Laser & Optoelectronics Progress, Volume. 60, Issue 4, 0410015(2023)
ω-net: A Secondary Feature Extraction Method for Multiple Medical Images
Figures & Tables(16)
Table 1. Dataset summary
View tableTable 1. Dataset summary
Dataset Number of images Size Segmentation part Image（Label） format Kagglelung 267 512×512 Lung TIF（TIF） Liver 420 512×512 Liver PNG（PNG） dsb2018 546 256×256 Cell PNG（PNG） Brats18-19 22151 512×512 Brain tumor NPY（NPY）
Table 2. Comparison of experimental results of brain tumor segmentation
View tableTable 2. Comparison of experimental results of brain tumor segmentation
Region Parameter Unet ResUnet DenseUnet Unet++ ResUnet++ Fcn16s［ 17 ］ω-net WT D 0.6754 0.7368 0.6713 0.6825 0.7609 0.7326 0.7896 TC 0.6330 0.6864 0.5508 0.7964 0.7513 0.7893 0.8490 ET 0.5145 0.6058 0.4951 0.6033 0.6369 0.6536 0.6946 WT P 0.8104 0.8641 0.7784 0.9072 0.8723 0.9070 0.8314 TC 0.7779 0.6715 0.6627 0.9354 0.9155 0.8607 0.8995 ET 0.6465 0.8080 0.6332 0.7952 0.7238 0.8240 0.7236 WT S 0.7469 0.7875 0.7667 0.7084 0.7962 0.7576 0.8326 TC 0.7552 0.9204 0.7620 0.8036 0.7626 0.8687 0.8839 ET 0.6563 0.6105 0.5679 0.6745 0.7406 0.6891 0.7591 WT DH 3.44 3.08 3.47 3.06 3.03 2.93 2.98 TC 2.06 2.47 3.05 1.77 2.07 1.93 1.67 ET 3.82 3.32 3.76 3.25 3.39 3.15 3.14 Number of parameters /106 118.42 124.42 155.12 139.74 142.41 512.25 133.81
Table 3. Comparison of cell segmentation experiment results
View tableTable 3. Comparison of cell segmentation experiment results
Network RMIoU DH S A P D Number of parameters /106 Unet 0.6931 5.23 0.7408 0.9373 0.9080 0.7805 118.42 CENet［ 18 ］0.7116 5.20 0.7578 0.9453 0.8998 0.8041 139.73 Unet++ 0.7488 5.09 0.8309 0.9553 0.8720 0.8355 139.74 AttUnet 0.7196 5.08 0.7726 0.9475 0.9020 0.8109 133.05 R2Unet 0.7221 6.15 0.6823 0.9248 0.8642 0.6105 149.12 R2AttUnet 0.7410 5.04 0.8177 0.9431 0.8683 0.8212 150.47 DenseUnet 0.6947 5.31 0.7344 0.9309 0.9129 0.7774 155.12 ω-net 0.7544 4.96 0.8315 0.9574 0.8789 0.8401 133.81
Table 4. Comparison of lung segmentation experiment results
View tableTable 4. Comparison of lung segmentation experiment results
Network RMIoU DH S P D Number of parameters /106 Unet 0.9507 6.07 0.9718 0.9685 0.9716 118.42 CENet 0.9538 6.03 0.9851 0.9680 0.9736 139.73 Unet++ 0.9560 5.93 0.9850 0.9704 0.9748 139.74 AttUnet 0.9549 5.97 0.9809 0.9733 0.9742 133.05 R2Unet 0.9467 6.15 0.9821 0.9637 0.9698 149.12 ω-net 0.9791 5.92 0.9819 0.9768 0.9759 133.81
Table 5. Comparison of liver segmentation experiment results
View tableTable 5. Comparison of liver segmentation experiment results
Network RMIoU DH S A P D Number of parameters /106 Unet 0.8185 6.62 0.8988 0.9865 0.8931 0.9001 118.42 CENet 0.8567 7.14 0.9498 0.9884 0.8990 0.9225 139.73 Unet++ 0.8437 6.55 0.9771 0.9855 0.8530 0.9142 139.74 AttUnet 0.8409 7.87 0.9417 0.9883 0.8974 0.9132 133.05 R2Unet 0.7132 8.64 0.9565 0.9541 0.8872 0.8225 149.12 R2AttUnet 0.7558 7.46 0.7072 0.9296 0.9036 0.8454 150.47 ω-net 0.8972 6.21 0.9435 0.9901 0.9493 0.9458 133.81
Table 6. Ablation experiment
View tableTable 6. Ablation experiment
Model RMIoU DH S P D Unet 0.8185 6.62 0.8988 0.8931 0.9001 ω-net（first feature map） 0.8043 7.40 0.9001 0.9237 0.8909 ω-net（second feature map） 0.8069 7.94 0.9392 0.8515 0.8926 ω-net（third feature map） 0.8540 6.92 0.9404 0.9250 0.9249 ω-net（fourth feature map） 0.8394 7.32 0.9423 0.8768 0.9126 ω-net（third feature map）+PSA 0.8780 6.92 0.9429 0.9403 0.9235 ω-net（third feature map）+PSA+CA 0.8972 6.21 0.9435 0.9493 0.9458
Tools
Get Citation
Copy Citation Text
Hao Wu, Yang Xu, Bin Cao. ω-net: A Secondary Feature Extraction Method for Multiple Medical Images[J]. Laser & Optoelectronics Progress, 2023, 60(4): 0410015
Paper Information
Category: Image Processing
Received: Nov. 29, 2021
Accepted: Jan. 5, 2022
Published Online: Feb. 14, 2023
The Author Email: Xu Yang (xuy@gzu.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20