Enhancing Ophthalmic Triage: identification of new clinical features to support healthcare professionals in triage

Purpose of review Vision Language Models are an emerging paradigm in artificial intelligence that offers the potential to natively analyze both image and textual data simultaneously, within a single model. The fusion of these two modalities is of particular relevance to ophthalmology, which has historically involved specialized imaging techniques such as angiography, optical coherence tomography, and fundus photography, while also interfacing with electronic health records that include free text descriptions. This review then surveys the fast-evolving field of Vision Language Models as they apply to current ophthalmologic research and practice. Recent findings Although models incorporating both image and text data have a long provenance in ophthalmology, effective multimodal Vision Language Models are a recent development exploiting advances in technologies such as transformer and autoencoder models. Summary Vision Language Models offer the potential to assist and streamline the existing clinical workflow in ophthalmology, whether previsit, during, or post-visit. There are, however, also important challenges to be overcome, particularly regarding patient privacy and explainability of model recommendations.

show abstract

An education-based intervention investigating the accuracy of community-based optometrists evaluating limbal anterior chamber depth

Jindal,

Agnew,

Siriwardena

et al. 2024

Preprint

View full text Add to dashboard Cite

Introduction In the UK, referrals for suspected primary angle closure (PAC) by community-based optometrists (CBO) to the hospital eye service show moderate accuracy. This study aimed to assess the inter-observer agreement of limbal anterior chamber depth (LACD) between CBOs and ophthalmologists and evaluate the impact of an education intervention. Methods Using a 7-point LACD grading scale, consultant ophthalmologists graded an LACD image dataset developed from 100 patients attending Moorfields Eye Hospital. Two sets of 84 images were utilised for two LACD online assessments. CBO were recruited and undertook assessments before and after a PAC education package (EP) between August 2023 and January 2024. Results Fifty-two optometrists completed the initial LACD assessment with a median of 19.0 (IQR 9.3–24.8) years post-registration experience. Using the 7-point LACD grading scale, mean weighted kappa (Kw) for pre-EP was moderate, 0.42 (95%CI, 0.36–0.48), increasing to 0.47 (0.42–0.53) post-EP. Conversion to the 4-point grading scale, Kw was substantial, pre-EP 0.61, (0.56–0.66) and post-EP Kw 0.64 (0.58–0.69). An LACD threshold (< 25%) sensitivity and specificity pre-EP was 86.0% (79.9%-91.3%) and 84.6 (82.3–86.9), respectively. Post-EP sensitivity decreased to 78.2% (74.1%-85.1%) (p = 0.049) but specificity increased to 90.4 (88.1–92.8) (p < 0.001). Conclusion The 4-point LACD grading scale would be more practical for a case-finding setting. An accessible EP improved agreement and specificity using the joint Royal College of Ophthalmologists/College of Optometrists PAC referral threshold. This EP may reduce false positive PAC referrals and could be rapidly disseminated to CBO. Further research is needed to assess real-world PAC referrals after an education intervention.

show abstract

Enhancing Ophthalmic Triage: identification of new clinical features to support healthcare professionals in triage

Cited by 4 publications

References 43 publications

An education-based intervention investigating the accuracy of community-based optometrists evaluating limbal anterior chamber depth

An education-based intervention investigating the accuracy of community-based optometrists evaluating limbal anterior chamber depth

Vision language models in ophthalmology

An education-based intervention investigating the accuracy of community-based optometrists evaluating limbal anterior chamber depth

Contact Info

Product

Resources

About