Classification of diabetic retinopathy based on improved deep forest model

Medical image is of great significance for medical diagnosis and clinical research. In previous studies, many supervised learning methods have been applied to the classification of medical images. In this paper, we propose an improved deep forest model, called MFgcForest (Multi-class feature Extraction deep Forest), for multi-classification of diabetic retinas. The main process is to input the raw data into the MFgcForest algorithm, firstly, the subsamples generated from the multi-grain scans are input to two random forests for classification to determine whether the patient is sick or not, secondly, some features are eliminated according to the performance of the classification to avoid their negative impact on the classification results, and finally the filtered features are input to the cascade forest to obtain the final prediction results and improve the classification The performance of the classification is improved. In this study, the Kaggle diabetic retinal image dataset is selected, and KNN, SVM and RF are used as comparison models to verify the effectiveness of the algorithm. The results show that the MFgcForest model proposed in this paper has better performance compared with other models, and can effectively improve the accuracy of predictive classification of 2–4% of diabetic data, which has important theoretical significance and It has important theoretical significance and practical value for diabetes diagnosis.