A computational method for the semi-automated quantitative analysis of tympanic membrane perforations and tympanosclerosis.

Tympanic membrane pathological findings such as perforations and tympanosclerotic plaques (along with conductive hearing loss and purulent drainage) are the most typical trademarks of either active or burned-out chronic otitis media. These findings are normally evaluated by an expert using visual subjective analysis. In this work, we present a computational semi-automated method to quantify these pathologies. We also present a validation study of the method using a subset of 39 cases randomly extracted from a set of more than 2000 cases (perforations and tympanosclerosis) of the Hospital de Clínicas de Porto Alegre. The validation was performed using Ground Truth images, generated by experts, and has shown promising results.