Author Affiliations
Terahertz Technology Innovation Research Institute, Terahertz Spectrum and Imaging Technology Cooperative Innovation Center, Shanghai Key Lab of Modern Optical System, University of Shanghai for Science and Technology, Shanghai 200093, Chinashow less
Fig. 1. Spectra of four pure samples of L-Glu, D-MI, CMH, and GABA
Fig. 2. Measured spectra and calculated spectra. (a) Mixture sample of L-Glu, D-MI, and CMH; (b) mixture sample of all four components
Fig. 3. Spectra of 10 mixture samples. (a) Before wavelet transform; (b) after wavelet transform
Fig. 4. RMSE of three SVR parameters based on leave-one-out cross validation. (a) Parameter c when g=0.01and e=0.01; (b) parameter g when c=0.25 and e=0.01; (c) parameter e when c=0.25 and g=0.01
Fig. 5. Actual and predicted concentrations of 10 mixture samples. (a) NAA; (b) NE
Component | Actual concentrationna1 /% | Calculated concentrationnc1 /% | Root mean squarederror ERMS1 | Average root meansquared error /% |
---|
L-Glu | 4.65 | 4.65 | 0 | | D-MI | 4.65 | 4.55 | 0.0215 | 4.65 | CMH | 4.65 | 4.10 | 0.1180 | |
|
Table 1. Concentration results of mixture sample of L-Glu, D-MI and CMH
Component | Actual concentrationna2 /% | Calculated concentrationnc2 /% | Root mean squarederror ERMS2 | Average root meansquared error /% |
---|
L-Glu | 4.90 | 4.50 | 0.0816 | | D-MI | 4.90 | 4.90 | 0 | 5.44 | CMH | 4.90 | 4.60 | 0.0612 | | GABA | 4.90 | 4.80 | 0.0204 | |
|
Table 2. Concentration results of mixture sample of all four components
Samplenumber | NAA massm1 /mg | NE massm2 /mg | Mass of otherfive componentsm3 /mg |
---|
1 | 3.02 | 1.13 | - | 2 | 9.85 | 7.02 | - | 3 | 11.99 | 15.08 | - | 4 | 9.20 | 3.82 | - | 5 | 1.20 | 9.98 | - | 6 | 4.13 | 8.91 | - | 7 | 15.20 | 12.03 | - | 8 | 12.87 | 3.19 | - | 9 | 4.86 | 4.80 | - | 10 | 7.02 | 13.03 | - |
|
Table 3. Parameters of mixture samples including NAA and NE
Component | RMSE ERMS | Average RMSE | Correlationcoefficient R | Average correlationcoefficient |
---|
NAA | 0.0040 | | 0.9913 | | NE | 0.0040 | 0.0040 | 0.9914 | 0.99135 |
|
Table 4. RMSE and correlation coefficient between predicted and actual concentrations of NAA and NE in mixture
Number ofsamples | RMSE ERMSn | Correlationcoefficient Rn |
---|
6 | 0.0125 | 0.9026 | 8 | 0.0055 | 0.9853 | 10 | 0.0040 | 0.9914 |
|
Table 5. Accuracy of algorithm models under different sample numbers
Algorithm | RMSE ERMSa | Correlationcoefficient Ra |
---|
Partial least squares | 0.0231 | 0.8052 | BP nerve network | 0.0175 | 0.8353 | Support vector regression | 0.0040 | 0.9914 |
|
Table 6. Prediction accuracy of different algorithms