Abstract
Diabetic Retinopathy and Diabetic Macular Edema (DME) represent one of the main causes of blindness in developed countries. They are characterized by fluid deposits in the retinal layers, causing a progressive vision loss over the time. The clinical literature defines three DME types according to the texture and disposition of the fluid accumulations: Cystoid Macular Edema (CME), Diffuse Retinal Thickening (DRT) and Serous Retinal Detachment (SRD). Detecting each one is essential as, depending on their presence, the expert will decide on the adequate treatment of the pathology. In this work, we propose a robust detection and visualization methodology based on the analysis of independent image regions. We study a complete and heterogeneous library of 375 texture and intensity features in a dataset of 356 labeled images from two of the most used capture devices in the clinical domain: a CIRRUSTM HD-OCT 500 Carl Zeiss Meditec and 179 OCT images from a modular HRA + OCT SPECTRALIS® from Heidelberg Engineering, Inc. We extracted 33,810 samples for each type of DME for the feature analysis and incremental training of four different classifier paradigms. This way, we achieved an 84.04% average accuracy for CME, 78.44% average accuracy for DRT and 95.40% average accuracy for SRD. These models are used to generate an intuitive visualization of the fluid regions. We use an image sampling and voting strategy, resulting in a system capable of detecting and characterizing the three types of DME presenting them in an intuitive and repeatable way.
Highlights
Due to the consumption habits in the developed countries, diabetes-related diseases have been increasing in prevalence [1]
Among the most common consequences of this pathology, we can find the Diabetic Retinopathy [4] and the Diabetic Macular Edema (DME). The latter is characterized by fluid accumulations in between the retinal layers, producing the progressive vision deterioration that may result in severe blindness [5,6]
To the best of our knowledge, only two works tried to generate a segmentation on the three clinical types we present, the work of Samagaio et al [19] which used texture, intensity and clinical-based features to localize Cystoid Macular Edema (CME), Diffuse Retinal Thickening (DRT) and Serous Retinal Detachment (SRD) fluid accumulations in OCT images and the work of de Moura et al [20], which uses these detections to further generate a segmentation of the different fluid accumulations
Summary
Due to the consumption habits in the developed countries, diabetes-related diseases have been increasing in prevalence [1]. Diabetes mellitus is characterized by a defective use or lack of insulin in the body, resulting in an increase in sugar levels in blood. This causes a progressive damage in different parts of the human body, specially in its most sensible structures like the retinal and choroidal vascularity [2,3]. Among the most common consequences of this pathology, we can find the Diabetic Retinopathy [4] and the Diabetic Macular Edema (DME) The latter is characterized by fluid accumulations in between the retinal layers, producing the progressive vision deterioration that may result in severe blindness [5,6].
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
