Investigative Ophthalmology & Visual Science Cover Image for Volume 65, Issue 7
June 2024
Volume 65, Issue 7
Open Access
ARVO Annual Meeting Abstract  |   June 2024
Sjögren's Syndrome Diagnosis Using Dry Eye Clinical Data using Deep Learning
Author Affiliations & Notes
  • Mathieu Lamard
    Universite de Bretagne Occidentale, Brest, France
    INSERM, LaTIM, UMR 1101, Brest, France
  • Anas-Alexis Benyoussef
    Service d'Ophtalmologie, CHRU de Brest, Brest, France
    INSERM, LaTIM, UMR 1101, Brest, France
  • Rachid Zeghlache
    Universite de Bretagne Occidentale, Brest, France
    INSERM, LaTIM, UMR 1101, Brest, France
  • Divi Cornec
    Service de Rhumatologie, CHRU de Brest, France
    INSERM, LBAI UMR 1227, Brest, France
  • Ikram Brahim
    INSERM, LBAI UMR 1227, Brest, France
    INSERM, LaTIM, UMR 1101, Brest, France
  • Footnotes
    Commercial Relationships   Mathieu Lamard None; Anas-Alexis Benyoussef None; Rachid Zeghlache None; Divi Cornec None; Ikram Brahim None
  • Footnotes
    Support  None
Investigative Ophthalmology & Visual Science June 2024, Vol.65, 5727. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Mathieu Lamard, Anas-Alexis Benyoussef, Rachid Zeghlache, Divi Cornec, Ikram Brahim; Sjögren's Syndrome Diagnosis Using Dry Eye Clinical Data using Deep Learning. Invest. Ophthalmol. Vis. Sci. 2024;65(7):5727.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

Purpose : (1) To develop a deep learning algorithm using dry eye clinical data,such as symptoms of xerophthalmia and Schirmer's test findings, for efficient prediction of Sjögren's Syndrome (SS).
(2) To combine H&E stained labial gland biopsy images with dry eye clinical data to enhance SS classification.

Methods : We used the DIApSS (Diagnostic Suspicion of Primitive Syndrome Sjögren's - Brest Cohort, NCT03681964), an observational study. The clinical data included gender, xerophthalmia symptoms, Schirmer's test results, and Anti-Ro/SSA antibody levels (U/mL). The dataset comprised 76 confirmed SS patients and 90 non-SS subjects. According to Schirmer’s test, the non-SS patients were categorized as follows: severe dry eyes (0-5mm) 39%, moderately dry (5-10mm) 14%, mildly dry (10-15mm) 10%, and normal (>15mm) 37%. The total of 166 patients was divided into training (99), validation (33), and testing (34) groups. Our method uses a Multi-Layer Perceptron (MLP) using PyTorch that processes clinical data and dry eye examination results, and a CNN encoder, inception_V3 backbone, that analyzes the H&E stained labial gland biopsy images. The outputs from both models are summed and fed into a classification head to predict the SS diagnosis.

Results : Evaluating the diagnostic precision using only the clinical data, we obtained an Area Under the Curve (AUC) of 0.95 , Accuracy of 0.82, Cohen's kappa score 0.51, and a recall of 0.83. When combining dry eye clinical data alongside H&E stained labial gland biopsy images we obtained an AUC of 0.98, Accuracy of 0.86, kappa score 0.69, and a recall of 0.86. Cohen's kappa score indicates that both models show a moderate degree of agreement between the predictions and actual diagnoses.

Conclusions : Our deep learning approach shows promising results in predicting Sjögren's Syndrome diagnosis. The integration of dry eye clinical data and H&E stained labial gland biopsy images significantly enhances SS classification accuracy. Future work will focus on including additional Lissamine staining examination results, such as the Oxford score, and Ocular Surface Staining (OSS). This will allow us to investigate the importance of various dry eye examinations when diagnosing SjD.

This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.

 

Flowchart illustrating the Deep learning model using dry eye clinical data and labial glan biopsies for Sjögren's Syndrome diagnosis.

Flowchart illustrating the Deep learning model using dry eye clinical data and labial glan biopsies for Sjögren's Syndrome diagnosis.

×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×