Kernel principal component spectral hashing method for diabetic fundus image classification

Abstract

The invention discloses a kernel principal component spectrum hashing method for diabetic fundus image classification. In this method, the diabetic fundus image data is firstly preprocessed and segmented, and the processed fundus image data is converted into a vector form; then the nonlinear feature information in the fundus image data is extracted by the kernel principal component analysis algorithm; and then the data is transformed into In the form of a binary code, the fundus image sample data is represented by the eigenvalue and eigenfunction value of the Laplace-Beltrami operator; finally, the sample eigenfunction value is converted into a binary code using the threshold value, and the nearest neighbor algorithm is used to carry out the process in the Hamming space Efficient classification of diabetic fundus images. The invention can fully extract complex nonlinear diabetic fundus image data features, has high classification accuracy, and can effectively reduce the computational complexity of large-scale fundus image classification.

Description

Technical field: [0001] The invention relates to medical image classification and detection, in particular to a kernel principal component spectrum hashing method for diabetic fundus image classification. Background technique: [0002] Diabetes is a disease with a high incidence rate and has become a major threat to human health. Early detection of diabetes can effectively reduce the cost of treatment, on the contrary, the treatment of advanced diabetes is expensive. Diabetes often leads to retinal abnormalities, a microvascular complication of diabetes known as diabetic retinopathy. Fundus images can be used to monitor retinal abnormalities, so fundus image classification has now become an effective method for non-invasive detection of diabetes. The classification accuracy of fundus image diagnosis is evaluated by using sensitivity and specificity. Sensitivity refers to the percentage of abnormal fundus being correctly classified, and specificity refers to the percentage ...

AI Technical Summary

Problems solved by technology

Early detection of diabetes can effectively reduce the cost of treatment, in contrast to the treatment of advanced diabetes is expensive

Images