»
NIR »
A total of 24 soils were prepared per active pharmaceutical ingredient (API) corresponding to four replicates per concentration level.
Samples were scanned using a prototype NIR-CI system comprising of a mercury-cadmium-telluride (MCT) detector with a detection range of 1260 - 2500 nm and spectral resolution of 10 nm. The pixel size was 30 × 30 µm and the optical sensitive area 9.2 x 6.9 mm2
. Data acquisition
time to capture a single 384 ×288 × 125 datacube was 5 seconds. The illumination source consisted of four 12V, 20W halogen lamps. This prototype NIR-CI system was designed utilizing components which off er the potential for portable systems with fast acquisition times. To this end, the main benefi t of this system is the fact that the spectrograph employed is a Fabry-Pérot fi lter (Rikola, Oulu, Finland). The benefi ts of Fabry-Pérot fi lters compared to tunable fi lters such as Acousto-Optic Tunable Filters and Liquid Crystal Tunable Filters are their small size and weight, speed of wavelength tuning and high optical throughput [8].
The NIR hyperspectral datacubes obtained were calibrated to correct for the instrument response
by measuring a white standard (I0 )
and the dark current (d) and using equation (1) to convert sample measurement intensity (I) from counts to spectroscopic units [9]. The NIR refl ectance data (R) obtained was then converted into absorbance (A) data by equation (2).
R = (I − d) ∕ (I0 A = log10 (1 ∕ R)
− d)
(1) (2)
Analysis of the NIR, hyperspectral datacubes was conducted using both existing and in-house R software routines (R 2.11.1, R Foundation for Statistical Computing, Vienna, Austria).
The wavelength range employed in the analysis was 1260-2170 nm. The NIR datacubes were processed by means of a median fi lter based on a 3 × 3 square neighborhood to remove speckle noise brought about by defective pixels. The NIR datacubes were then auto-scaled to remove baseline shifts and to standardize the variance. A classifi cation function was then employed to diff erentiate stainless steel pixels from residue containing pixels. The classifi cation function was based on the histograms (intensity in the x-axis and the number of pixels per intensity value in the y-axis) of the blank and API soiled stainless steel coupons. Whereas the blank coupons followed a standard normal distribution in the NIR range understudy, soiled coupons with API residues showed a bimodal density. Also, stainless steel coupons had higher absorption values across the NIR range studied (1260 – 2170 nm). The cut-off point in the classifi cation function for detecting residue pixels was the value corresponding to the lower one percentile (P < 0.01) of the standard normal distribution (z = -2.326). Since the secondary peak corresponding to residue pixels causes the corresponding auto-scaled stainless steel pixels to bias away from mean zero, an estimate of the bias was obtained (distance between the primary peak and zero) and subsequently removed to center the density of the stainless steel pixels on zero.
This classifi cation technique was based on two wavelengths for each API. The selected wavelengths corresponded to those wavelengths that resulted in the lowest standard error associated with a linear regression of number of API classifi ed pixels and applied soil concentration. Quantifi cation of API residues on stainless steel coupons was then conducted by building a calibration model based on linear regression of
106 | | September/October 2013 - 15TH ANNIVERSARY ISSUE
number of pixels identifi ed as residue by the classifi cation function versus concentration (µg residue 50 mm-2). The 95% Confi dence Interval (CI) and 95% Prediction Interval of the linear models were also calculated [10].
The Limit of Detection was also calculated according to the International Cooperation on Harmonisation (ICH) [11] using the equation:
Limit of Detection (LOD) = y + 3.3 × σ
Where y is the residue pixel count for a blank sample and σ is the standard error.
Results and Discussion
Classifi cation of API Residues and Stainless Steel Background
The auto-scaled hyperspectral datacubes corresponding to the diff erent API soils were masked using a classifi cation function to discriminate pixels as API residue or stainless steel background. The thresholded masked images of stainless steel coupons soiled with diff erent concentrations of sulfadimidine sodium salt and sulfacetamide sodium salt are displayed in Figure 2. In these masked images, at any given wavelength, pixels classifi ed as API residue pixels are displayed using their raw absorbance values whereas stainless steel pixels are set to the minimum raw absorbance value of the API classifi ed pixels. It can be observed in Figure 2 that the API soils were correctly extracted from the stainless steel background by the classifi cation function. The two wavebands selected for use in the classifi cation function were 1580 and 2140 nm in the case of sulfadimidine sodium salt, and 1480 and 2140 nm for sulfacetamide sodium salt. It is important to note that this selection was conducted by assessing which combination of two wavelengths from the studied range resulted in linear models of API classifi ed pixels versus applied residue concentration with a minimum standard error associated.
The images presented in Figure 2 suggest that the drying pattern of the API soils was not homogeneous for both APIs under study. The dried API soils showed a ring-like structure with the API solids concentrating on the ring with only thin deposits of API observed inside the ring. Previous
Figure 2. Masked Chemical Images of API residues on stainless steel coupons. Row (a) sulfadimidine sodium salt displayed at 1580 nm (b) sulfacetamide sodium salt displayed at 1480 nm.
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56 |
Page 57 |
Page 58 |
Page 59 |
Page 60 |
Page 61 |
Page 62 |
Page 63 |
Page 64 |
Page 65 |
Page 66 |
Page 67 |
Page 68 |
Page 69 |
Page 70 |
Page 71 |
Page 72 |
Page 73 |
Page 74 |
Page 75 |
Page 76 |
Page 77 |
Page 78 |
Page 79 |
Page 80 |
Page 81 |
Page 82 |
Page 83 |
Page 84 |
Page 85 |
Page 86 |
Page 87 |
Page 88 |
Page 89 |
Page 90 |
Page 91 |
Page 92 |
Page 93 |
Page 94 |
Page 95 |
Page 96 |
Page 97 |
Page 98 |
Page 99 |
Page 100 |
Page 101 |
Page 102 |
Page 103 |
Page 104 |
Page 105 |
Page 106 |
Page 107 |
Page 108 |
Page 109 |
Page 110 |
Page 111 |
Page 112 |
Page 113 |
Page 114 |
Page 115 |
Page 116 |
Page 117 |
Page 118 |
Page 119 |
Page 120 |
Page 121 |
Page 122 |
Page 123 |
Page 124 |
Page 125 |
Page 126 |
Page 127 |
Page 128 |
Page 129 |
Page 130 |
Page 131 |
Page 132 |
Page 133 |
Page 134 |
Page 135 |
Page 136 |
Page 137 |
Page 138 |
Page 139 |
Page 140 |
Page 141 |
Page 142 |
Page 143 |
Page 144 |
Page 145 |
Page 146 |
Page 147 |
Page 148 |
Page 149 |
Page 150 |
Page 151 |
Page 152 |
Page 153 |
Page 154 |
Page 155 |
Page 156