Evaluation of cooked rice texture by near infrared spectroscopy | |
| Author | Weena Srisawas |
| Call Number | AIT Diss. no.FB-09-04 |
| Subject(s) | Rice Cookery (Rice) Near infrared spectroscopy |
| Note | A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Food Engineering and Bioprocess Technology, School of Environment, Resources and Development |
| Publisher | Asian Institute of Technology |
| Series Statement | Dissertation ; no. FB-09-04 |
| Abstract | Rice eating quality is influenced not only by the individual preference but also by the variety, postharvest handling and storage conditions, milling process and cooking method. Besides sensory evaluation, several instrumental measurement techniques have been proposed for predicting cooked rice eating quality attributes. This research investigated the potential of an alternative non-destructive test technique using near infrared spectroscopy (NIRS) for quantifying cooked rice eating quality characteristics as compared to the conventional test procedures based on the physicochemical and pasting properties of milled rice, and instrumental measurements of cooked rice textural attributes. Rice samples were collected from different locations throughout Thailand to cover wide range of rice varieties and properties. A total of 230 indica-type rice samples consisting of 43 varieties was analyzed for apparent amylase content (AAC), Kjeldahl protein content (KPC), alkali spreading value (ASV), gel consistency (GEL) and grain elongation ratio (ELR). The pasting properties of milled rice were measured by a Rapid Visco Analyzer (RVA). The NIR spectra of both whole grain milled rice and flour samples in 1100-2500 nm wavelength range were used for the prediction of physicochemical and pasting properties of rice. The moving window partial least square regression (MWPLSR) resulted in the informative region for predicting AAC and KPC from the milled rice spectra in 1160-1950 and 1108-1804 nm wavelength ranges with corresponding improvement of 13 and 16% standard error of prediction (SEP), respectively. The models developed using partial least square regression (PLSR) for predicting AAC from NIR spectra of rice flour showed better accuracy than the milled rice spectra with coefficient of determination (r²) and SEP of 0.945 and 1.22%, respectively for the validation data set. In contrast, the prediction of KPC was more accurate based on NIR spectra of whole grain milled rice indicating r² of 0.955 and SEP of 0.19%. Models for other physicochemical properties such as ASV (r² = 0.737) and GEL (r² = 0.654) were only moderately accurate. None of the pasting properties was successfully modeled by NIRS with high accuracy. However, the PLSR models showed relatively better performance for the secondary parameters derived from the selected viscosities known as the primary parameters on the RVA viscogram. Among three primary parameters consisting of peak, hot paste and cooled paste viscosities, only peak viscosity could be modeled (r² = 0.678) by NIRS. In case of the secondary parameters, the model developed for breakdown viscosity was better (r² = 0.735) than the models for predicting setback (r² = 0.605) and total setback (r² = 0.522) viscosities. The eating quality of cooked rice is conventionally assessed by a well-trained sensory panel in a straightforward manner. However, such procedures are cumbersome involving high cost and labor, and not practical for routine quality control. Instrumental evaluation of cooked rice texture is an alternative provided accurate predictions of sensory attributes could be made. One major factor affecting cooked rice eating quality has been known to be the water uptake upon cooking. However, this factor has been conspicuously overlooked by the most researchers. The effects of cooking water-to-rice (W/R) ratio on sensory characterization of cooked rice eating quality for 14 varieties of Thai rice were investigated in relation to their physicochemical and pasting properties. Milled rice samples were cooked using five W/R ratios ranging from 1.3 to 2.5 on weight basis and analyzed by twelve trained sensory evaluation panelists, instrumental measurements using texture profile analysis (TPA) and back extrusion testing (BET), and NIRS. Three-way ANOVA and principal component analysis of sensory eating quality attributes indicated the intensity of sensory hardness to be the main characteristic of cooked rice. Sensory hardness (Hs) of cooked rice decreased with increasing WIR ratio whereas sensory stickiness (Ss) showed opposite trend. The overall acceptability scores based on visual, texture and flavor attributes reached peak levels and corresponded to the optimum W IR ratios for different rice varieties, and were highly related with Hs and Ss. Thus, PLSR models could be used first for determining the optimum W IR ratio, Hs and Ss with r² values of 0.991,0.966 and 0.969, respectively from the physicochemical properties of milled rice. Subsequently, peak overall acceptability scores corresponding to the optimum WIR ratio could be estimated from the developed linear relationships with Hs, Ss or SslHs with reasonable accuracy (r² = 0.86-0.94). The sensory textural attributes of cooked rice could be successfully predicted by the models developed from NIR spectra of milled or cooked rice. The PLSR models based on NIR spectra of cooked rice predicted Hs and Ss slightly better than glossiness (Gs) with r² and SEP in the ranges of 0.89-0.92 and 0.31-0.32 unit score on 9-point sensory intensity scales, respectively. Results showed that NIRS-based models predicted Hs, Ss and Gs scores of cooked rice with higher accuracy and less SEP compared to the TPA and BET-based models. However, the performance of a back-propagation neural network (BPNN) was superior compared to the PLSR models for predicting cooked rice sensory textural attributes with r² and SEP values ranging from 0.925-0.945 and 0.21-0.29 sensory score, respectively. Models developed from NIR spectra of cooked rice resulted in almost the same accuracy as indicated by the predictive models based on the physicochemical properties with the SEP of 0.2 to 0.3 unit sensory score. The direct predictions based on NIR spectra of cooked rice were independent of W IR ratio and thus offered an added advantage over the two-step predictions by NIR spectra of milled rice. In summary, the eating quality attributes of cooked rice such as Hs, Ss and Gs could be successfully estimated with reasonable accuracy from the physicochemical properties, NIR spectra of cooked rice, two-step prediction from NIR spectra of milled rice, pasting characteristics, and instrumental textural measurements by TPA and BET, respectively in decreasing order of the performance of developed models. |
| Year | 2009 |
| Corresponding Series Added Entry | Asian Institute of Technology. Dissertation ; no. FB-09-04 |
| Type | Dissertation |
| School | School of Environment, Resources, and Development (SERD) |
| Department | Department of Food, Agriculture and Natural Resources (Former title: Department of Food Agriculture, and BioResources (DFAB)) |
| Academic Program/FoS | Food Engineering and Bioprocess Technology (FB) |
| Chairperson(s) | Athapol Noomhorm;Jindal, Vinod K.; |
| Examination Committee(s) | Rakshit, Sudip Kumar;Manukid Parnichkun;Warunee Thanapase; |
| Scholarship Donor(s) | AIT Fellowship; |
| Degree | Thesis (Ph.D.) - Asian Institute of Technology, 2009 |