagnosis apart from laboratory tools and can provide several benefits for high-yield, noninvasive, and rapid response (within a few minutes).” 

Because network models can be applied simply and in most clinical settings if configured using online interfaces or mobile applications, the team believes their deep learning prototype could assist clinicians especially in developing countries, where specialists and laboratory facilities are limited. The approach could reduce the number of unnecessary cornea specialist consultations and facilitate telehealth consultation.