Comparisons of deep learning algorithms for diagnosing bacterial keratitis via external eye photographs

Abstract

Bacterial keratitis (BK), a painful and fulminant bacterial infection of the cornea, is the most common type of vision-threatening infectious keratitis (IK). A rapid clinical diagnosis by an ophthalmologist may often help prevent BK patients from progression to corneal melting or even perforation, but many rural areas cannot afford an ophthalmologist. Thanks to the rapid development of deep learning (DL) algorithms, artificial intelligence via image could provide an immediate screening and recommendation for patients with red and painful eyes. Therefore, this study aims to elucidate the potentials of different DL algorithms for diagnosing BK via external eye photos. External eye photos of clinically suspected IK were consecutively collected from five referral centers. The candidate DL frameworks, including ResNet50, ResNeXt50, DenseNet121, SE-ResNet50, EfficientNets B0, B1, B2, and B3, were trained to recognize BK from the photo toward the target with the greatest area under the receiver operating characteristic curve (AUROC). Via five-cross validation, EfficientNet B3 showed the most excellent average AUROC, in which the average percentage of sensitivity, specificity, positive predictive value, and negative predictive value was 74, 64, 77, and 61. There was no statistical difference in diagnostic accuracy and AUROC between any two of these DL frameworks. The diagnostic accuracy of these models (ranged from 69 to 72%) is comparable to that of the ophthalmologist (66% to 74%). Therefore, all these models are promising tools for diagnosing BK in first-line medical care units without ophthalmologists.

Infectious keratitis: A review

“…Algorithms such as ResNet, DenseNet, ResNeXt, SENet, VGG and EfcientNet can potentially develop models for image diagnosis of BK. 54,57 A study from Thailand used three algorithms, DenseNet121, REstNet50, VGG19 to classify images of patients with infectious keratitis. The test accuracy (F1 score) was higher for VGG19 (78%) followed by DenseNet121 (71%) and REstNet50 (68%).…”

“…57 Investigators using external eye photographs to assess deep learning frameworks in BK have reported that the diagnostic accuracy of different models ranged from 69% to 72%; comparable to ophthalmologists (66% to 74%). 54 In areas or circumstances where patients are unable to access ophthalmic care, the ability to diagnose and assess microbial keratitis through artificial intelligence using external eye photos, such as could be taken with a mobile phone, may allow for appropriate therapy to be commenced without delay. 2,5,[54][55][56] 3 | HERPES SIMPLEX KERATITIS…”

“…54 In areas or circumstances where patients are unable to access ophthalmic care, the ability to diagnose and assess microbial keratitis through artificial intelligence using external eye photos, such as could be taken with a mobile phone, may allow for appropriate therapy to be commenced without delay. 2,5,[54][55][56] 3 | HERPES SIMPLEX KERATITIS…”

“…Deep learning algorithms are increasingly being recognised as having potential for screening and making management recommendations for patients with painful red eyes 54 ; distinguishing active corneal infection from scarring 55 ; and differentiating between causal organisms in keratitis 56 —for example between fungal and BK 57 . Convolutional neural networks apply very effectively deep learning for image classification.…”

“…Convolutional neural networks apply very effectively deep learning for image classification. Algorithms such as ResNet, DenseNet, ResNeXt, SENet, VGG and EfcientNet can potentially develop models for image diagnosis of BK 54,57 . A study from Thailand used three algorithms, DenseNet121, REstNet50, VGG19 to classify images of patients with infectious keratitis.…”