Purchase this article with an account.
Afshan Nanji, Anat Galor, William J Feuer, James Wong, Madhura G Joag, Juan Carlos Murillo, Ibrahim Osama, Nabeel M Shalabi, Carol L Karp; Evaluation of high resolution optical coherence tomography among novice users in the diagnosis of ocular surface squamous neoplasia. Invest. Ophthalmol. Vis. Sci. 2014;55(13):2448.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To assess the utility of anterior segment high resolution optical coherence tomography (HR OCT), among novice users, in the diagnosis of ocular surface squamous neoplasia (OSSN).
The HR OCTs of 28 patients with biopsy-proven diagnoses were selected for grading. Of the 28 diagnoses, 14 were OSSN and 14 were non-OSSN lesions, including 3 conjunctival melanomas, 3 lymphoma or lymphoid lesions, 2 pterygia, 2 complexion-associated melanosis or primary acquired melanosis lesions, 2 nevi, 1 Salzmann nodular degeneration, and 1 anterior basement membrane dystrophy. Thirty-nine participants across a variety of medical training levels took part in the study. All were inexperienced with HR OCT interpretation. A brief lecture on anterior segment HR OCT was given, following which all participants were asked to evaluate the 28 OCTs for the diagnosis of OSSN. No clinical history or slit lamp images were provided.
Among novice users of HR OCT, after a brief training, sensitivity for detecting OSSN was 65% (range 21-95% across OCTs) and specificity for correctly ruling out OSSN was 91% (range 62% to 100% across OCTs). Of the OCTs that were not OSSN, 10 of the 14 had a specificity of more than 90%. Neither level of training nor cornea specialization was significantly associated with a correct diagnosis of OSSN by HR OCT alone (p >0.9 for each). Among all OCTs, the Kappa coefficient of inter-rater agreement was 0.45, which is considered fair agreement. There was also a correlation between the investigators' assessment of particular OCTs as being easy, moderate, and difficult to grade and the percentage of graders who correctly identified those photos as OSSN. The percentage of graders correctly identifying OSSN in pairs of OCT images deemed by experienced users as easy, moderate, or difficult to grade was 28 (72%), 20 (51%), and 13 (33%), respectively. More of the "easy", or classic, images were correctly identified as OSSN by graders than the "difficult" images (p=0.007). "Moderate" OCT images were intermediate between "easy" and" difficult" but not statistically different from either (p>0.1).
HR OCT has a moderate sensitivity for identification of OSSN and high specificity for ruling out OSSN among novice users, following a brief training in the technology. HR OCT may be used as an adjunctive technology in the evaluation of ocular surface lesions.
This PDF is available to Subscribers Only