Abstract
Purpose :
Anterior segment optical coherence tomography (AS-OCT) has been shown to be useful in quantifying intraocular inflammation. Limitations in AS-OCT image quality may compromise utility. We aimed to develop and validate criteria to assess the quality of AS-OCT images in the context of use for the assessment of inflammation.
Methods :
A cross-sectional study using AS-OCT images acquired within a specialist eye hospital from participants (aged 6 – 16) diagnosed with childhood onset uveitis. A novel three-level grading system (of good, limited or poor utility for quantifying anterior chamber cells) was defined, as were image artefact / quality limiting features (lid, eyelash, cropping, flare, movement). Three independent experts graded 2825 images, using the quality criteria described. Intraclass correlation coefficients (ICC) were calculated for each parameter to quantify agreement.
Results :
There was good inter-grader agreement scores when assessing image quality with ICC 0.85 (95%CI 0.84-0.87) between graders. There was complete agreement at either end of the confusion matrix with no ‘good’ images labelled as ‘poor’ by other readers, and vice versa no ‘poor’ images labelled as ‘good’. Similarly, there was good agreement when assessing presence of lash (0.96 95%CI 0.938-0.98), movement (0.972, 95%CI 0.95-0.99), flare (0.82 95%CI 0.80-0.84) and cropping (0.90, 0.88-0.92). ‘Gold standard’ labels (gradings undertaken by the most senior observer) were good for 681 images (24%) and limited for 1845 (65%). The most commonly occurring artefact or quality limiting feature was lash artefact (36% images).
Conclusions :
Wider clinical and community care implementation of AS-OCT assessment of ocular inflammation will be reliant on standardised, validated image quality assessment criteria (IQAC). The novel IQAC described here are robust with good interobserver agreement, and excellent agreement on the differentiation between ‘good’ and ‘poor’ quality images. The large proportion of images graded as ‘limited’ suggest the need for further work to subdivide this classification, using the specific image quality limiting features, for which we also report good interobserver agreement. Work is currently underway to further validate these IQAC through citizen science assessment of images by the public on the Zooniverse platform (the world’s largest platform for crowd-sourced science).
This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.