In the Laboratory

edited by
Cost-Effective Teacher Harold H. Harris
University of MissouriSt. Louis
St. Louis, MO 63121

Microscale Colorimetric Analysis

Using a Desktop Scanner and Automated Digital Image Analysis
Douglas J. Soldat,* Phillip Barak, and Brian J. Lepore
Department of Soil Science, University of WisconsinMadison, Madison, WI 53706; *djsoldat@wisc.edu

In recent decades, several standard colorimetric reactions channel value of 159 is

for chemical analysis have been miniaturized to microwells on
microplates, including methods useful for environmental mea- 159
A = log = 0. 205
surements. Advantages of method miniaturization are a reduc- 255
tion in reagents required, improved safety, reduced waste stream,
and increased sample throughput. However, the widespread use In this article, we demonstrate that digital image analysis
of microscale techniques employing microplates in classroom of a scanned microplate image can substitute for a spectropho-
settings is likely limited by the cost of microplate readers, cur- tometer for several common quantitative microscale procedures.
rently around $10,000 for a basic unit. This finding allows for cost effective and microscale quantifica-
Recently, Kohl et al. (1) showed that the principle of ab- tion of several compounds to be demonstrated in the laboratory.
sorbance could be easily demonstrated using colored solutions Additionally, popular teaching and learning activities such as
and digital image analysis. However, the authors speculated that water-quality monitoring can now be performed accurately and
a camera and image analysis could not generally substitute for a inexpensively using digital image analysis.
spectrophotometer except under special conditions. Although
spectrophotometers read peaks of specific wavelengths, absor- Materials and Methods
bance spectra tend to be relatively broad, and measurements
at specific wavelengths are highly autocorrelated with those of To demonstrate that digital image analysis can substitute
nearby neighbors, which implies that broadband intensity data for a spectrophotometer for quantitative colorimetric analysis,
of red, green, and blue channels may be adequate for digital we prepared standard curves for previously published colorimet-
colorimetric quantification. According to the BeerLambert law, ric methods for microscale determination of ammonium (3),
the absorption of light by a compound is equal to the product bromide (4), nitrate (5), and phosphate (6, 7) ions. Standard
of the molar absorptivity, the path length of the sample, and the microplates, each with an 8 12 array of flat-bottomed 350L
concentration of the compound in the solution. Absorbance can microwells, were prepared with appropriate standards and color
be calculated as development solutions using an electronic 12-channel pipet.
Using the transparency scanner feature (Figure 1) of a com-
I mercially available desktop scanner (Epson Perfection 4990
A = log
I0
where A is the absorbance, I is the intensity of the light after
passing through the solutions, and I0 is the intensity of the light
before passing through the solutions.
Although the ultimate definition of color and color space
resides with the CIE (International Commission on Illumina-
tion) colorimetric system based on CIE XYZ tristimulus values,
the coding of digital color images in digital photography and
scanning more often uses a linear RGB system that is more
familiar to the modern, computer-savvy student and instructor
(2). The linear RGB color model uses intensity data from three
channels (red, green, and blue) to form a single color for each
pixel. Numerically, the color is represented by three numbers. In
a 24-bit color image, there are 28, or 256, possible values (0255)
for each of the red, green, and blue channels. When white light
is transmitted through a colored sample, absorbance of an im-
age can be calculated using red, green, and blue channel data by
taking the log of the maximum intensity (255) divided by the
actual red, green, or blue value from the image. For example, Figure 1. Scanner operating in transparency mode with white light
the absorbance in the red channel of an 8-bit image with a red from the lid passing through the 96-well microplates to the light
sensors below. The template with cut-outs positions the microplates
Current address: Biological Systems Engineering, University of consistently and blocks stray light. During the scanning operation, the
WisconsinMadison, Madison, WI 53706. scanner lid is closed and parallel to the scannerbed.

Division of Chemical Education www.JCE.DivCHED.org Vol. 86 No. 5 May 2009 Journal of Chemical Education 617
 In the Laboratory
Figure 2. Image produced by a transparency scan of a microplate
with ammonium standards (020 mg L1 N in NH4+) developed
according to Sims et al. (3). Parallax due to the sensor location on
the vertical axis is apparent in the left- and rightmost columns and
suggests the use of ellipses rather than spheres to describe regions of
interest within each microwell. (Shown in color on p 533.)
Photo, Long Beach, CA, $449), a 200 dpi, 48-bit (216 values per
primary color) digital image of the 96-microwell plates (available
from Fisher Scientific) was obtained. We built a simple opaque
polycarbonate template that covers the entire scanner bed with
a cut-out to hold microplates in the same location (Figure 1) so
that the scanned images could be cropped automatically to the
size of the microplate with the software provided by the scanner
manufacturer. The scanned image was saved in TIFF format,
a lossless format. An example of a scanned image is shown in
Figure 2.
The scanned microplate image was opened with public
domain image analysis software, ImageJ version 1.40 (8). This
software splits all pixels within an image into their red, green,
and blue components. To automate the image analysis process,
we developed a custom macro.1 This macro initially requires as
input the x and y coordinates (in pixels) of the top left micro
well, the number of rows and columns of microwells on the
microplate, the number of pixels between them, and the size and
eccentricity of the elliptical region of interest for each microwell.
Use of a template to hold microplates in a repeatable location
on the scanner ensures that these values do not need to be re
adjusted for subsequent scans. An elliptical region of interest was
chosen because the three-dimensional objects viewed from the
central axis of the scanner created a slight overlap of the image of
walls and well bottoms. The magnitude of the overlap increased
with distance from the central axis of the scanner bed (Figure2).
When conducting image analysis on the scanned images, one
could either analyze pixels in a smaller circle avoiding the image
of the walls or use an ellipse as the region of interest instead of a
complete circle; we chose the latter option because it allowed a
greater number of pixels to be analyzed per well without chang-
ing the scanner resolution.
The macro was designed to export the mean, median,
and mode values of red, green, and blue color channels from
a 424-pixel ellipsoid at the center of each microplate well into
618 Journal of Chemical Education Vol. 86 No. 5 May 2009 www.JCE.DivCHED.org Division of Chemical Education
80
red
70
Relative Pixel Count
60
50
40
30
green blue
20
10
0
46000 48000 58000 60000
Intensity of Pixel
Figure 3. Results of image analysis by ImageJ of an elliptical selection
of a microwell on a microplate determining nitrate by Szechrome
NAS. The red and green channels were more tightly grouped than
the blue channel, and the red and green channels were also more
sensitive to changes in nitrate concentration than the blue channel
(see Figure 4).
an ImageJ results spreadsheet. These values were then manually
transferred to a standard electronic spreadsheet. After examining
the images and their central tendencies for the effects of occa-
sional bubbles and dust particles, we selected median values for
red, green, and blue channels of the region of interest to calculate
absorbance values, as described in the introduction. The macro
is written as a text file, which permits simple editing of the fields
to further customization.
For comparison, we measured absorbance values for each
microplate with a standard microplate reader (Bio-Tek Power-
wave XS microplate scanner; Winooski, VT, $12,400) at the
appropriate wavelength for each method, within five minutes
of the digital scan.
Hazards
The use of a desktop scanner as a microplate reader presents
no innate hazards. A method for analyzing phosphate is pre-
sented in the online material with the following hazards: Ammo-
nium molybdate is an irritant to eyes, skin, and the respiratory
system. Antimony potassium tartrate is toxic and can be fatal if
swallowed. It is also an irritant to eyes, skin, and the respiratory
system. Monopotassium phosphate is an irritant to eyes, skin,
and the respiratory system. Sulfuric acid is corrosive and inhala-
tion can produce damaging effects to the upper respiratory tract.
Ingestion or contact with skin can cause severe burns. Contact
with eyes can cause severe burns or blindness.
Results
Image analysis of the colorimetric reactions in the mi-
croplates produced tightly clustered red, green, and blue values
for each microwell (Figure 3), which allowed for collection of
single central values of each to be transformed into absorbance.
The minimum requirement for a transparency scanner image
 In the Laboratory

A B
0.8
0.07
0.7
red channel red channel
green channel 0.06 green channel
0.6
blue channel blue channel
0.05

Absorbance
Absorbance

0.5

0.04
0.4

0.03
0.3

0.02
0.2

0.1 0.01

0.0 0.00
0 5 10 15 20 0.0 0.2 0.4 0.6 0.8 1.0

Ammonium Concentration as N / (mg L1) Phosphate Concentration as P / (mg L1)

C D
0.20 1.0

0.18 0.9
red channel
0.16 green channel 0.8
blue channel
0.14 0.7
Absorbance

Absorbance
0.12 0.6
red channel
0.10 0.5 green channel
0.08 0.4
blue channel

0.06 0.3

0.04 0.2

0.02 0.1

0.00 0.0
0 5 10 15 20 0.00 0.05 0.10 0.15 0.20 0.25

Nitrate Concentration as N / (mg L1) Bromide Concentration / (mg L 1

)
Figure 4. Calibration curves created using absorbance values from red, green, or blue channels in scanned images for the colorimetric reactions
of (A) ammonium by the phenate method, (B) phosphate by ascorbic acid, (C) nitrate with Szechrome NAS, and (D) bromide by phenol red.

as a substitute for traditional absorbance measurements is the
ability to produce calibration curves using at least one of either
the red, green, or blue channels for established colorimetric
reactions. For ammonium, bromide, nitrate, and phosphate
ions, this clearly occurs as seen in Figure 4. In some cases, such
as phosphate, all three primaries were correlated with analyte
concentration and the line with steepest slope should be chosen
as the most sensitive. In other cases, one or more of the primaries
are relatively unresponsive to the analyte concentration, but in
all cases at least one of the three primaries was correlated with
analyte concentration. The primary color that is most sensitive
for each analytical method generally follows the standard wave-
length chosen for analysis by spectrophotometry. For example,
the red primary (~630 nm) is close to the wavelength usually
used for ammonium by phenate, 667 nm, and green (~530
nm) is close to that used for bromide analysis by phenol red,
575 nm. Interestingly, the phosphate measurements, usually
read at a wavelength of 850 nm, that is, in the infrared range,
are found here to be quite usable with the red primary owing to
the broad absorbance peak of the colored complex. Somewhat
more difficult to explain to students is the inverse nature of the
color measured. For example, for ammonium ion analysis, the
yellow blank is the result of the addition of red and green pri-
maries. As the absorption of the red wavelengths increases with
increasing ammonium concentrations, the remaining mixture
appears greener.
For these colorimetric reactions, the use of a scanner oper-
ating in transparency mode may be favorably compared to the
dedicated microplate reader (Figure 5), sometimes matching
the exceptional dynamic range of the microplate reader but at
a much lower cost.
Conclusions
The results of this study indicate that digital image
analysis can replace a microplate reader for several microscale

Division of Chemical Education www.JCE.DivCHED.org Vol. 86 No. 5 May 2009 Journal of Chemical Education 619
 In the Laboratory

A B
bromide (green channel, 575 nm) phosphate (red channel, 850 nm)
0.8 ammonium (red channel, 667 nm) 0.14 nitrate (red channel, 570 nm)

0.7 0.12
Absorbance (Scanner)

Absorbance (Scanner)
0.6
0.10
y = 0.009 + 0.204x y = 0.002 + 0.226x
0.5 r 2 = 0.997 r 2 = 0.999
0.08
0.4
y = 0.028 + 0.151x
0.06
0.3
r 2 = 0.999

0.04 y = 0.003 + 0.234x

0.2 r 2 = 0.999

0.1 0.02

0.0 0.00
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7

Absorbance (Microplate Reader) Absorbance (Microplate Reader)

Figure 5. Relationship between absorbance of the red, green, or blue channel compared to absorbance as determined by microplate reader
at the appropriate wavelength for (A) ammonium by the phenate method and bromide by phenol red and (B) nitrate with Szechrome NAS
and phosphate by ascorbic acid.

quantitative colorimetric reactions. This finding has sev- Literature Cited

eral important implications. First, quantitative microscale
1. Kohl, S. K.; Landmark, J. D.; Stickle, D. F. J. Chem. Educ. 2006,
analysis can be introduced inexpensively into the teaching
83, 644646.
and learning laboratories, allowing students to gain experi-
2. Hearn, D.; Baker, M. P. Computer Graphics, 2nd ed.; Prentice Hall,
ence with microscale techniques and analytical methods,
Upper Saddle River, NJ, 1994.
and encouraging the use of multiple calibration curves on
3. Sims, G. K.; Ellsworth, T. R.; Mulvaney, R. L. Commun. Soil Sci.
a single microplate for improved quality control of results.
Plant Anal. 1995, 26, 303316.
Second, digital image analysis can add extra value to the
4. Lepore, B. J.; Barak, P. Soil Sci. Soc. Am. J., in press.
learning experience, especially during the conversions of
5. Rowe, R.; Todd, R.; Waide, J. Appl. Environ. Microbiol. 1977, 33,
red, green, and blue values into absorbance, which enables
675680.
students to understand the principles behind absorbance
6. DAngelo, E.; Crutchfield, J.; Vandiviere, M.; J. Environ. Qual.
and color science.
2001, 30, 22062209.
The microscale nature of these methods leads to a reduction
7. Avila-Segura, M.; Lyne, J. W.; Meyer, J. M.; Barak, P. Commun.
in reagents and waste, especially for alkaline or acid reagents, and
Soil Sci. Plant Anal. 2004, 35, 547557.
a reduction in cost when the reagents are relatively expensive.
8. ImageJ Home Page. http://rsb.info.nih.gov/ij/ (accessed Feb
In addition, this method has the potential to be made portable
2009).
for analyzing results in remote locations by operating the scan-
9. Frankenberger, W. T., Jr.; Tabatabai, M. A.; Adriano, D. C.;
ner on a generator or storage battery with an inverter, neither
Doner, H. E. Bromine, Chlorine, and Fluorine. In Methods of
of which is likely to work for a standard microplate reader. We
Soil Analysis: Part 3- Chemical Methods; Sparks, D. L., Ed.; ASA-
have chosen to present here results pertaining to several ions of
CSSA-SSSA: Madison, WI, 1996; pp 833867.
environmental interestammonium, nitrate, phosphate, and
10. Greenman, N. S.; Mulvaney, R. L.; Sims, G. K. Commun. Soil Sci.
bromideand expect these to bring water contamination and
Plant Anal. 1995, 26, 25192529.
tracer studies within the reach of student and citizen scientists
operating with a minimum of laboratory resources, but we have
Supporting JCE Online Material
found similar results for chloride (9) and urea (10), and expect
http://www.jce.divched.org/Journal/Issues/2009/May/abs617.html
the technique to be more widely applicable to many time-
honored colorimetric reactions at a microscale. Abstract and keywords
Full text (PDF)
Note Links to cited URLs and JCE article
1. The macro is available on the Internet at http://mywebspace. Supplement
wisc.edu/pwbarak/web/AnalyzeMicroplateStatistics.txt (accessed Feb Student and instructor notes for phosphate determination
2009). Macro file used to automate the image analysis process

620 Journal of Chemical Education Vol. 86 No. 5 May 2009 www.JCE.DivCHED.org Division of Chemical Education