Aim
We aim to develop a use-error-robust ML algorithm for optical property prediction from DRS spectra. Approach: We developed a wavelength-independent regressor (WIR) to predict optical properties from DRS data. For validation, we generated 1520 simulated DRS spectra with the forward Monte Carlo model, where μa=0.44<math><mrow><msub><mi>μ</mi><mi>a</mi></msub><mo>=</mo><mn>0.44</mn></mrow></math> to 2.45cm−1<math><mrow><mn>2.45</mn><mtext> </mtext><msup><mrow><mi>cm</mi></mrow><mrow><mo>-</mo><mn>1</mn></mrow></msup></mrow></math>, and μ′s=6.53<math><mrow><msubsup><mrow><mi>μ</mi></mrow><mrow><mi>s</mi></mrow><mrow><mo>'</mo></mrow></msubsup><mo>=</mo><mn>6.53</mn></mrow></math> to 9.58cm−1<math><mrow><mn>9.58</mn><mtext> </mtext><msup><mrow><mi>cm</mi></mrow><mrow><mo>-</mo><mn>1</mn></mrow></msup></mrow></math>. We introduced common use-errors, such as wavelength miscalibrations and intensity fluctuations. Finally, we collected 882 experimental DRS images from 170 tissue-mimicking phantoms and compared performances of the WIR model, a dense neural network, and the MCI model.
Conclusions
The WIR model presents reliable use-error-robust optical property predictions from DRS data.
Results
When compounding all use-errors on simulated data, the WIR model best balanced accuracy and speed, yielding errors of 1.75% for μa<math><mrow><msub><mi>μ</mi><mi>a</mi></msub></mrow></math> and 1.53% for μ′s<math><mrow><msubsup><mrow><mi>μ</mi></mrow><mrow><mi>s</mi></mrow><mrow><mo>'</mo></mrow></msubsup></mrow></math>, compared to the MCI's 50.9% for μa<math><mrow><msub><mi>μ</mi><mi>a</mi></msub></mrow></math> and 24.6% for μ′s<math><mrow><msubsup><mrow><mi>μ</mi></mrow><mrow><mi>s</mi></mrow><mrow><mo>'</mo></mrow></msubsup></mrow></math>. Regarding experimental data, WIR model had mean errors of 13.2% and 6.1% for μa<math><mrow><msub><mi>μ</mi><mi>a</mi></msub></mrow></math> and μ′s<math><mrow><msubsup><mrow><mi>μ</mi></mrow><mrow><mi>s</mi></mrow><mrow><mo>'</mo></mrow></msubsup></mrow></math>, respectively. The errors for MCI were about eight times higher. Conclusions: The WIR model presents reliable use-error-robust optical property predictions from DRS data.
Significance
Machine learning (ML)-enabled diffuse reflectance spectroscopy (DRS) is increasingly used as an alternative to the computation-intensive inverse Monte Carlo (MCI) simulation to predict tissue's optical properties, including the absorption coefficient, μa<math><mrow><msub><mi>μ</mi><mi>a</mi></msub></mrow></math> and reduced scattering coefficient, μ′s<math><mrow><msubsup><mrow><mi>μ</mi></mrow><mrow><mi>s</mi></mrow><mrow><mo>'</mo></mrow></msubsup></mrow></math>. Aim: We aim to develop a use-error-robust ML algorithm for optical property prediction from DRS spectra. Approach: We developed a wavelength-independent regressor (WIR) to predict optical properties from DRS data. For validation, we generated 1520 simulated DRS spectra with the forward Monte Carlo model, where μa=0.44<math><mrow><msub><mi>μ</mi><mi>a</mi></msub><mo>=</mo><mn>0.44</mn></mrow></math> to 2.45cm−1<math><mrow><mn>2.45</mn><mtext> </mtext><msup><mrow><mi>cm</mi></mrow><mrow><mo>-</mo><mn>1</mn></mrow></msup></mrow></math>, and μ′s=6.53<math><mrow><msubsup><mrow><mi>μ</mi></mrow><mrow><mi>s</mi></mrow><mrow><mo>'</mo></mrow></msubsup><mo>=</mo><mn>6.53</mn></mrow></math> to 9.58cm−1<math><mrow><mn>9.58</mn><mtext> </mtext><msup><mrow><mi>cm</mi></mrow><mrow><mo>-</mo><mn>1</mn></mrow></msup></mrow></math>. We introduced common use-errors, such as wavelength miscalibrations and intensity fluctuations. Finally, we collected 882 experimental DRS images from 170 tissue-mimicking phantoms and compared performances of the WIR model, a dense neural network, and the MCI model. Results: When compounding all use-errors on simulated data, the WIR model best balanced accuracy and speed, yielding errors of 1.75% for μa<math><mrow><msub><mi>μ</mi><mi>a</mi></msub></mrow></math> and 1.53% for μ′s<math><mrow><msubsup><mrow><mi>μ</mi></mrow><mrow><mi>s</mi></mrow><mrow><mo>'</mo></mrow></msubsup></mrow></math>, compared to the MCI's 50.9% for μa<math><mrow><msub><mi>μ</mi><mi>a</mi></msub></mrow></math> and 24.6% for μ′s<math><mrow><msubsup><mrow><mi>μ</mi></mrow><mrow><mi>s</mi></mrow><mrow><mo>'</mo></mrow></msubsup></mrow></math>. Regarding experimental data, WIR model had mean errors of 13.2% and 6.1% for μa<math><mrow><msub><mi>μ</mi><mi>a</mi></msub></mrow></math> and μ′s<math><mrow><msubsup><mrow><mi>μ</mi></mrow><mrow><mi>s</mi></mrow><mrow><mo>'</mo></mrow></msubsup></mrow></math>, respectively. The errors for MCI were about eight times higher. Conclusions: The WIR model presents reliable use-error-robust optical property predictions from DRS data.