Transfer learning and wavelength selection method in NIR spectroscopy to predict glucose and lactate concentrations in culture media using VIP-Boruta.

Regression models are constructed to predict glucose and lactate concentrations from near-infrared spectra in culture media. The partial least-squares (PLS) regression technique is employed, and we investigate the improvement in the predictive ability of PLS models that can be achieved using wavelength selection and transfer learning. We combine Boruta, a nonlinear variable selection method based on random forests, with variable importance in projection (VIP) in PLS to produce the proposed variable selection method, VIP-Boruta. Furthermore, focusing on the situation where both culture medium samples and pseudo-culture medium samples can be used, we transfer pseudo media to culture media. Data analysis with an actual dataset of culture media and pseudo media confirms that VIP-Boruta can effectively select appropriate wavelengths and improves the prediction ability of PLS models, and that transfer learning with pseudo media enhances the predictive ability. The proposed method could reduce the prediction errors by about 61% for glucose and about 16% for lactate, compared to the traditional PLS model.

Glucose Monitoring in Cell Culture with Online Ultrasound-Assisted Near-Infrared Spectroscopy.

Near-infrared (NIR) spectroscopic analysis for suspensions often requires a pretreatment by dilution or filtering of suspended solids to maintain the sensitivity of the measurements. An online ultrasound-assisted spectroscopy (UAS) unit enabling pretreatment-free and noncontact analysis for bioprocessing is proposed and evaluated with a model suspension containing 3-µm-diameter polystyrene latex particles (PSLs) with the density of 5.1 × 108 counts/ml and Chinese hamster ovary (CHO) cell culture whose cell density was 3.2 × 107 cells/ml. The online UAS uses acoustic radiation force generated by ultrasonic standing waves. Suspended matter such as the PSLs and CHO cells can be localized at nodal planes in the suspensions by the acoustic radiation force. Hence, in the case of the online UAS, incoming light can pass through the suspensions more easily than that in the case of a conventional spectroscopy. Its effectiveness was evaluated by the predictive capability of a calibration model for glucose concentrations in the model suspensions and the CHO cell cultures. The calibration models were constructed by use of a partial least-squares regression (PLS-R) in the range of 4900-4200 cm-1 region after the pretreatment of second-order Savitzky-Golay filter. The calibration model built from the NIR spectra acquired with the online UAS could predict the glucose concentration in the CHO cell cultures with a measurement error of 0.6%. It was validated that the glucose concentrations in the flowing model suspension were able to be monitored by the online UAS with a measurement error of 8%. The newly developed online UAS for cell culture monitoring allows us to promote a wider use of NIR spectroscopy. For example, in the applications to the biopharmaceutical and cell-therapy industries, the online UAS enables simpler and easier monitoring of cell cultures because cleaning and sterilization of monitoring tools after cell culturing can be eliminated.

9-kW peak power and 150-fs duration blue-violet optical pulses generated by GaInN master oscillator power amplifier.

Blue-violet optical pulses of 9-kW peak power and 150-fs duration were obtained from a GaInN master oscillator power amplifier system using a nonlinear dispersion compensator. Seed pulses from a dispersion-compensated GaInN mode-locked semiconductor laser diode were stretched to 3-ps duration using a nonlinear dispersion compensator with a spatial light modulator that added second-order phase dispersion to an optimized nonlinear phase dispersion compensating the higher-order dispersion of the optical pulses. The stretched phase-optimized pulses were efficiently amplified to 3.0 nJ by a GaInN semiconductor optical amplifier. The amplified pulses were subsequently compressed using a linear pulse compressor, yielding 1.4-nJ femtosecond pulses. The obtained results show the highest peak-power ever reported for an electrically-pumped semiconductor gain medium.

140-fs duration and 60-W peak power blue-violet optical pulses generated by a dispersion-compensated GaInN mode-locked semiconductor laser diode using a nonlinear pulse compressor.

Blue-violet optical pulses of 140-fs duration and 60-W peak power were obtained from a dispersion-compensated GaInN mode-locked semiconductor laser diode using a nonlinear pulse compression technique. Wavelength-dependent group velocity dispersion expressed by third-order phase dispersion was applied to the optical pulses using a pulse compressor with a spatial light modulator. The obtained optical pulses had the shortest duration ever obtained for a mode-locked semiconductor laser diode using edge-emitting type devices.

A high-peak-power UV picosecond-pulse light source based on a gain-switched 1.55 microm laser diode and its application to time-resolved spectroscopy of blue-violet materials.

We generated sub-kilowatt peak-power and 6-ps duration 390-nm optical pulses via the fourth harmonic generation of amplified optical output from a gain-switched 1.55-microm laser diode. We obtained a power-conversion-efficiency of 12% from 1.55-microm to 390-nm light, and subsequently applied the ultraviolet pulses to time-resolved spectroscopy of blue-violet luminescent materials, including a Coumarine dye solution and nitride semiconductor materials using single-photon and two-photon excitation schemes.