Lung X-ray image classification method based on K-means clustering and GAN

• Main Authors: LIU KUN, NING XIAOLIN
• Format: Patent
• Language: Chinese; English
• Published: 06.08.2021
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; HANDLING RECORD CARRIERS; HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATIONTECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING ORPROCESSING OF MEDICAL OR HEALTHCARE DATA; INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTEDFOR SPECIFIC APPLICATION FIELDS; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

The invention provides a lung X-ray image classification method based on K-means clustering and GAN. The method comprises the following steps of firstly, extracting image feature information by using a ResNet-20 network, learning features of each type of lung X-ray images with labels, averaging the extracted features of each type of images, and taking the averaged value as an image clustering center, extracting the features of the unlabeled image and the clustering center, calculating the Euclidean distance with the weight, and carrying out label prediction to obtain unlabeled image rough classification, and finally, sending a rough classification result, a small number of label images, an image generated by the generator and a large number of label-free images into a discriminator network for fine classification. Experimental results show that compared with the most advanced lung X-ray image classification method (the accuracy is 94%), the method uses less than 30% of labeled samples, and efficient performan