Research On Machine Learning Algorithm Based On Ensemble Learning

Computer-aided diagnosis technology has become one of the important research topics in the field of medical imaging research.We study the application of machine learning in computer-aided diagnosis of glaucoma and the application of deep learning in medical image segmentation.The research is based on the diagnosis of glaucoma.The medical images mainly include color fundus photography and optical tomography report.Based on the prior knowledge of glaucoma diagnosis,quantitative features that can identify glaucoma are extracted from medical image reports as priori features.The priori features extracted include cup-to-disk ratio,morphology and scale characteristics of optic nerve fiber layer thickness.Furthermore,we design an integrated learning algorithm based on priori features to achieve automatic diagnosis of glaucoma.We design two ensemble learning algorithms from two views;The first algorithm is used to integrate the prediction of the classifier based on cup-to-disk ratio and optic nerve fiber layer thickness.The second algorithm is based on a metric learming method to learn from two views,and make the two types of sample points as separated as possible in the Euclidean space.Finally,the effectiveness of the proposed algorithms are verified on a clinical dataset.In the first integrated algorithm,support vector machine and logistic regression are used as classifier,and in the second algorithm,support vector machine and K-nearest neighbor algorithm are used as classifier.Both of the above integrated algorithms have achieved good predictive performance on the clinical data.We also use the convolutional neural network in deep learning to automatically identify the optic disc line in the color fundus.The results of the automatic segmentation are compared to the results manually labeled by physician.The results show that the results of automatic segmentation of the algorithm are highly consistent with the results marked by the physician.