Author Affiliations
1 School of Information Sciences & Technology, Northwest University, Xi'an, Shaanxi 710127, China;2 Key Laboratory for Radiomics and Intelligent Sense of Xi'an, Northwest University, Xi'an, Shaanxi 710127, Chinashow less
Fig. 1. Framework of FMT rapid reconstruction based on autoencoder
Fig. 2. Diagram of the non-homogeneous cylinder phantom. (a) Model of non-homogeneous cylinder phantom; (b) distribution of shot points at plane of z=15 mm
Fig. 3. Result diagram of single-source reconstruction. (a)-(c) Stereogram of reconstruction results, the 2D cross-section views at planes of x=0 mm and z=15 mm with original data; (d)-(f) stereogram of reconstruction results, the 2D cross-section views at planes of x=0 mm and z=15 mm with AE method
Fig. 4. T1 and T2 reconstruction results using AE method under different dimensionality. (a) Reconstruction results of T1; (b) reconstruction results of T2
Fig. 5. Result diagram of double-sources reconstruction. (a)-(c) Stereogram of reconstruction results, the 2D cross-section views at planes of x=0 mm and z=15 mm with original data; (d)-(f) stereogram of reconstruction results, the 2D cross-section views at planes of x=0 mm and z=15 mm with AE method
Fig. 6. Single source reconstruction results in digital mouse experiments using AE method under different dimensionality
Fig. 7. Results of single source in digital mouse experiments. (a)-(c) Stereogram of reconstruction results, the 2D cross-section views at planes of x=11.9 mm and z=16.4 mm with original data; (d)-(f) stereogram of reconstruction results, the 2D cross-section views at planes of x=11.9 mm and z=16.4 mm with AE method
Fig. 8. Reconstruction results of T1 and T2 when compressed to different dimensions using AE. (a) Reconstruction results of T1; (b) reconstruction results of T2
Fig. 9. Results of double-sources in digital mouse experiments. (a)-(c) Stereogram of reconstruction results, the 2D cross-section views at planes of x=11.9 mm and z=16.4 mm with original data; (d)-(f) stereogram of reconstruction results, the 2D cross-section views at planes of x=11.9 mm and z=16.4 mm with AE method
Organ | /mm-1 | /mm-1 | /mm-1 | /mm-1 |
---|
Muscle | 0.0052 | 10.80 | 0.0068 | 10.30 | Heart | 0.0083 | 6.73 | 0.0104 | 6.60 | Lungs | 0.0133 | 19.70 | 0.0203 | 19.50 | Liver | 0.0329 | 7.00 | 0.0176 | 6.60 | Bone | 0.0060 | 60.09 | 0.0030 | 30.74 |
|
Table 1. Optical parameters for non-homogeneous cylinder phantom
Method (dimensionality) | LE /mm | WCLE /mm | NRMSE /mm-1 | Time /s | Dice |
---|
IVTCG | 0.73 | 0.2546 | 3.3640 | 7.68 | 0.5200 | AE+IVTCG(50) | 0.73 | 0.2438 | 0.0239 | 4.22 | 0.6667 | AE+IVTCG(100) | 0.73 | 0.1789 | 0.0186 | 3.98 | 0.6667 | AE+IVTCG(150) | 0.79 | 0.2621 | 0.0204 | 4.37 | 0.6000 | AE+IVTCG(200) | 0.79 | 0.4100 | 0.0280 | 4.48 | 0.5200 | AE+IVTCG(300) | 1.26 | 0.6047 | 0.0318 | 4.76 | 0.4444 |
|
Table 2. Reconstruction results of single source non-homogeneous cylinder phantom simulation experiments using AE method
Number of excitation source | LE /mm | WCLE /mm | NRMSE /mm-1 | Time /s | Dice |
---|
36 | 0.73 | 0.1789 | 0.0186 | 3.98 | 0.6667 | 18 | 0.73 | 0.2094 | 0.0334 | 4.21 | 0.6667 | 9 | 0.73 | 0.5235 | 0.0236 | 4.85 | 0.6667 | 6 | 1.17 | 0.4118 | 0.0474 | 4.02 | 0.4700 | 3 | 1.26 | 1.2600 | 0.0534 | 4.17 | 0.4000 |
|
Table 3. Quantitative simulation results of single source non-homogeneous cylinder using AE method under different number of excitation sources
Noise level /% | LE /mm | WCLE /mm | NRMSE /mm-1 | Time /s | Dice |
---|
5 | 0.73 | 0.6723 | 0.0152 | 4.09 | 0.6667 | 10 | 0.73 | 0.6879 | 0.0159 | 4.20 | 0.6667 | 15 | 0.73 | 0.1461 | 0.0192 | 4.33 | 0.6667 | 20 | 0.79 | 0.2033 | 0.0179 | 4.01 | 0.6000 | 25 | 1.26 | 0.7213 | 0.0164 | 4.80 | 0.4000 |
|
Table 4. Quantitative simulation results of single source non-homogeneous cylinder using AE method under different noise levels
Method | Target | LE /mm | WCLE /mm | NRMSE /mm-1 | Time /s | Dice |
---|
IVTCG | T1 | 0.6169 | 0.4915 | 0.3291 | 24.3438 | 0.2361 | T2 | 1.3971 | 0.9369 | AE+IVTCG | T1 | 0.6169 | 0.4826 | 0.0491 | 10.5170 | 0.4444 | T2 | 1.3286 | 0.4598 |
|
Table 5. Quantitative simulation results of double sources non-homogeneous cylinder phantom simulation experiments using AE method
Method | LE /mm | WCLE /mm | NRMSE /mm-1 | Time /s | Dice |
---|
IVTCG | 0.40425 | 0.4268 | 0.2509 | 5.8112 | 0.4000 | AE+IVTCG | 0.40425 | 0.5635 | 0.0343 | 1.7308 | 0.5700 |
|
Table 6. Quantitative simulation results of single source in digital mouse experiments using AE method
Method | Target | LE /mm | WCLE /mm | NRMSE /mm-1 | Time /s | Dice |
---|
IVTCG | T1 | 0.6169 | 1.2876 | 0.2617 | 47.776 | 0.3333 | T2 | 0.6833 | 0.9369 | AE+IVTCG | T1 | 0.6169 | 1.2876 | 0.0220 | 20.171 | 0.4000 | T2 | 0.5169 | 0.4579 |
|
Table 7. Quantitative simulation results of double sources in digital mouse experiments using AE method