Deep Learning Based Fluid Segmentation in Retinal Optical Coherence Tomography Images

Macular Edema (ME) is the accumulation of fluid in the macular region of the eye, and it may lead to the distortion of center vision. It often occurs in diabetic retinopathy. It is important to measure fluid accumulation in ME patients for disease monitor. Segmentation of the fluid region in the retinal layer is an essential step for quantitatively analysis. However, manual segmentation is time consuming and also subjective. In this paper, a new deep learning based segmentation method is proposed. The attention mechanism is introduced to automatically locate the fluid region, which can reduce the number of parameters compared to typical two-stage approaches. In addition, dense skip connection makes the segmentation result more accurate. Joint losses are used, including cross entropy loss, dice loss and regression loss. The proposed method is evaluated on a public available dataset, results show that the proposed method can adapt to the OCT scans acquired by various imaging scanning devices, and this method is more effective than other methods.