Nondestructive determination and visualization of protein and carbohydrate concentration of Chlorella pyrenoidosa in situ using hyperspectral imaging technique

As a new food resource approved by the National Health Commission of China, Chlorella pyrenoidosa (C. pyrenoidosa) has become commercialized in recent years with the advantages of comprehensive nutrition, especially rich in protein content. Monitoring the growth information of C. pyrenoidosa is crucial for optimizing the culture environment and increasing microalgae yield. Traditional methods for the detection of microalgae bioproducts are time-consuming and expensive. In this study, a fast visual and non-invasive method based on visible/near infrared (VIS/NIR) hyperspectral imaging (HSI) combined with chemometric methods was developed to predict the biomass, carbohydrate and protein in the cultures of C. pyrenoidosa. Twelve data preprocessing approaches, 3 feature selection methods, and 4 calibration models were used to establish and optimize the estimation models. The prediction results showed that the effects of autoscaling preprocess combined with CARS-MLR for biomass (R2p = 0.9788, RPD = 7.6503), wavelet transform (WT) combined with iRF-RFR for carbohydrate (R2p = 0.9935, RPD = 27.0385), and S-G preprocess combined with SA-RFR (R2p = 0.9677, RPD = 12.9928) for protein obtained the best performances, respectively. Moreover, visualization maps of the distribution and abundance of these components in the liquid suspension of C. pyrenoidosa were obtained based on the optimal models. This study showed that HSI technology combined with chemometric methods can accurately predict the biomass, carbohydrate, and protein contents of C. pyrenoidosa in situ, which has the potential as a fast and nondestructive approach for monitoring microalgal growth information.