Residual network-based liver cancer multi-parameter nuclear magnetic resonance image classification method
The invention relates to a multi-parameter liver cancer nuclear magnetic image classification method based on a residual network. For nuclear magnetic data of an arterial phase, a delay phase and T2WI, in preprocessing, lesion parts of nuclear magnetic images of three parameters are segmented and fu...
Saved in:
Main Authors | , , , |
---|---|
Format | Patent |
Language | Chinese English |
Published |
22.08.2023
|
Subjects | |
Online Access | Get full text |
Cover
Loading…
Summary: | The invention relates to a multi-parameter liver cancer nuclear magnetic image classification method based on a residual network. For nuclear magnetic data of an arterial phase, a delay phase and T2WI, in preprocessing, lesion parts of nuclear magnetic images of three parameters are segmented and fused, and the image quality and the number can reach the training degree through image enhancement; and constructing a new camouflage classification residual network CCRNet for image classification. In the network, a multi-parameter feature extraction module MFE is provided, and the module is used for improving extraction of fused multi-parameter nuclear magnetic data complementary information; an improved asymmetric receptive field ARF is provided to enhance the distinguishability of the features, and the lesion feature information is further accurately identified; a multi-level feature fusion module MLFF is provided to combine feature information from each level to form a more discriminative fusion feature. Accord |
---|---|
Bibliography: | Application Number: CN202310330284 |