Attention-guided model for abdominal computed tomography liver tumor segmentation
Owing to the low frequency of liver tumors in abdominal computed tomography(CT)images and the lim-itations of artificial and traditional segmentation methods,this study presents a high-efficiency two-stage attention-guided liver tumor segmentation model.This model consists of a liver organ segmentation module and a liver tumor segmentation module.In the first module,a convolutional neural network generates a segmented image of the liver organ,which is superimposed with the original CT image to obtain a new input image.This image is subsequently imported into the liver tumor segmentation module to obtain accurate liver tumor segmentation images.Each module incorporates a specific loss function for supervised training.This study uses the data set from a liver tumor segmen-tation challenge,conducting experimental research and qualitative comparative analysis.Our model outperforms ex-isting models across several metrics,including the Dice coefficient.The improvements in liver tumor segmentation accuracy and the precision of tumor localization result in segmented images that more closely align with the real boundaries of liver tumors.
abdominal organ segmentationmedical image processingabdominal CTliver segmentationliver tumor segmentationdeep learningattention guideloss function
国家科技期刊平台
NSTL
万方数据
