基于感知相似度的甲状腺细胞图像识别方法OA

Perceptual similarity is an important branch of artificial intelligence image processing and one of the important meas-ures for effectively recognizing thyroid cells.To address the problem of low diagnostic accuracy of thyroid cancer caused by the scarcity of thyroid cell image samples and insufficient model training,a thyroid cell image recognition method based on percep-tual similarity is proposed.Through image morphology processing,noise information in thyroid cell images is eliminated,and the detection sensitivity of the core region is enhanced.By perceiving the similarity of normal sample images,thyroid cancer is excluded,and by perceiving the similarity of disease samples,thyroid cancer types are recognized.The average recognition ac-curacy of 4 types of thyroid cancer,including papillary cancer,follicular cancer,medullary cancer and undifferentiated cancer is verfied in a tertiary hospital in a certain city which is 99.7％.It's proved that the proposed method can effectively improve the accuracy of thyroid cell image recognition.