Automated Image Quality Assessment for Anterior Segment Optical Coherence tomograph

Optical Coherence Tomography (OCT) is an imaging technique for diagnosing eye disorders. Image quality assessment (IQA) of OCT images is essential for the diagnosis and study of eye diseases, but manual IQA is time consuming and subjective. Recently, automated IQA methods based on deep learning (DL) have achieved good performance. However, few DL-based methods focus on images of the anterior segment of the eye (AS-OCT). AS-OCT is essential for diagnosing diseases such as anterior uveitis, a significant cause of visual morbidity. Moreover, few of these methods identify the factors that affect the quality of the images (called quality factors in this paper). This adversely affects the acceptance of these methods. In this study, we define, for the first time to the best of our knowledge, the quality level and six quality factors of AS-OCT for the clinical context of anterior chamber inflammation. We then develop an automated framework based on multi-task learning to assess image quality and to identify the existing of quality factors in AS-OCT images. The effectiveness of the framework is demonstrated in experiments.