Assessing the accuracy of virtual clinic diagnosis of benign lid lesions based on patient self-generated photographs
Author: Mumta Kanda
Base Hospital / Institution: Moorfield’s Eye Hospital, London
ePoster presentation
Abstract ID: 24-209
Purpose
To analyse the diagnostic accuracy and quality of patient self-generated photos in a benign lid lesion virtual clinic at a tertiary referral centre in London, UK.
Methods
Adult patient referrals were triaged to a benign lid lesion virtual clinic. Patients submitted self-captured photos prior to clinic. Photographs from 300 consecutive patients submitted April – May 2021 were reviewed by two independent observers and graded on quality using a rating scale. Virtual diagnosis was compared to clinic, biopsy, surgical, or imaging diagnosis.
Results
300 patients (57% female; median age 40 (16-90)) submitted 895 photographs. 84.5% of photographs were graded excellent (40.5%) or adequate (44.0%). There was no statistical difference (chi-sq p>0.05) between age groups in quality of images. Virtual clinic diagnosis was verified in 236 (79%) cases: clinic review in 55 (23%), histopathology in 80 (34%), in-surgery review without biopsy in 98 (42%), and imaging in 3 (1%). 232 (98%) lesions were benign, 2 (0.8%) were pre-malignant (bowen’s disease + actinic keratosis), and 2 (0.8%) were malignant (basal cell carcinoma, ocular lymphoma). There was correct virtual clinic identification of lesions as benign in 99% of benign lesions, as pre-malignant in 50% of pre-malignant lesions, and as cancerous in 100% of malignant lesions. An actinic keratosis was labelled a ‘benign lesion’ in the virtual clinic. Photographs from this patient were graded as adequate quality. Correct diagnosis was made in 190 (81%) patients. There was no statistical difference (chi-sq p>0.05) between the quality of images in those with a correct vs. incorrect diagnosis.
Conclusion
Patients, irrespective of age, can provide adequate quality photographs to aid virtual consultations for lid lesions, where poor video resolution might otherwise make assessment challenging. Photographs reliably identify benign lesions but may be less reliable for borderline or pre-malignant lesions, even with adequate photo quality. Therefore, we recommend use of self-generated photos in benign lesion virtual clinics and in-person review where suspicious features are present.
Additional Authors
First name | Last name | Base Hospital / Institution |
---|---|---|
Gabriella | Guervara | Moorfield’s Eye Hospital, London |
Hardeep | Kandola | Moorfield’s Eye Hospital, London |
Laxmi | Raja | Moorfield’s Eye Hospital, London |
Swan | Kang | Moorfield’s Eye Hospital, London |