Research On Medical Image Classification Based On Deep Forest

The study of medical images facilitates the early detection and treatment of diseases and plays a significant role in the early recovery of patients.The classification of medical images by machine learning provides scientific and effective detection results for health care workers and patients,saves health care resources,and enables patients to receive timely treatment.In this paper,based on deep forest theory,we propose an optimization algorithm combined with an improved deep forest algorithm to classify medical images,and conduct experiments on three medical image datasets on Kaggle,and the results show that the method effectively improves the accuracy of medical image classification,and the main research content is as follows:For the medical image classification problem,this paper proposes an improved deep forest algorithm.The multi-grain scanning part of the deep forest increases the diversity of features,but some subsamples affect the overall classification effect,and the subsamples generated by the multi-grain scanning are screened to improve the overall model prediction accuracy,and in the empirical part the method is compared with the K-nearest neighbor algorithm,support vector machine algorithm,random forest algorithm and traditional In the empirical section,the method is compared with K-nearest neighbor algorithm,support vector machine algorithm,random forest algorithm and the original deep forest algorithms.To address the problem of model optimization,this paper proposes a Lévy flight-based spotted hyena optimization algorithm,which introduces the Lévy flight strategy into the process of finding the optimal parameters,increases the global search capability of parameters,avoids the problem that the traditional method tends to fall into local optimality,and provides reasonable improvements to the parameter selection and tuning of the model,and compares the genetic algorithm,particle swarm algorithm,Differential Evolution Algorithm,Bottleneck Sea Sheath Algorithm,Harris Hawk Optimization Algorithm,and the original Speckled Hyena Optimization Algorithm,the traditional six optimization algorithms,to illustrate the performance of the improved Speckled Hyena Optimization Algorithm,while using the improved Speckled Hyena Optimization Algorithm to optimize two parameters in the weighted deep forest,namely,the number of subtrees in the random forest and the sliding window and the size of the sliding window,and the performance is compared with the unoptimized deep forest on the dataset.In summary,this paper takes three medical image datasets as research objects,i.e.,brain tumor image dataset,ECG image dataset and pneumonia image dataset,and optimizes the improved deep forest model by the improved speckled hyena optimization algorithm to realize the diagnostic classification research of various medical images,improve the accuracy and timeliness of the classification model,and can provide effective solutions for doctors and patients,which are useful in several directions of medical It can provide effective solutions for doctors and patients,and has good promotion and application value in several directions of medical treatment.