Retinal vessel segmentation based on an improved deep forest

To improve the diagnosis efficiency of diseases related to retinal blood vessels, this article proposes a novel retinal blood vessels segmentation algorithm by combining the advantages of convolutional neural network (CNN) and cascade forest (CF). Firstly, we use the contrast limited adaptive histogram equalization (CLAHE) algorithm to enhance the color fundus retinal image. Secondly, we randomly select some image patches to train the CNN feature extraction module and the CF classification module. Finally, the image patches from the test image are sent to the trained model to complete the retinal blood vessel segmentation. The algorithm is verified on the DRIVE, STARE, and CHASE_DB1 datasets. The sensitivity reaches 0.8206, 0.8762, and 0.7705, the accuracy reaches 0.9531, 0.9611, and 0.9559, and the area under the ROC curve reaches 0.9770, 0.9899, and 0.9767, respectively. The comprehensive performance of our method is better than that of some state‐of‐the‐art methods.