Computerized Image Analysis of Nails Affected by Fungal Infection: Evaluation Using Digital Photographs and Manually Defined Areas

September 2004 | Volume 3 | Issue 5 | Original Article | 489 | Copyright © September 2004


Robert Baran MD, Adele Sparavigna MD, Michele Setaro, Federico Mailland MD

Abstract
Despite the relevant increase in clinical trials on the efficacy of various systemic and/or topical antifungal agents in onychomycosis therapy, the evaluation of the results is largely subjective. The aim of this study was to set up and ensure an objective, reproducible and reliable method to measure nail plate involvement in onychomycosis. In order to validate a specifically designed software for the computerized image analysis of affected areas of the nail, standardized clinical pictures of onychomycosis were prepared by six different clinicians using a sample of 11 affected nails. Diseased areas and total nail plates were measured both on the clinical pictures and on their drawings traced by the different clinicians on transparent tapes adhering to sample nails. The computerized procedure was undertaken by a trained operator who was not a dermatologist. The variation coefficients of measurements on clinical pictures (automatically detected) and on drawings were compared. In addition, the agreement between automatic evaluation and drawing was evaluated by means of Bland-Altman analysis. To consider the effect of possible variations linked to different operators using the computerized method, 11 clinical pictures (one for each clinical case considered) were selected and submitted to computerized image analysis by six different trained operators. The computerized detection of affected nail areas showed a coefficient of variation (vc=8.5%) lower than that observed on drawings (vc=14.7%). The two methods showed appreciable agreement, as demonstrated by Bland-Altman plot. The coefficient of variation of image analysis conducted by six different operators was very low for the total area calculation (vc=0.9%) and acceptable for pathological area detection (vc=4.8%). Based on the results obtained, we conclude that automatic evaluation is a reliable and helpful method for the measurement of the clinical involvement of the nail plate in onychomycosis and for the evaluation of therapies, since it can increase the objectivity and reproducibility of data. However, in a minority of difficult cases, expert dermatological evaluation is needed.