Detection and analysis of diabetic macular edema deformation in OCT images using levelset segmentation

This study offers a new method for segmenting Diabetic Macular Edema (DME)  deformations in Optical Coherence Tomography (OCT) images using a morphological Chan-Vese level. Diabetic Retinopathy (DR) is a severe complication that, left unattended or disregarded, leads to impaired vision. Conventional segmentation methods fail to accurately identify the DME deformations in OCT due to their subtlety, complexity, varying qualities, and other macular pathologies. While traditional segmentation methods are utilized, the inherent variability in OCT images’ quality and the common occurrence of concurrent retinal pathologies lead to multiple challenges in accurately pinpointing DME. Patient movement significantly impacts image clarity; different machines used make it difficult to calibrate them all similarly; the flow of early DME trends is quite subtle and is hard to decipher within the retina’s complexity. The modifications presented in the proposed method are designed to enhance the original Chan-Vese model to process the OCT images robustly, considering their unique characteristics. In particular, the proposed method is aimed at enhancing robustness to high-intensity speckled noise and large disparities in contrast inherent to natural images originating from the retina. The proposed method achieves better segmentation of fluid-filled regions that appear in DME cases without depending on strong gradients in OCT images. The simulations yield a marked improvement in the accuracy and efficiency of DME segmentation compared to traditional methods.