Search

We present a compact, fast, and versatile fiber-optic probe system for real-time determination of tissue optical properties from spatially resolved continuous-wave diffuse reflectance measurements. The system collects one set of reflectance data from six source-detector distances at four arbitrary wavelengths with a maximum overall sampling rate of 100 Hz. Multivariate calibration techniques based on two-dimensional polynomial fitting are employed to extract and display the absorption and reduced scattering coefficients in real-time mode. The four wavelengths of the current configuration are 660, 785, 805, and 974 nm, respectively. Cross-validation tests on a 6 x 7 calibration matrix of Intralipid-dye phantoms showed that the mean prediction error at, e.g., 785 nm was 2.8% for the absorption coefficient and 1.3% for the reduced scattering coefficient. The errors are relative to the range of the optical properties of the phantoms at 785 nm, which were 0-0.3/cm for the absorption coefficient and 6-16/cm for the reduced scattering coefficient. Finally, we also present and discuss results from preliminary skin tissue measurements.

Absorption (mu(a)) and reduced scattering (mu(s)') spectra of turbid media were quantified with a noncontact imaging approach based on a Fourier-transform interferometric imaging system (FTIIS). The FTIIS was used to collect hyperspectral images of the steady-state diffuse reflectance from turbid media. Spatially resolved reflectance data from Monte Carlo simulations were fitted to the recorded hyperspectral images to quantify mu(a) and mu(s)' spectra in the 550-850-nm region. A simple and effective calibration approach was introduced to account for the instrument response. With reflectance data that were close to and far from the source (0.5-6.5 mm), mu(a) and mu(s)' of homogeneous, semi-infinite turbid phantoms with optical property ranges comparable with those of tissues were determined with an accuracy of +/-7% and +/-3%, respectively. Prediction accuracy for mu(a) and mu(s)' degraded to +/-12% and +/-4%, respectively, when only reflectance data close to the source (0.5-2.5 mm) were used. Results indicate that reflectance data close to and far from the source are necessary for optimal quantification of mu(a) and mu(s)'. The spectral properties of mu(a) and mu(s)' values were used to determine the concentrations of absorbers and scatterers, respectively. Absorber and scatterer concentrations of two-chromophore turbid media were determined with an accuracy of +/-5% and +/-3%, respectively.

We present a new, to our knowledge, method for extracting optical properties from integrating sphere measurements on thin biological samples. The method is based on multivariate calibration techniques involving Monte Carlo simulations, multiple polynomial regression, and a Newton-Raphson algorithm for solving nonlinear equation systems. Prediction tests with simulated data showed that the mean relative prediction error of the absorption and the reduced scattering coefficients within typical biological ranges were less than 0.3%. Similar tests with data from integrating sphere measurements on 20 dye-polystyrene microsphere phantoms led to mean errors less than 1.7% between predicted and theoretically calculated values. Comparisons showed that our method was more robust and typically 5-10 times as fast and accurate as two other established methods, i.e., the inverse adding-doubling method and the Monte Carlo spline interpolation method.

We describe a method for determining the reduced scattering and absorption coefficients of turbid biological media from the spatially resolved diffuse reflectance. A Sugeno Fuzzy Inference System in conjunction with data preprocessing techniques is employed to perform multivariate calibration and prediction on reflectance data generated by Monte Carlo simulations. The preprocessing consists of either a principal component analysis or a new, extended curve-fitting procedure originating from diffusion theory. Prediction tests on reflectance data with absorption coefficients between 0.04 and 0.06 mm(-1) and reduced scattering coefficients between 0.45 and 0.99 mm(-1) show the root-mean-square error of this method to be 0.25% for both coefficients. With reference to practical applications, we also describe how the prediction accuracy is affected by using relative instead of absolute reflectance data, by imposing measurement noise on the reflectance data, and by changing the number and the position of detectors.