Automatic Classification of Slit-Lamp Photographs by Imaging Illumination.
Journal
Cornea
ISSN: 1536-4798
Titre abrégé: Cornea
Pays: United States
ID NLM: 8216186
Informations de publication
Date de publication:
01 Jun 2023
01 Jun 2023
Historique:
received:
16
03
2023
accepted:
25
04
2023
pmc-release:
01
12
2024
medline:
2
6
2023
pubmed:
2
6
2023
entrez:
2
6
2023
Statut:
aheadofprint
Résumé
The aim of this study was to facilitate deep learning systems in image annotations for diagnosing keratitis type by developing an automated algorithm to classify slit-lamp photographs (SLPs) based on illumination technique. SLPs were collected from patients with corneal ulcer at Kellogg Eye Center, Bascom Palmer Eye Institute, and Aravind Eye Care Systems. Illumination techniques were slit beam, diffuse white light, diffuse blue light with fluorescein, and sclerotic scatter (ScS). Images were manually labeled for illumination and randomly split into training, validation, and testing data sets (70%:15%:15%). Classification algorithms including MobileNetV2, ResNet50, LeNet, AlexNet, multilayer perceptron, and k-nearest neighborhood were trained to distinguish 4 type of illumination techniques. The algorithm performances on the test data set were evaluated with 95% confidence intervals (CIs) for accuracy, F1 score, and area under the receiver operator characteristics curve (AUC-ROC), overall and by class (one-vs-rest). A total of 12,132 images from 409 patients were analyzed, including 41.8% (n = 5069) slit-beam photographs, 21.2% (2571) diffuse white light, 19.5% (2364) diffuse blue light, and 17.5% (2128) ScS. MobileNetV2 achieved the highest overall F1 score of 97.95% (CI, 97.94%-97.97%), AUC-ROC of 99.83% (99.72%-99.9%), and accuracy of 98.98% (98.97%-98.98%). The F1 scores for slit beam, diffuse white light, diffuse blue light, and ScS were 97.82% (97.80%-97.84%), 96.62% (96.58%-96.66%), 99.88% (99.87%-99.89%), and 97.59% (97.55%-97.62%), respectively. Slit beam and ScS were the 2 most frequently misclassified illumination. MobileNetV2 accurately labeled illumination of SLPs using a large data set of corneal images. Effective, automatic classification of SLPs is key to integrating deep learning systems for clinical decision support into practice workflows.
Identifiants
pubmed: 37267474
doi: 10.1097/ICO.0000000000003318
pii: 00003226-990000000-00307
pmc: PMC10689570
mid: NIHMS1895890
doi:
Types de publication
Journal Article
Langues
eng
Sous-ensembles de citation
IM
Subventions
Organisme : NEI NIH HHS
ID : P30 EY005722
Pays : United States
Organisme : NEI NIH HHS
ID : R01 EY031033
Pays : United States
Informations de copyright
Copyright © 2023 Wolters Kluwer Health, Inc. All rights reserved.
Références
JAMA. 2016 Dec 13;316(22):2402-2410
pubmed: 27898976
NPJ Digit Med. 2018 Aug 28;1:39
pubmed: 31304320
Nat Biomed Eng. 2020 Aug;4(8):767-777
pubmed: 32572198
Bull World Health Organ. 2001;79(3):214-21
pubmed: 11285665
Sci Rep. 2021 Jan 12;11(1):586
pubmed: 33436781
Sci Rep. 2020 Oct 20;10(1):17851
pubmed: 33082530
Comput Methods Programs Biomed. 2021 May;203:106048
pubmed: 33765481
MMWR Morb Mortal Wkly Rep. 2014 Nov 14;63(45):1027-30
pubmed: 25393221
Med Image Anal. 2022 May;78:102427
pubmed: 35344824
Lancet Glob Health. 2021 Feb;9(2):e144-e160
pubmed: 33275949
Nat Commun. 2021 Jun 18;12(1):3738
pubmed: 34145294
Ophthalmology. 2017 Nov;124(11):1678-1689
pubmed: 28942073
Community Eye Health. 2009 Dec;22(71):33-5
pubmed: 20212922
Semin Ophthalmol. 2020 May 18;35(4):210-215
pubmed: 32644878
Stat Med. 1998 Apr 30;17(8):857-72
pubmed: 9595616
Am J Ophthalmol. 2020 Oct;218:128-135
pubmed: 32445703
Ophthalmology. 2017 Jul;124(7):962-969
pubmed: 28359545
Eye Vis (Lond). 2020 Apr 16;7:22
pubmed: 32322599
Stat Med. 2000 May 15;19(9):1141-64
pubmed: 10797513
Comput Biol Med. 2021 Oct;137:104675
pubmed: 34425417
Asia Pac J Ophthalmol (Phila). 2020 Mar-Apr;9(2):88-95
pubmed: 32349116
Indian J Ophthalmol. 2022 Apr;70(4):1131-1138
pubmed: 35325999
Lancet Glob Health. 2017 Dec;5(12):e1221-e1234
pubmed: 29032195