基于双阶段注意力机制与可逆变换的医学图像融合

沈瑜,袁祯,刘佳英,白珊,强振凯

Medical image fusion based on dual-stage attention mechanism and invertible transformation

Yu SHEN,Zhen YUAN,Jiaying LIU,Shan BAI,Zhenkai QIANG

表 1 各病例融合图像的客观评价结果

Tab.1 Objective evaluation result of fused images for each case

病例 方法 SD SF AG CC SCD PSNR/dB VIF MS-SSIM 病例1 CNP 43.6912 13.4352 2.5146 0.7513 0.3604 14.7931 0.3062 0.6036 NSST 44.6915 15.2364 3.6235 0.7792 0.4062 15.3642 0.4065 0.5931 MATR 45.0009 13.4218 3.7140 0.8585 0.3033 17.8022 0.6381 0.2624 CDDFuse 55.0304 15.6447 4.1472 0.8640 1.0586 19.3783 0.8957 0.8061 EMMA 63.9454 12.5992 3.3849 0.9018 1.7036 18.2285 0.5267 0.8148 SwinFusion 52.0736 16.0239 4.3561 0.8684 0.5986 18.5632 0.6931 0.2813 MMAE 58.2865 13.4399 3.3697 0.8955 1.4230 19.9769 0.6247 0.8164 BSAFusion 66.8800 18.1830 4.7909 0.8632 1.4037 16.4599 0.5423 0.3911 FATFusion 31.5446 22.5042 5.3812 0.7990 1.7508 12.2301 0.1494 0.6179 MBHFuse 67.6712 17.5006 4.4093 0.9067 1.7901 16.8958 0.5402 0.3959 CrossFuse 66.4745 11.2239 2.9035 0.9083 1.6939 14.8062 0.5564 0.2261 本文方法 69.0206 25.5996 5.2558 0.9098 0.6500 17.5385 0.7666 0.8172 病例2 CNP 45.2790 29.4450 4.9373 0.6927 0.8727 10.1246 0.3191 0.5961 NSST 39.2303 16.2598 5.2248 0.7234 0.3992 12.8197 0.2257 0.2060 MATR 45.4259 23.9303 6.2787 0.6873 0.2803 14.0141 0.6283 0.2358 CDDFuse 65.2071 29.4547 7.6431 0.7896 1.6384 15.5131 0.6629 0.7536 EMMA 72.9314 20.9976 6.2079 0.7804 1.7300 14.2355 0.5284 0.7335 SwinFusion 59.2150 30.0445 8.0129 0.7709 1.4593 15.0673 0.6002 0.2524 MMAE 70.8482 25.2982 6.7522 0.7834 1.6298 14.7633 0.6235 0.7463 BSAFusion 66.8708 33.3345 8.4601 0.7429 1.4039 14.2714 0.5343 0.3321 FATFusion 56.5790 30.3244 7.7434 0.7268 1.3804 15.1629 0.7196 0.7555 MBHFuse 73.2663 31.8790 7.9370 0.7888 1.6942 14.0994 0.6166 0.3147 CrossFuse 67.8622 13.9737 4.1944 0.7565 1.6032 13.5881 0.5464 0.1962 本文方法 67.1445 35.2224 8.8516 0.7964 1.6173 15.2154 0.6595 0.7621 病例3 CNP 43.8249 16.3485 3.5793 0.7014 0.8589 11.5829 0.3853 0.5715 NSST 40.5782 13.1943 3.6839 0.6619 0.4928 12.3462 0.4123 0.4812 MATR 50.1103 16.3469 4.6163 0.6475 0.5786 13.5929 0.7438 0.2504 CDDFuse 58.7600 19.5557 5.3171 0.6680 1.0714 15.0229 0.8588 0.7630 EMMA 59.5400 17.2863 5.0076 0.7038 1.3965 14.5586 0.6243 0.7335 SwinFusion 56.1268 19.8078 5.5398 0.6890 1.1771 14.4677 0.8607 0.2619 MMAE 57.4685 17.6996 4.9085 0.7079 1.4567 15.2774 0.7335 0.7576 BSAFusion 62.4856 22.4648 6.6407 0.6659 1.1634 13.7203 0.6870 0.4304 FATFusion 67.8917 22.1793 6.0303 0.6507 1.1753 13.4980 0.8624 0.7528 MBHFuse 58.2416 20.2579 5.5603 0.7053 1.3947 14.7126 0.8336 0.3152 CrossFuse 39.0810 11.3517 3.1958 0.7480 1.1878 15.5592 0.6658 0.1936 本文方法 65.0521 25.0053 6.0577 0.7896 1.6637 14.7614 0.8098 0.7698