黄斑水肿(ME)被认为是增殖性糖尿病视网膜病变的主要适应症之一, 由于视网膜下层中蛋白质沉积物的积累, ME引起视网膜肿胀。光学相干断层成像(OCT)通过显示黄斑病变的横截面视图提供了ME的早期检测。文章提出了一种新的算法, 即改进遗传算法和最佳熵法相结合的方法, 可以在前期工作的基础上提高时间效率, 从5s提高至0.7s。算法首先对图像预处理, 排除血管和散斑噪声对分割效果的影响, 再将信息熵作为遗传算法(GA)的适应度函数, 将最大熵作为遗传算法的收敛准则。经过遗传操作, 得到最佳阈值, 对黄斑水肿OCT图像进行分割, 然后对得到的二值图像进行边界平滑处理, 得到最终的分割图像。所述方法对300张黄斑水肿OCT图像进行了分割, 与Canny算子、OTSU分割算法和临床医生手动分割结果相比较。结果表明, 算法能够快速准确提取黄斑水肿区域, 为临床诊断和治疗提供了定量分析的工具。

Macular Edema (ME) is considered to be one of the main indications for proliferative diabetic retinopathy, causing retinal swelling due to accumulation of protein deposits in the subretinal layer. Optical Coherence Tomography (OCT) provides early detection of ME by showing a cross-sectional view of macular lesions. On the basis of the preliminary work, a new algorithm is proposed, which is a combination of improved genetic algorithm and optimal entropy method. It can improve the time efficiency from 5s to 0.7s on the basis of previous work. Firstly, the image is preprocessed to eliminate the influence of blood vessel and speckle noise on the segmentation effect. Then the information entropy is used as the fitness function of Genetic Algorithm (GA), and the maximum entropy is used as the convergence criterion of the genetic algorithm. After the genetic operation, the optimal threshold is obtained. Therefore, the OCT image of the macular edema is segmented, and the obtained binary image is subjected to boundary smoothing to obtain the final segmented image. This method is used to segment 300 OCT images of macular edema, compared with Canny operator, OTSU segmentation algorithm and manual segmentation results of clinicians, respectively. The results show that the proposed algorithm can quickly and precisely extract the area of macular edema, and provides a quantitative analysis tool for clinical diagnosis and treatment.