Nl-Fr

View abstract

This abstract is assigned to session Poster Session - On Display Posters nr 300 ... 399
Presentation preference Poster presentation
TitleMachine learning-based diagnosis of uveitis diseases using routine peripheral blood test data
PurposeTo evaluate diagnostic accuracy of machine learning for predicting diagnosis of uveitis using peripheral blood test data.
MethodsFive thousand four hundred and twenty-three patients diagnosed with uveitis at the Department of Ophthalmology, Tokyo Medical University Hospital between April 2004 and March 2020 were analyzed. Ten types of uveitis comprising eight diagnosed at high frequencies [Vogt-Koyanagi-Harada (VKH) disease, sarcoidosis(SAR), Behcet disease (BD), herpetic iridocyclitis(HI), vitreoretinal lymphoma (VRL), idiopathic pediatric iridocyclitis (IPD), acute retinal necrosis (ARN), and endophthalmitis(END)], other classifiable uveitis (others), and unclassified uveitis were diagnosed by six machine learning algorithms [support vector machine linear (SVM-L), support vector machine: radial basis function (SVM-RBF), random forest (RF), decision tree (DT), naïve Bayes (NB), and linear discriminant analysis (LDA) using peripheral blood test data.
ResultsAccuracy (median ± standard deviation) using RF, SVM-RBF, SVM-L, LDA, DT and NB was 49±14%, 47±0.7%, 45±0.7%, 45±0.8%, 42±0.9% and 14±1%, respectively. Using RF, precision was 83.3% for BD, 77.5% for ARN, 77.1% for SAR, 75.0% for IPD, 61.5% for VRL, 54.0% for END, 50.0% for HI, 50.0% for VKH, 49.4% for others, and 45.7% for unclassified.
ConclusionMachine learning approach using routine peripheral blood test data may be useful to make a diagnosis of uveitis.
Conflict of interestNo
Authors 1
Last nameTSUBOTA
Initials of first name(s)K
DepartmentDepartment of Ophthalmology, Tokyo Medical University Hospital
CityTokyo
CountryJapan
Authors 2
Last nameUsui
Initials of first name(s)Y
DepartmentDepartment of Ophthalmology, Tokyo Medical University Hospital
CityTokyo
CountryJapan
Authors 3
Last nameNezu
Initials of first name(s)N
DepartmentDepartment of Ophthalmology, Tokyo Medical University Hospital
CityTokyo
CountryJapan
Authors 4
Last nameGoto
Initials of first name(s)H
DepartmentDepartment of Ophthalmology, Tokyo Medical University Hospital
CityTokyo
CountryJapan