Search

The Direct Epifluorescent Filter Technique (DEFT) and the reference method of counting total bacterial colonies on Petri dishes were compared. IDF Standards 128 (1985) and 161A (1995) were applied. A total of 496 samples of milk were analyzed. Colony forming units per microL milk were transformed to decimal logarithmic units: log (cfu/microL). The repeatability standard deviation, Sr = 0.114, was typical for a routine microbiological method. To study the carryover at different levels of bacteria, 3 tests were performed on milk samples of approximately 100, 700, and >1000 cfu/microL. For the first 2 experiments, no carryover was detected; in the milk sample with >1000 cfu/microL, the carryover was <0.12%. When the DEFT counts were regressed versus the reference method, the values of the slope and intercept were 0.92 and 0.17, respectively; the correlation coefficient was r = 0.84; and the residual standard deviation was Syx = 0.287. The paired t-test showed that the reference method and DEFT do not give significantly different results (p = 0.05).

Fat, protein, and total solids in fermented milk were determined by near-infrared reflectance spectroscopy (NIRS). A set of 107 samples from diverse types of fermented milk (whole, skimmed, with added flavors, and without added flavors) were used to calibrate the instrument by modified partial least-square regression. Global calibration, using all samples, was more effective than specific calibrations using fewer samples of each type. Standard errors of a calibration were 0.071 for fat, 0.075 for protein, and 0.083 for total solids. Values of the correlation coefficient square (R2) were 0.997 for fat, 0.981 for protein, and 1.000 for total solids. Calibration was validated with an independent set of 34 samples of the same types. Standard errors of validation were 0.08, 0.14, and 0.25 for fat, protein, and total solids, respectively, and values of R2 for the regression of measurements by NIRS versus reference methods were 1.00 for fat and total solids and 0.94 for protein. When Standard 128 of the International Dairy Federation was used to compare NIRS results with those from reference methods, no significant differences were found (p = 0.05).