Annals of Occupational Hygiene Advance Access originally published online on August 6, 2004
Annals of Occupational Hygiene 2004 48(6):519-532; doi:10.1093/annhyg/meh049
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
© 2004 British Occupational Hygiene Society Published by Oxford University Press;
Development and Evaluation of a Quantitative Video-fluorescence Imaging System and Fluorescent Tracer for Measuring Transfer of Pesticide Residues from Surfaces to Hands with Repeated Contacts
1 Battelle Memorial Institute, 505 King Avenue, Columbus, OH 43201, USA; 2 US Environmental Protection Agency, National Exposure Research Laboratory E205-04, Research Triangle Park, NC 27711, USA
* Author to whom correspondence should be addressed. Fax: +1 614 424 3638; e-mail: nishiomg{at}battelle.org
Received 21 June 2002; in final form 30 December 2003
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
A video imaging system and the associated quantification methods have been developed for measurement of the transfer of a fluorescent tracer from surfaces to hands. The highly fluorescent compound riboflavin (vitamin B2), which is also water soluble and non-toxic, was chosen as the tracer compound to simulate the transfer from surfaces to hands of pesticide residues deposited on carpeted and laminate surfaces of a residence. The system was designed around the unique properties of riboflavin. Excitation energy was centered near 440 nm (in the blue region of the visible spectrum); emitted energy was measured at 600 nm (in the red/orange region), well beyond the significant fluorescence peak maximum of natural skin. A video camera system with an image intensifier was interfaced to an image processing analysis software system. Quantification utilized chemometric techniques to account for the non-linearity of pixel detectivity and non-linear excitation strength. Method quantification and detection limits were approximately 0.1 and 0.02 µg/cm2, respectively. The relative error was
100% at the quantification limit, but <20% at higher levels. Transfer of riboflavin to hands, resulting in dermal loadings in the range 0.1–2.0 µg/cm2, were measured with this system from surfaces whose loadings approximated the pesticide levels that occur in homes after broadcast application.
Keywords: dermal exposure • pesticide exposure assessment • video-fluorescence imaging
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
Video-fluorescence imaging has been, and continues to be, a powerful technique for assessing dermal exposures in the occupational arena. Because of the ability to display both intensity and spatial information, the magnitude and distribution of contamination on the skin can be observed and displayed in real time. The VITAE system (Video Imaging Technique for Assessing Dermal Exposure), both in its original design (Fenske et al., 1986a
) and in the updated format (Fenske and Birnbaum, 1997), was the first system to employ this technique to image the whole body following occupational exposures. With this technique, fluorescent whitening agents (FWAs) that are used as optical brighteners in detergents were mixed with paints, fungicides and insecticides and co-applied under true occupational conditions. Some of the standard FWAs that were employed included Calcofluor, Uvitex OB and Tinopal CBS-X. Following occupational activities and removal of worker clothing, the contamination on the skin was readily observable under UVA light.
These types of studies have elucidated important issues for worker exposure and worker protection. In particular, the earliest studies demonstrated that pesticides and chlorophenols readily penetrate worker clothing (Fenske et al., 1986b, 1987; Fenske, 1988b; Fenske and Elkner, 1990; Methner and Fenske, 1994) and that gaps in the clothing around collars and cuffs act by a bellows effect to cause additional exposure to the forearm and torso areas (Fenske, 1988a). Quantitative video-fluorescence imaging of dermal surfaces after applications showed that there is no uniformity of deposition under clothing (Fenske, 1990; Archibald et al., 1994, 1995) for individual workers. Because of these widely varying deposition patterns, exposure estimates based on a single dermal patch per anatomical region (placed underneath clothing) showed poor agreement with exposure estimates based on video-fluorescence imaging (Fenske, 1990).
The video-fluorescence imaging of individual workers showed that exposures can be tied to activity patterns. For instance, the use, non-use or improper use of protective clothing can be readily discerned in tracer levels after pesticide applications (Fenske and Elkner, 1990; Methner and Fenske, 1994). Several authors were able to obtain quantitative measures of dermal exposure and dose and then show good agreement of these with the levels of pesticide metabolites excreted in urine following the application (Fenske et al., 1987; Fenske, 1988b; Archibald et al., 1994).
For the demonstration of occupational exposure, video-fluorescence imaging is a powerful teaching tool. The immediate feedback to workers of the visual image that combines both amount and distribution of chemical residue on the body has been cited as the main advantage of this technique, independent of the quantitative aspects. However, fluorescence can be observed with very high sensitivities relative to other spectroscopic tools. If working in a system where the background fluorescence signal is minimal, then a true signal can be amplified significantly and allow low levels of an analyte to be measured. The potential of high signal-to-noise with fluorescence is important for transferring the concepts of video-fluorescence imaging from the occupational exposure arena to the residential exposure arena.
The video-fluorescence imaging technique has been used in the residential environment. Fenske and Black had children play on Uvitex OB-treated lawns for 11 min and then imaged the fluorescence on exposed face, neck, hands, arms and legs (Black, personal communication). Using a previously determined ratio of dislodgeability for Uvitex OB versus chlorpyrifos (Black and Fenske, 1996), they were able estimate that a child playing on treated turf will pick up 80–140 ng chlorpyrifos/cm2 bare skin.
There continues to be a need to understand the mechanisms of exposure, as well as the magnitude, frequency and duration, for young children playing in their commonly encountered environments (Cohen Hubal et al., 2000a). The dermal transfer from contaminated surfaces continues to be an area that is not well characterized (Cohen Hubal et al., 2000b). Several recent studies have measured transfer of pesticides, or surrogates, from surfaces to hands under a limited number of conditions that are representative of the residential environment soon after a pesticide application. Camann et al. (1996) and Rodes et al. (2001) developed transfer efficiency test methods in which the hand loading was assessed following removal of the contamination by isopropanol wash or wipe. As such, loadings following each contact in a sequence could not be assessed. Video-fluorescence imaging techniques were used successfully by Brouwer et al. (1999), using a finely milled dust of the fluorescent compound Tinopal as a surrogate for soil or dust. Though transfer of these particles may not be representative of residue transfers following a residential application of pesticides, Brouwer and co-workers clearly demonstrated that fluorescent imaging techniques can be used to elucidate transfer efficiencies from multiple contacts.
The above-mentioned studies present important data for exposure resulting from surface contact. However, those studies have limitations, either through the use of particles to simulate the dried liquid film of a pesticide application or through the use of techniques where individual transfers of sequential contacts could not be assessed. Video-fluorescence imaging has the potential to provide valuable insights into the mechanics of sequential dermal transfers that are not readily apparent with bulk measurements from hand rinses. This is important, as studies have shown that pesticides are rapidly absorbed into the skin and thus not completely removed with rinses, washes or wipes (Lu and Fenske, 1999; Brouwer et al., 2000; Campbell et al., 2000).
In order to pursue studies of contact transfers from low surface loadings, a new imaging system was designed around commercially available hardware and software and specific to the properties of a completely safe fluorescent compound whose properties would allow low level quantification without significant interference from natural skin fluorescence. This tracer was applied via aqueous spray to carpets and laminate surfaces to simulate a residential application of a pesticide. After the residues dried, the transfers to hands with sequential contacts were measured (Cohen Hubal et al., 2003). This manuscript describes the imaging system and quantitation methods developed for these dermal transfer measurements, as well as analyses to characterize the precision and accuracy of the system and method.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
Tracer properties
Riboflavin (vitamin B2) was selected as the tracer. It is a water-soluble essential vitamin that functions as an enzyme cofactor. Any excess amount ingested, beyond the body's requirement for skin and muscle tone and metabolism of carbohydrate, fats and proteins, is excreted in urine. The recommended daily allowance is 1.6–2.6 mg/day for adults. Its main sources in the diet include leafy green vegetables, milk, cheese, nuts, citrus fruits and tomatoes. Riboflavin has been formulated in specialty cosmetics for skin, but is formulated with oils to enhance its dermal penetration. Dermal irritation as measured in the Draize test has not been reported. Riboflavin is considered completely non-toxic and, for this reason, is an extremely safe tracer to use in experiments with human subjects.
Riboflavin (C17H20N4O6, mol. wt 376) is both water- and acetone-soluble; the log KOW is –1.46 (Nahum and Horvath, 1980). It has three fused heterocyclic nitrogen rings with a polar chain containing hydroxyl functional groups. With its high water solubility (150 mg/l), riboflavin solutions approximating pesticide solutions can be readily prepared as an aqueous solution for spray application onto surfaces.
Riboflavin possesses spectroscopic properties that make it an ideal candidate as a tracer. Investigation of the excitation emission matrix for riboflavin shows that it absorbs in the 250–500 nm region of the electromagnetic spectrum, with strong absorption in the region of blue visible light (440–480 nm). Its peak fluorescence emission occurs in the green region of the spectrum (505–560 nm), but it has sufficient emission in the red/orange region (600 nm) to allow its detection well outside the interference of natural skin fluorescence (skin fluoresces in the blue/green region of the spectrum).
A comparison of the fluorescence strength of several candidate tracers is shown in Fig. 1. With UVA excitation at 380 nm, Uvitex OB has fluorescence output in the blue region (440 nm) that is many times greater than that of either riboflavin or fluorescein. However, in the red/orange region of the emission spectrum (600 nm) the output for the three compounds is similar. With blue light excitation at 440 nm, Uvitex OB is not detectable; fluorescein has output that is about double that of riboflavin, but the fluorescence of both compounds is of the same order of magnitude. Fluorescein is not soluble in water nor is it as safe as riboflavin for use with human subjects.
|
As with any highly fluorescent compound, riboflavin will undergo photodegradation. In aqueous solution under fluorescent light and incident sunlight (e.g. the laboratory setting) the degradation rate is 10%/h. For this reason, and to assure removal of interference from extraneous reflection from secondary sources, laboratory work here was performed in either a low light room or in a room with yellow filtered lights.
Fluorescence system and measurement technique
The system for image generation and collection consisted of an optically filtered excitation source and a charge-coupled device (CCD) video camera (black and white Canon 20iR) with an image intensifier (VARO Inc.) on the image collection plane. The fluorescence was excited by 96 W fluorescent lamps positioned above the image plane. These lamps were enclosed in a fan-cooled unit where the openings directed down towards the image plane were optically filtered with Schott BG28 elements. These filters passed blue excitation energy and significantly blocked the lamp energy in the 600 nm region of the spectrum, which could cause reflectance interference in the analytical signal. The excitation source and the emission collection camera lens were located in the same plane of the system, 40 cm above the image plane (where the hand was placed). Figure 2 shows the lay-out of the optical system.
|
A video camera unit was made up of several components. The exterior lens system consisted of an interference filter (Coherent) with wavelength centered at 600 nm and a band width of 10 ± 5 nm. The lens focused the image onto the front of the image intensifier, which multiplied the signal 10 000-fold and projected this image onto the back plane of the intensifier. Transfer optics were used to refocus this image onto the detector elements of the camera, which had a 4 x 6 mm CCD system. Figure 3 shows the filter transmittance and the fluorescence output for the excitation filters on the source (left) and the filter transmittance on the camera (right).
|
The hand that was being imaged was placed palm side up in an individually crafted hand mold made of potter's clay that was painted black. The clay mold was kept in a moist environment when not in use to maintain its precision fit to the hand.
In contrast to some of the fluorescent whitening agents, the maximum fluorescence emission of riboflavin is achieved when the individual molecules are well separated so that quenching does not occur. For riboflavin that is adsorbed into the epidermal layer, there is sufficient moisture and other blocking material to reduce quenching energy transfer in the skin. For measurement of riboflavin that is on the skin surface, we found that addition of a thin film of water to the skin surface assured uniformity of measurement. To achieve this, the hand containing riboflavin residues was held 10 cm from the outlet of a cool mist vaporizer for 5 s just prior to acquiring a measurement of optical fluorescence. This resulted in a thin ‘sheen’ of water on the surface that did not distort the position of the riboflavin on the hand. The difference in fluorescence intensity between a moist hand and a dry hand with the same riboflavin loading was a factor of 2. This difference dropped to 20% when there were very low loadings on hands, as the skin effectively served the purpose of the added water.
Data acquisition system
The images obtained by the camera were collected, averaged and stored by the Matrox image acquisition system (Matrox Electronic Systems Ltd, Canada). The Matrox system consists of a frame grabber data collection board (METEOR2), operated under Microsoft Windows, that allows the images to be collected from a video camera and averaged during a specified collection period. The data collection here averaged 32 images at a time. With this averaging, the speckle found in the raw intensifier image was removed so that the signal-to-noise ratio was improved significantly.
The Matrox Inspector Software (version 3.0), a hardware-independent application package, was used to manipulate and analyze the grabbed images. Manipulations included point-to-point mathematical operations (e.g. image-to-image subtraction), as well as geometric outlining. The software was also used to generate the frequency histogram of signal strength (i.e. gray level) versus number of pixels. Scripts written in Basic automated the data acquisition.
After image collection, the collected image voltage was overlaid on a background voltage from the camera. This pedestal (or background voltage) provided sufficient enhancement to the signal to allow observation of the image on the connected TV screen. This pedestal was then subtracted before applying the mathematical steps to linearize the image. Linearization was needed to correct for curvature of the image introduced by the image intensifier and other effects due to non-uniform pixel-to-pixel sensitivity in the camera system. To do so, the collected image was ratioed against the image of a ‘white surface’ which had been stored away at the time of alignment of the detection system to the image plane. The ratioed image was then linear with respect to the detectors and its zero energy was registered as a true zero. From here, the frequency histogram was generated.
The image manipulations are represented schematically in Fig. 4. Further data reduction, specifically application of a calibration function to obtain mass with simultaneous correction to the raw histogram non-linearities caused by the optical absorption properties (thickness) of the collected residue itself, were completed off-line in Excel. This conversion and adjustment is described further in the discussion of calibration below.
|
A measurement of the background fluorescence of the hand prior to contacting a riboflavin-treated surface was collected with every test. Pixel-by-pixel subtraction of the uncontaminated hand image from the riboflavin-contaminated hand image yielded the amount of fluorescence due only to riboflavin. The results of this process are shown in Fig. 5. This figure includes the images of a hand prior to and after contact with a riboflavin-treated surface. Both images have undergone the pixel-to-pixel linearization process. The resulting subtracted image is included to the right of the two original images. The histogram that results from the conversion process is shown to the right of the subtracted image.
|
Calibration method
Theoretical considerations for calibration
Ideally the signal strength of each pixel is directly related to the mass present in that pixel area. In such a case, a signal strength of 200 in one pixel would be equivalent in mass to 200 pixels each having a signal strength of one. In fact, this is true only if the layer of riboflavin is thin and uniformly deposited. In practice, a uniformly deposited thin layer does not occur with either contamination on a hand introduced by contact with a treated surface or with addition of known amounts for a calibration curve.
The signal strength observed is related to the thickness of the contamination. Surface layering can cause the same amount of material to yield a different signal depending on where in the layering the source of the signal lies, as lower layers of material have less excitation energy available and have more scattering of the fluorescent light than do the top-most layer. The relationship of the mass to the signal strength can, then, be quite complex, because there are multiple mixtures and wavelengths of light involved in the total process. Calibration must be approached with the thickness of the deposit taken into account.
The incident light is reduced as it penetrates through riboflavin layers. The total fluorescence detected is the sum of fluorescence emanating directly from the top-most layer, from subsequently lower layers and fluorescence reflected off the surface of the hand. Whether direct or reflected, the fluorescence generated in lower layers is more likely to be scattered or reabsorbed than fluorescence generated in upper layers. The relationship of mass to signal strength depends on the depth of generation in the sample according to the following equation:
 |
With increasing thickness, the signal strength increases monotonically, but at increasingly slower rate since S becomes larger with added depth. Therefore, each pixel signal strength level (1–256) must be independently scaled for its relationship to mass loading. In order to deal with <256 weighting factors and, because of the monotonically increasing nature of the relationship between S and fluorescence, it is possible to reduce the 256 different pixel strength levels to substantially fewer signal strength bins, each of which covers multiple signal strength levels.
Empirical approach to calibration using weighting factors to account for layering and to convert pixel signal strength level to mass
The goal of this process is to be able to determine the total mass on the hand based on the observed signal strength value of each pixel. The conceptual model is that the total mass on the hand is equal to the sum of the mass at each pixel, and the mass at each pixel is a function of the observed signal strength for that pixel. The function relating the signal strength value to the mass is expected to be non-linear due to layering effects and is somewhat dependent on the individual subjects due to differences in the reflectiveness of different people's skin. Below, a calibration of the system is described through a two-stage process based on 15 calibration points.
The calibration produces a function w(k) of the signal strength response for a pixel to the mass at the pixel area. The mass at the pixel equals w(k) where k is the integer value for the pixel signal strength value. In this way, the total mass, m, is given by:
 |
 |
For the work described here, 15 known amounts of riboflavin (2–61 µg, from an aqueous 100 µg/ml solution) were deposited on a subject's hand for the calibration curve. Each deposit was made and then spread out to cover 1–25 cm2 on the hand. In some cases, the same deposit amount was spread out over three or five different areas. The deposit was allowed to dry, then moistened as per the method described above and then imaged with the system. Each deposit of riboflavin was made so that the border of the riboflavin deposit was distinct and its intensity was as flat as possible. This process yielded 15 nominally top-hat shaped histograms of number of pixels versus signal strength level.
Let mi be the total mass from the ith sample. Then,
 |
The first step in the calibration process exploited this continuity; within small ranges of k the function w(k) should not change much. This allows the data to be binned and to effectively work with 10 different signal strength values rather than all 256. The data are now represented as:
 |
The second step of the process was to extend the 10 weighting factors to a continuous function. This was done by fitting a curve to the 10 points where each of the 10 weighting factors above was associated with the center of the bin (see Fig. 6). With this calibration curve, an unknown mass can be found either with the equation:
 |
 |
|
Quantification of dermal and surface contact areas
In order to calculate the transfer efficiency from surfaces following contact, both the surface contact area and the dermal contact area must be known. For simple presses to a surface, the surface contact area and the dermal contact area are identical. However, for a smudge contact of the hand with a surface, where the hand moves across the surface, the surface contact area is greater than the dermal contact area. The video image of the contaminated hand, minus the background image of the uncontaminated hand, can, in theory, be used to determine the dermal contact area, as the area of contamination on the hand is equivalent to the dermal contact area for simple press contacts. Though straightforward in theory, this version of the Matrox software was not well-adapted for this particular application. In addition, while dermal contact areas can, in theory, be measured using the video-fluorescence imaging system, measurement of the surface contact area cannot be done with an imaging system, since the surface cannot be imaged to show the area that was contacted.
To determine the dermal and surface contact areas, a different approach was used here. For each subject, a series of three handpresses was obtained under controlled conditions. The hand surface was first coated with a thin layer of yellow Tempra paint and then the subject was instructed to press or smudge the hand onto black paper using either a calibrated light (0.1 p.s.i.) or heavy press (1 p.s.i.). The resulting handpresses were imaged using the video-fluorescence imaging system and the area of contamination was integrated using the Matrox software and mouse-drawn segmentation of the individual areas. The press and smudge areas were then used whenever a dermal contact or surface contact area was required in the calculations of transfer efficiency, recognizing that these areas may be very good approximations of an initial press, but may be less accurate for subsequent contacts of a hand with a surface. This approach to dermal and surface contact area measurement was used as a first approximation and may be reasonable within the range of error of the technique.
Each subject had a distinctly different pattern to the areas of contact for the simple presses. An example of the series of paint press images, and the integrated areas, for one subject are shown in Figure 7. The dermal contact area for a light press for 5 adult women averaged 55 cm2, and the dermal contact area for a heavy press averaged 86 cm2, indicating and 60% increase in dermal contact area between a light and heavy press. This average dermal contact area of 55 cm2 areas well with literature data, that being 154 cm2 for projected area of a woman's hand, and a dermal contact area of 35% of projected dermal area (US EPA, 1996; Rodes et al, 2001).
|
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
Stability of the system
A section of cardboard was cut to the approximate size and shape of a typical hand used in the dermal transfer experiments. Spatially, cardboard has fairly uniform fluorescence and its intensity is similar to the levels observed in the riboflavin transfer experiments. The cardboard hand was placed in the optical plane that was used for the hand measurements and the video-fluorescence image of this cardboard hand was collected at multiple times over a day and over a week to assess temporal stability of the overall system. Once the image was collected, the area of the hand was outlined and the total intensity of the included area was determined as:
 |
The intensity of the hand was first determined by drawing the included area five times from a single video image. The intensity of this one image was 757 028 ± 2651 pixel intensity units; this indicates that the error on generating a histogram is minimal, of the order of 0.4%. The intensity of the hand measured over 5 h from six different images was 730 168 ± 29 082; this variability of 4% represents the type of variability that might be expected during the timespan of one set of hand contact experiments. The intensity of the imaged hand over 7 days from seven different images was 737 822 ± 36 696; this weekly variability of 5% is only slightly more than the daily variability of the system. When the hand was imaged from several different locations within the optical plane, the variation in intensity was 15%. The locations chosen for this latter assessment were more widely spaced than any used in the conduct of the actual hand contact experiments and thus represent an unrealistically high value of measurement variability.
The moistening and imaging of the hand appears to be a stable and reproducible procedure as well. To test this, a subject's hand was pressed onto a riboflavin-treated surface and then imaged four times. Each image was collected after moistening the hand for 5 s in the mist of a cool mist vaporizer. The 5% variation in total intensity for the four measurements suggests that moistening does not introduce any additional variability to the measurements.
Precision, accuracy and detection limit of the system and the calibration method
A nine point calibration curve was generated on the hands of four different subjects. The riboflavin (0.5–15 µg) was spread over 5 cm2 to give loadings of 0.1–3.0 µg/cm2. There were three calibration curves generated on the first subject, each on a different day. A single calibration curve was generated on the other three subjects, each on different days. The weighting factors (ng/pixel) of the calibration curve were determined using a bin number of 8. The recalculated masses for these calibration histograms are given in Table 1, with intra-subject and inter-subject assessment of variability. As shown there, the relative mean square error (used to incorporate both bias and variability) for riboflavin averaged 21 and 13%, respectively, for repeated analyses on one hand and repeated analyses on different hands. The absolute mean square error averaged 0.7 µg for repeated measures on one subject and 0.5 µg for measures on unrelated subjects. We are somewhat puzzled that the repeated measures on the same individual had greater variability than on three unrelated subjects and have assumed that this represents the maximum possible variability of the system.
|
The quantification limit of this system for riboflavin is
0.1 µg/cm2, with a detection limit of 0.02 µg/cm2. Typical levels that have been measured with this system have been in the range 0.1–2 µg/cm2 for experiments where moist, sticky and dry hands were pressed onto riboflavin-treated laminate and carpet surfaces with surface loadings of 2 and 10 µg/cm2 (Cohen Hubal et al., 2004). Subsequent to these initial measurements, we opened the camera lens and lowered the detection limit of the system to 0.01 µg/cm2, and this system configuration was used in the experiments described below.
Accuracy of blind measurements on hands
As a test of the accuracy of the overall system, three different subjects contacted either a carpeted or laminate surface with a riboflavin surface loading of 0.2 µg/cm2. Subjects pressed their hand onto the surface either once, twice, four times or eight times before having the hand imaged with the video-fluorescence imaging system. After imaging, the hand was rinsed in a bag of water; the water was subsequently analyzed by conventional spectrofluorimetry for riboflavin content. After rinsing, the hand was imaged again and the loss of riboflavin from the hand was calculated as the difference between the amounts in the pre-wash and post-wash fluorescence images. Calibration data as described above were collected for the three subjects and the calibration weighting factors were averaged across subjects to give an average weighing factor for each bin.
The pair-wise comparison of the amount of riboflavin measured by video-fluorescence imaging with the amount measured in the wash solution is shown in Fig. 8. Several data points were excluded from this analysis because they were made at the detection limit of the system (0.01 µg/cm2). Using a log-linear mixed model, with adjustment for the repeated measures aspect of the data, the model yielded a statistically significant (P < 0.00010) estimate of the slope between the ln(difference in pre-wash and post-wash image amounts) and ln(wash solution amount) of 0.86, with a standard error of 0.1162.
|
Effect of skin tone
Fenske and Birnbaum (1997)
found that with the VITAE system, using Uvitex OB as the fluorescent tracer and measurement at 440 nm, there is a measurable effect due to skin tone. Since skin tones may affect reflectance and thus affect the magnitude of the weighting factor for a bin, we used weighting factors (ng/pixel) from individual calibration curves to assess the effect of skin tone. The weighting factors from the calibration curve generated on one person with a dark skin tone were compared with the weighting factors from three people with relatively light skin tones. We used a constrained linear model with measurement error to estimate the full distribution of the pixel intensity weight vector for each individual across all bin levels. The model was fitted via Markov chain Monte Carlo (MCMC) simulation to account for non-normality in the parameter distributions due to the linear model constraints. This approach allowed us to more accurately estimate the 2.5th and 97.5th percentiles, or an approximate 95% confidence interval, of the distribution of weights for each individual across all bin levels. Within each bin, pairwise comparisons among confidence intervals were made to identify statistically significant differences. Within the uncertainty and variability in the data, we did not find evidence of a statistical difference between bin values for skin with light and dark tones. This does not indicate that person-to-person differences do not exist, only that the variation in the data is too great to observe a statistically significant effect due to skin tone. These results are not entirely surprising, given the difference in monitored wavelength used here versus that used by Fenske and co-workers. Pale skin has approximately equal reflectance in the blue, green and red regions of the spectrum, whereas brown skin has much less reflectance in the green region, but nearly equal amounts in the red and blue. Since our measurements are being made in the red/orange region of the spectrum, the system is probably seeing approximately the same degree of reflectance from all skin tones. In contrast, the VITAE system with measurement in the green spectral region is making measurements in a region affected by skin tone.
System application
An example of the system performance is shown in Fig. 9. The first column of the figure includes six hand images, each corrected for background skin fluorescence, in a sequence of repetitive contacts with a laminate surface with a riboflavin surface loading of 10 µg/cm2. The hand was smudged on the contaminated surface five times and then the sixth smudge was made onto an untreated laminate surface. As shown in the sequence, there is a gradual increase in riboflavin loading on the hand in images 1–5 and then a slight loss with the final smudge to a clean surface.
|
The corresponding histograms of pixel count versus pixel signal strength for these six hand images are included in the second column of Fig. 9. In this series one can begin to identify the trend of increasing signal at higher signal strength with each successive press. However, this trend is far more evident when the data are plotted as the total signal strength at each gray level, as shown in the third column of Fig. 9. It is these final histograms, then, that are adjusted by the weighting factor to give the mass on the hand. The quantified masses of riboflavin on the hand in Fig. 9 are 19, 32, 35, 35 and 47 µg (sequential contact with the riboflavin-treated surface) and 43 µg (after a smudge to a clean surface) or, expressed in terms of average mass loading in the dermal contact area, 0.24, 0.40, 0.45, 0.45 and 0.59 µg/cm2 after presses to the riboflavin surface and 0.54 µg/cm2 after the smudge to the clean surface.
|
CONCLUSIONS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
The system developed and tested here was constructed with off-the-shelf components and commercial software. This is essentially a ‘plug-and-play’ data acquisition and reduction system, which is operated in a modular manner to provide diverse operations for data analysis in a spatial and/or 3-dimensional context. The system is compact; the image collection unit is about the same size as the image display monitor and the Matrox Inspector computer unit.
The total method is virtually risk free. Riboflavin is non-toxic; in addition, because it is water soluble, residues can be removed after an experiment via washing with soap and warm water. With riboflavin, visible light is used as the energy source, thus avoiding the UVA light sources that are used to excite the fluorescent whitening agents of other methods. The methodology is also flexible. In contrast to the fluorescent whitening agent tracers, riboflavin can be readily removed from the skin after a test, so that multiple tests can be carried out in succession. With the fluorescent whitening agents, natural dermal abrasion over 5–7 days must be used to rid the skin of fluorescent residues before another test can be undertaken. The addition of multivariate analysis allows the effective treatment of layering and quenching during calibration.
One disadvantage of the system is the lack of photostability of riboflavin, which necessitates the use of low light conditions. As with any fluorescence-based system, this system is sensitive to ambient light. Even the video display monitor is angled away from the image collection camera to avoid spurious reflectance. The use of riboflavin also requires moistening of the hand before imaging to lessen the impact of molecular layering and quenching and to maximize signal strength. And, finally, this system was developed with the treatment of the hand as essentially a flat plane. While this is acceptable for this simple application, a much more elaborate lighting system, of the order of the FIVES dodecahedral illumination system (Roff, 1994), must be considered for whole body analyses or ones requiring analysis of a 3-dimensional object.
The system is sufficiently sensitive (detection limit 0.02 µg/cm2) to allow analysis of residues transferred from surfaces that have a fluorescent tracer loading equivalent to pesticide loadings in homes following applications. The average relative error is <20% above the quantification limit of 0.1 µg/cm2. The utility of riboflavin as a tracer surrogate for pesticides needs to be compared with some of the other more frequently used fluorescent tracers. In addition, this system needs to be tested under transfer tests with ever lower surface contamination levels, to more accurately reflect the levels of pesticides that are found indoors one or more days after an application.
|
ACKNOWLEDGEMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
The authors acknowledge the laboratory assistance of K. Andrews, J. Sowry and M. McCauley of Battelle Memorial Institute and the assistance of the recruited subjects. This work was funded by the US EPA under contract no. 68-D-99-011 to Battelle. It has been subjected to agency review and approved for publication. Work with human subjects was approved by Institutional Review Boards at Battelle and EPA.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSCONCLUSIONSACKNOWLEDGEMENTSREFERENCES |
|---|
Archibald BA, Solomon KR, Stephenson GR. (1994) Estimating pirimicarb exposure to greenhouse workers using video imaging. Arch Environ Contam Toxicol; 27: 126–9.[Web of Science][Medline]
Archibald BA, Solomon KR, Stephenson GR. (1995) Estimation of pesticide exposure to greenhouse applicators using video imaging and other assessment techniques. Am Ind Hyg Assoc J; 56: 226–35.[Web of Science][Medline]
Black KG, Fenske RA. (1996) Dislodgeability of chlorpyrifos and fluorescent tracer residues on turf: comparison of wipe and foliar wash sampling techniques. Arch Environ Contam Toxicol; 31: 563–70.[CrossRef][Web of Science][Medline]
Brouwer DH, Kroese R, van Hemmen JJ. (1999) Transfer of contaminants from surfaces to hands: experimental assessment of linearity of the exposure process, adherence to the skin, and area exposed during fixed pressure and repeated contact with surfaces contaminated with a powder. Appl Occup Environ Hyg; 14: 231–9.[CrossRef][Medline]
Brouwer DH, Boeniger MF, van Hemmen J. (2000) Hand wash and manual skin wipes. Ann Occup Hyg; 44: 501–10.
Camann DE, Majumdar TP, Harding HJ, Ellenson WD, Lewis RG. (1996) Transfer efficiency of pesticides from carpet to saliva-moistened hands. In Proceedings of the International Specialty Conference on Measurement of Toxic and Related Air Pollutants; Air and Waste Management Association, Pittsburgh, PA, pp. 532–40.
Campbell JL, Smith MA, Eiteman MA, Williams PL, Boeniger MF. (2000) Comparison of solvents for removing pesticides from skin using an in vitro porcine model. Am Ind Hyg Assoc J; 61: 82–8.
Cohen Hubal EA, Sheldon LS, Zufall MJ, Burke JM, Thomas KW. (2000a) The challenge of assessing children's residential exposure to pesticides. J Expos Anal Environ Epidemiol; 10: 475–86.
Cohen Hubal EA, Sheldon LS, Burke JM, McCurdy TR, Berry MR, Rigas MI, Zartarian VG, Freeman NCG. (2000b) Children's exposure assessment: a review of factors influencing children's exposure and data available to characterize and assess that exposure. Environ Health Perspect; 108: 475–86.
Cohen Hubal EA, Suggs JC, Nishioka MG, Ivancic WA. (2004) Characterizing pesticide residue transfer efficiencies using a fluorescent imaging technique. J Expos Anal Environ Epidemiol; in press.
Fenske RA. (1988a) Comparative assessment of protective clothing performance by measurement of dermal exposure during pesticide applications. Appl Ind Hyg; 3: 207–13.
Fenske RA. (1988b) Correlation of fluorescent tracer measurements of dermal exposure and urinary metabolite excretion during occupational exposure to malathion. Am Ind Hyg Assoc J; 49: 438–44.[Web of Science][Medline]
Fenske RA. (1990) Nonuniform dermal deposition patterns during occupational exposure to pesticides. Arch Environ Contam Toxicol; 19: 332–7.[CrossRef][Web of Science][Medline]
Fenske RA, Birnbaum SG. (1997) Second generation video imaging technique for assessing dermal exposure (VITAE System). Am Ind Hyg Assoc J; 58: 636–45.[Web of Science][Medline]
Fenske RA, Elkner KP. (1990) Multi-route exposure assessment and biological monitoring of urban pesticide applicators during structural control treatments with chlorpyrifos. Toxicol Ind Health; 6: 349–71.[Web of Science][Medline]
Fenske RA, Leffingwell JT, Spear RC. (1986a) A video imaging technique for assessing dermal exposure I. Instrument design and testing. Am Ind Hyg Assoc J; 47: 764–70.[Web of Science][Medline]
Fenske RA, Wong SM, Leffingwell JT, Spear RT. (1986b) A video imaging technique for assessing dermal exposure II. Fluorescent tracer testing. Am Ind Hyg Assoc J; 47: 771–5.[Web of Science][Medline]
Fenske RA, Horstman SW, Bentley RK. (1987) Assessment of dermal exposure to chlorphenols in timber mills. Appl Ind Hyg; 2: 143–7.
Lu C, Fenske RA. (1999) Dermal transfer of chlorpyrifos residues from residential surfaces: comparison of hand press, hand drag, wipe, and polyurethane foam roller measurements after broadcast and aerosol pesticide applications. Environ Health Perspect; 107: 463–7.
Methner MM, Fenske RA. (1994) Pesticide exposure during greenhouse applications. Part II. Chemical permeation through protective clothing in contact with treated foliage. Appl Occup Environ Hyg; 9: 567–74.
Nahum A, Horvath C. (1980) Evaluation of octanol-water partition coefficients by using high-performance liquid chromatography. J Chromatogr; 192: 315–22.
Rodes CE, Newsome JR, vander Pool RW, Antley JT, Lewis RG. (2001) Experimental methodologies and preliminary transfer factor data for estimation of dermal exposure to particles. J Expos Anal Environ Epidemiol; 11: 123–39.[CrossRef][Web of Science][Medline]
Roff MW. (1994) A novel lighting system for the measurement of dermal exposure using a fluorescent dye and an image processor. Ann Occup Hyg; 38: 903–19.
US Environmental Protection Agency. (1996) Exposure factors handbook, US EPA/600/P-95/002Ba. US EPA, Washington, DC.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||