ResNet model-based liver cancer nuclear magnetic image classification method
The invention relates to a liver cancer nuclear magnetic image classification method based on a ResNet model, and belongs to the field of deep learning and medical image classification. According to T2WI nuclear magnetic data, dimension reduction is carried out on 3D data through preprocessing, a se...
Saved in:
Main Authors | , , , |
---|---|
Format | Patent |
Language | Chinese English |
Published |
12.08.2022
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | The invention relates to a liver cancer nuclear magnetic image classification method based on a ResNet model, and belongs to the field of deep learning and medical image classification. According to T2WI nuclear magnetic data, dimension reduction is carried out on 3D data through preprocessing, a sequence 2D image with clear lesion is screened out, a lesion part is segmented out, and image quality and number can reach a trainable degree and maintain relative balance through image enhancement; through a ResNet101 model, migration learning is introduced to solve the general problem of small medical image data volume; an attention mechanism CBAM is introduced into a network structure, spatial attention and channel attention are utilized to enable the network to learn lesion features with more weights, and lesion feature relations are obtained globally; and finally, the residual structure is improved and adjusted, so that the residual structure is more suitable for feature extraction and training of the medical i |
---|---|
Bibliography: | Application Number: CN202210598442 |