Chronic rhinitis (CR) is among the most frequent inflammatory diseases of ear-nose-throat (ENT) covering up to 30% of the population. Different forms of CR require different treatment tactics, which indicates the need for an efficient tool for differential diagnostics of CR. Optical coherence tomography (OCT) is a promising tool for fast non-invasive evaluation of nasal mucosa, which, however, requires further interpretation of the obtained diagnostic image. In this paper, we provide a comparative analysis of several machine learning approaches that aim at automated differential diagnostics of CR based on diagnostic OCT images of 78 patients aged between 28 and 74 ages. Gradient boosting decision trees (GBT) approach reveals the best classification accuracy (98% and 94% for binary and diagnostic classification, respectively). It shows that proposed approaches have potential for automated classification of CR OCT images.

慢性鼻炎（CR）是最常见的耳鼻喉（ENT）炎性疾病，覆盖了30％的人口。不同形式的CR需要不同的治疗策略，这表明需要用于差动诊断CR的有效工具。光学相干断层扫描（OCT）是用于鼻粘膜快速非侵入性评估的有前途的工具，但是，这需要对获得的诊断图像进行进一步的解释。在本文中，我们基于对78例年龄在28至74岁之间的患者的诊断性OCT图像，对几种旨在自动进行CR的鉴别诊断的机器学习方法进行了比较分析。梯度提升决策树（GBT）方法显示出最佳的分类准确性（二进制和诊断分类分别为98％和94％）。