Open AccessBone metastasis segmentation based on Improved U-NET algorithm
Author(s) -
Jingyi Zhang,
Mengge Huang,
Tao Deng,
Ye Cao,
Qiang Lin
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1848/1/012027
Subject(s) - segmentation , computer science , artificial intelligence , feature (linguistics) , image segmentation , medical diagnosis , bone metastasis , deep learning , pattern recognition (psychology) , image (mathematics) , computer vision , algorithm , metastasis , radiology , medicine , cancer , philosophy , linguistics
Whole body bone scan image analysis is widely used in nuclear medicine to assist nuclear medicine physicians in the detection of bone metastases. At present, the analysis of whole-body bone scan images mainly relies on the manual reading of nuclear medicine doctors. The doctors, based on personal knowledge and experience, look for abnormal lesion locations and diagnose them by examining the whole-body bone scan images. However, this method is prone to misdiagnosis and missed diagnosis. To solve the above problems, this study proposes an image segmentation method based on deep learning, which can automatically identify the location of bone metastases, so that doctors can make more accurate diagnosis. The Methods Attention mechanism was added to the jump connection of the original U-NET network to enhance the image feature selection. Experiments show that the algorithm in this study teaches traditional U-Net to show better results on the three indicators of MIoU Dice and MAP.