Kelsey Kleven

Zoology 152

Tuesday 1:20



Analyzing for the Presence of Iron in Small Intestinal Villi


                Anemia, a very serious problem among prematurely born infants, is sometimes caused by inadequate iron absorption by the small intestine.  To better understand iron absorption, intestinal tissue samples from newborn rats were harvested and prepared onto slides.  The standard histochemical iron staining was quantitated by the standard hand counting method and compared to the newer digital imaging method, ImageJ software.  These methods were compared to determine the most objective and practical way to quantitate iron.  Hand counting allowed statistically acceptable consistency.  The time invested in personalizing ImageJ software for the iron measurement was more than anticipated.  Iron quantitiation required adjustment of the program for each individual slide. Due to time constraints and complexity of analysis, the consistency analysis of the digital imaging method was not performed.  Hand counting appears to be the most practical method for current needs. If the relative importance of intestinal iron measurement increases for the lab, the digital method may be more practical for future experiments.




            Anemia in premature infants is currently one of the most prevalent problems in neonatal intensive care units (Samanci et al., 1996).  In a broad sense, it develops when there is insufficient erythropoiesis, or production of red blood cells.  Red blood cells are made after the body senses anemia and produces a hormone called erythropoietin.  This hormone stimulates bone marrow stem cells to become red blood cells (Lappin, 1996).  These red blood cells must then survive, multiply, incorporate iron and produce hemoglobin, the oxygen-carrying component of blood.  Often there are multiple reasons why anemia occurs.  One reason is inadequate iron absorption in the small intestine (Kling and Winzerling, 2002).

            The majority of the body’s iron is absorbed by cells lining the first portion of the small intestine, an area known as the duodenum (Andrews, 1999).  The tubular small intestine has, long, finger-like projections approximately 1mm in length known as villi which protrude into its center and improve absorption by increasing the surface area (Roy and Enns, 2000).  (See Figure 1)

            To better study iron absorption in premature infants, the Kling lab is planning to quantify intestinal iron content in newborn rats fed different forms and different amounts of iron. In one group, the pups are dam fed.  In the other group, the pups receive rat milk substitute (RMS) diet.  The purpose of having these two groups is to study the differences in iron status depending on the source of the diet.  Several measures of iorn status are employed; i.e. plasma iron, red cell iron, and tissue iron (Dubuque et al, 2002).  After experimental feeding, intestinal tissues will be harvested and prepared on microscope slides.  These slides will be stained for iron and evaluated either microscopically (hand counting) or digitally to quantify the iron content in the small intestinal villi. 

            Hand counting is the standard method used when analyzing tissue slides.  However, the process is very time consuming, subjective, and error-prone.  When hand counting, the number of villi stained positively for iron, i.e. blue, and the number of individual lining cells that contain blue stain within each villi will be counted.  Some cells or regions of a villi are densly blue, where other cells or regions of the same villi contain a speck of blue, indicating very little iron.   Currently, it is difficult to quantify this varying degree of iron staining.  Hence, digital imaging methods are being investigated.  Digital methods are available and being used in scientific literature in other research labs on campus and elsewhere (Ruifrok, 1997).

            Digital methods potentially provide a more objective and precise measurement of a visually-displayed image.  ImageJ, version 1.29, is an open source image processing application written in Java by Wayne Rasband of the National Institute of Mental Health in Bethesda Maryland.  Versions of ImageJ exist for Linux, Windows, and Macintosh platforms (Rasband, 2003).  ImageJ can display, edit, analyze, process, save, and print 8-bit, 16-bit and 32-bit images.  It can read many different imaging formats including TIFF, GIF, and JPEG, and is capable of measuring distances and angles, scaling, rotating, flipping, and has eight levels of magnification.  Also, ImageJ allows contrast enhancement, density profiling, sharpening, edge detecting, and for pixel measurements to be calibrated in units such as micrometers.  Examples of its use include adjusting histogram values for color image analysis and density measurements (Ruifrok, 1997, Furness et al, 1997, Goldfarb, 2001).  Furthermore, a region of interest (ROI) within the image can be user-defined so that only significant regions will be analyzed.  Most importantly, however, it allows for users to create plug-ins in Java to perform complex operations on images opened in ImageJ (Rasband, 2003, Research Services Branch, 2003).  A plug-in is a series of commands, which the program and computer reads and interprets to perform certain desired functions.  In the case of imaging programs, plug-ins can be written to repeat the same function multiple times, each time with a different image file (Rasband, 2003).  Plug-ins can be a great time saver and allow users to develop their tools and commands specific to their project.  They greatly expand the capabilities of any imaging program.

            It is my intention to compare the degree of iron staining using microscopic hand counting versus digital imaging software to develop an accurate and efficient way to quantify the amount of iron absorbed.  Thus, I will quantify the proportion of villi staining positive for iron and the proportion of individual cells within each villi staining positive for iron and then determine the degree of iron-staining.  Based on the advanced capabilities of digital imaging programs, I hypothesized that ImageJ would be more feasible than hand counting.  I predicted that it will require less time, will not induce the fatigue associated with manual counting, and that the increased precision will out way any disadvantages observed.


Materials and Methods:

            This research project is under the guidance and supervision of Dr. Pamela Kling of the Department of Pediatrics and Division of Neonatology at Meriter Hospital.  She is also an assistant professor with the University of Wisconsin - Madison and bases her research work at the University of Wisconsin Hospital.

            Archived tissue iron stains produced from previous Kling lab experiments examining the GI tract of neonatal rats were used to compare the two different methods of quantification. These samples were taken sometime within the past two years and stored.  The slides came from experiments which span the range of potential experimental iron feeding.  These slides were made after harvesting the segments of small intestine from experimental animals.  Two-centimeter tubular sections of the small intestine were cut, opened and fixed using paraformaldehyde to preserve the tissues.  After fixing, tissues were mounted on paraffin blocks and sliced very thin (approximately 2 mm) with a microtome. (Dvorak et al, 2000)  The slides were then stained for iron using Gomori’s Method which contains Hydrochloric Acid and Potassium Ferrocyanide (Luna, 1968). 

            First, hand counting by a Zeiss light microscope was used to view the slides, and a five chamber, Fisher Scientific Cell Counter was used to track the number of cells analyzed.  While analyzing slides, the reader was blind as to which sample was in question as a way to avoid unwanted bias.  When applying this method, each slide was read three times. 

            The first reading counted the total number of villi on that particular slide.  When counting villi, strict rules were developed and followed to eliminate subjectivity and discrepancies.  These are as follows:

1.) Only count whole villi.  This is important because sometimes the villi are sectioned into smaller segments making it nearly impossible to determine which segments are a part of which villi.  Also, villi may be cut at different angles giving an inaccurate and inconsistent sample.  Another important thing consideration is that depending on the part of the small intestine being analyzed, villi lengths greatly vary.  Therefore, the counter must familiarize him or her self with the length of villi corresponding with the source and not count incomplete villi.  The villi tip is crucial when looking for iron because most absorption takes place in the tips of villi.  (See figure 2)

2.) Do not count free-floating villi unless they represent a whole villi.  This means that they must be the approximate length of a villi and also contain the actual tip of that villi.  (See figure 3)

The purpose of getting an initial villi count is to get a base number for determining the percentage of villi containing iron. 

            After counting villi, we determined whether cells are stained positive for iron, i.e. blue.  If there was no visible iron, that slide was labeled as negative.  The positively stained villi were counted, evaluated, and classified into three different categories.  A villi was reported as having 1-25% of its cells stained blue, 25-75% of its cells stained blue, or 75-100% of its cells stained blue.  These categories were further labeled at +1, +2, or +3.  If there were no intestinal endothelial cells stained blue, it was labeled as +0.  The third time the slide was read, an evaluation was made to quantify the number of villi positive for visible iron within the intravillous space.  (See Table 1)  Once the slide had been read and counted for iron, it was given an overall ranking depicting its degree of iron absorption.  Each slide containing visible iron was labeled as +0, +1, +2, or +3 as well as positive or negative for iron in the intravillous space. 

            The comparison (digital) method visualized the same slides with a Nikon -Eclipse TE2000U inverted microscope with Plan Flour lenses.  Next, a Diagnostic Instruments Inc. -Spot Insight QE cooled color camera was used and the images were captured with Diagnostic Instruments Inc. -Spot Advanced software v3.5.6 for Windows.  The pictures were taken at the lab of Dr. Jing Zheng of Meriter and saved on either a Zip disk or CD-R.  The pictures were then transferred onto a iMac G4 and analyzed using ImageJ, version 1.29.  

            To use ImageJ, two plug-ins were developed to perform the analysis.  The first plug-in, named, detects iron concentrations, accented by shades of blue, in the villi and turns those areas white.  The rest of the image is then turned black to accent the contrast between iron concentrated and areas without iron content.  The plug-in also calculates the percentage of iron area in relation to the total image area.  Regions of interest (ROI) are also supported.  The user can select an area of the image to perform the analysis on.  The plug-in then ignores the rest of the image and only analyzes the ROI.

            The villi images are stored as RGB tiff files.  Each pixel contains 8 bits of red, 8 bits of blue, and 8 bits of green.  Each 8 bits can specify one of up to 256 shades.  Thus the number of colors available for each pixel is 256 x 256 x 256 = 16,777,216, enough colors to represent every visible color in the spectrum.

            First, the IronTest code calculates the height and width (in pixels) of the image and checks to see if a Region of Interest (ROI) is defined.  A ROI allows for the user to specify exactly which part of each image is important and needs analysis.  In analyzing for the presence of iron, a ROI is necessary because each picture is of part of a whole slide, which has irrelevant areas.  The plug-in works as follows:

1. For each pixel, the IronTest plug-in stores three variables - the degree of red, green, and blue.  The research done by Ruifrok is similar in that he also worked with color ananlysis using RGB values (Ruifrok, 1997).

2. If the amount of red in a pixel is below a specified threshold, the amount of green is below a specified threshold, and the amount of blue is above a specified threshold, the pixel is turned to white.  Otherwise, the pixel is turned to black.  The thresholds can be manipulated to accurately represent the presence of iron.  However, because no two pictures are the same, different threshold values must be determined for each individual picture.  The desired thresholds for each picture are determined through trial-and-error and comparing an actual picture with the white-on-black result.

3. Counters add up the number of white pixels and the number of black pixels.  Once all the pixels are turned either white or black, the percentage of the image or ROI containing iron can be calculated and displayed in a pop-up window.

4. The produced black and white picture can then be saved and compared to the original image.

            The second plug-in developed for this project is named and allows for batch analysis of images.  A user must create a text file that lists all images to be analyzed and save the text file in the same directory that the images are stored.  When running the plug-in from within ImageJ, a box pops up giving the user the opportunity to select the text file to be used.  The IronTestBatch code works very similar to the IronTest code.

1. The code reads the text file defined by the user.

2. For each image defined in the file, it calculates the percentage of blue pixels and writes the value into a results window.

The logic for determining which pixels are blue and which ones are not is identical to the IronTest code except that the IronTestBatch code does not actually show the white-on-black images.  The IronTestBatch only determines if the pixel is blue (according to the specified standards for each individual picture) and counts the number of blue and non-blue pixels.

Data analysis:           

            The data of nine slides counted three times each were analyzed by Statview, version 5, and are expressed as mean ± 2 standard deviations (95 % confidence intervals) when assessing precison.  Many assays used in research laboratories utilize a cutoff of 5% coefficent of variation to assess rejection for insufficient precision (PJ Kling, 2003).  Coefficients of variation were then utilized to determine the precision between results with hand counting.  The coefficient of variation represents an average deviation from the mean (Statview, 1999).  To obtain the data and a mean, each slide was counted blindly in triplicate. 

            Analysis of variance (ANOVA) was used to determine if differences in percent of the villi positive for iron between experimental treatments could be observed with hand counting of each slide once.  ANOVA determines the significance of the factors in a model by calculating how much the variability can be explained by those factors.  When examining differences between treatment groups, data are expressed by the mean ± two standard deviations of the mean.



            Of the nine representative slides examined (A - I), I counted between 50 ± 4 villi on a slide to as many as 230 ± 31 villi on each 2 cm length of small intestine on each slide.  Blinded counting of the same slides yielded a coefficient of variation range from between 0.021 to 0.634. See Table 2 for display of all coefficients of variation.

            When counting the percent of the villi positive for iron in representative slides in triplicate, I observed between 1.2 ±2 % to 69 ± 19% of the villi to be positive.  See Figure 4 for graphic display of 9 different slides (A - I).  The coefficient of variation ranged from 0.138 to the highest coefficient of variation to .867.  See Table 2 for coefficients of variation. When the percent of blue staining was statistically anaylized, most of the coefficients of variation were inside the usual limits for acceptance.  In repetitive assays, acceptable coefficients of variations are usually <15% or .150, but when the mean number is close to zero, higher numbers are acceptable (40% or .400) (Chard, 1990).  Additionally, low coefficients of variation were observed when evaluating the precision with scoring of 0, 1+, 2+, 3+ and intravillous staining, as shown in table 2. Each slide was completely assessed and recorded in a worksheet within 15-20 minutes.

            To determine whether the manual method was sufficient to discriminate differences between treatment groups after a single counting, I examined another set of slides with 4 different treatments, 5 different rats from each treatment group (total of 20 rats). These included RMS control, RMS plus enteral iron, dam fed and dam fed plus enteral iron. The findings are shown in Figure 5. There were statistically significant differences between treatment groups by ANOVA (p<0.0001), with dam fed and dam fed with iron differing from the RMS control group.  A post hoc test was also performed on this set of data to show whether or not the results corresponding with each treatment group are statistically different from each other; in other words, to show that the differences in treatment groups are not due to chance.

            Due to time constraints and the complexity of the project, the statistical results obtained from the digital imaging method are incomplete.  However, the time to required to learn and obtain proficiency with the microscope and camera was 2 hours. To learn the language and write the plug-ins took approximately another 16 hours of combined effort with a consultant who has experience with other types of computer programming and imaging.  After the plug-ins were designed, they had to be applied. 

            To begin analysis of a single 2cm length of intestine, the slide required scanning into 7 discreet images. This process required twenty-five minutes.  Some images required re-scanning as a result of unforeseen difficulties.  Transforming the seven images into the white-on-black image required a total of 140 minutes and a sophisticated computer (Macintosh with OS 10 capabilities).  After reporting the time required and complexity of these analyses to Dr. Kling, she suggested to terminate the assessment of precision with the digital counting methods.



            When comparing hand counting slides and using a digital method, both methods have advantages and disadvantages.  Disadvantages of hand counting include that it’s variable and causes physical fatigue and eyestrain that comes with looking into a microscope for extended periods of time.  However, hand counting had its advantages.  The most significant advantages was that hand counting is easily taught and learned in a one-on-one setting.  Another very important advantage of hand counting is that it acknowledges the presence of iron within the intravillous space which might be especially valuable information with future studies.

            Using a digital method and imaging program was more precise and less prone to subjectivity.  However, it too had disadvantages.  First of all, writing plug-ins was time consuming and required much extra work and knowledge.  Second, only one region of interest (ROI) could be defined at a time.  This means that only one villi could be accurately analyzed at a time.  A third disadvantage was that the RGB threshold values were determined by trial-and-error for each picture separately.  This was also time-consuming and made it difficult to analyze more than one picture at a time.  Another important disadvantage of using a digital method was that is not possible to calculate the total percentage of blue villi on a slide because an entire tissue sample cannot be analyzed at one time.  Instead, each slide had to be analyzed in multiple segments and the overlap of these segments had to be eliminated.

            The major objective of this project was to develop a more feasible method to use when analyzing slides compared to the standard hand counting method.  In particular, we were investigating the practicality of learning and using a digital imaging method. ImageJ is a great program with seemingly unlimited possibilities provided one has the resources to take advantage of all the program has to offer.  However, for the purposes of the Kling lab, the disadvantages associated with ImageJ were too significant for it to be used as the primary method for analyzing the iron content in the small intestinal villi.  Much more time will be required to further develop an efficient method to do the necessary analysis.  This time includes educating ourselves on further use of ImageJ and learning to write successful plug-ins.  Currently, the research being done in the Kling lab does not require the accuracy of which ImageJ is capable.  For example, the primary endpoint within their research on anemia comes from other sources; tissue iron concentration is a secondary quantification.  Also, based on our archived slides, the differences in staining from each treatment group being studied were great enough that they were seen without extensive analysis.  However, if in future studies more subtle differences are seen between treatment groups being analyzed and a more precise analysis is needed, ImageJ would be more feasible and a more worthwhile investment.    

            Hand counting, on the other hand, served all the same necessary purposes that ImageJ did without the added costs and time commitment.  The results obtained from hand counting are more general (i.e. +1, +2, +3), but for the purpose intended, the results were acceptible.  Because of this, hand counting is definitely a practical route provided the iron content in the small intestinal villi is not the primary experimental outcome.  Our hypothesis was not supported by our data.

            The results from this project will be used when making decisions as to the most practical way to quantify iron in the small intestinal villi.  By determining exactly how much iron stains positive with respect to each treatment, connections can be made as to the influence of diet, iron supplement and rhEpo therapy on iron absorption.  If there is little or no iron staining within the villi, it could mean that the iron has been quickly absorbed, which can be confirmed by high plasma iron levels which will have already been obtained, which may be the case with rhEpo therapy (PJ Kling, 2003). 

            Understanding the role rhEpo treatment and iron supplementation play in treating anemia provides an endless opportunity for more extensive research.  In particular, future research will examine what happens to the iron once it has been absorbed.  The data generated from this project will aid in understanding the role of iron supplementation in improving anemia in premature infants and aid the design of additional projects in this important field.

Al-Manhal International School.  Work of the small intestine [Online].  [April 24, 2003].


Andrews NC.  1999.  Disorders of iron metabolism.  New Eng. J. of Med. 341:1986-1995


Chard T.  1990.  Laboratory Techniques in Biochemistry and Molecular Biology.  El Sevier, Amsterdam-New York-Oxford: 216-219.


Dubuque SH, Dvorak B, Woodward SS, McCuskey RS, and Kling PJ.  2002.  Iron-deficient erythropoiesis in neonatal rats.  Biol. Neonate. 81:51-57


Dvork B, McWilliam DL, Williams CS, Dominguez JA, Machen NW, McCuskey RS, and Philipps AF.  2000.  J Pediatr Gastroenterol Nutr.  31:118-120.


Furness PN, Rogers-Wheatley L, and Harris KP.  1997.  Semiautomatic quantification of macrophages in human renal biopsy specimens in proteinuric states.  Jour of Clin Path.  50(2):118-122.


Goldfarb MF.  2001.  Analysis of casein using two-dimensional electrophoresis, western blot, and computer imaging.  Adv in Exp Med and Biol.  501:535-539.


Kling PJ.  2003.  Assistant Professor & M.D.  Department of Pediatrics, Division of Neonatology, Meriter Hospital, Madison, WI 53715.  Pers. comm.


Kling PJ, and Winzerling JJ.  2002.  Iron status and the treatment of the anemia of prematurity.  Clin Perinatol.  29:283-294.


Lappin TRJ, and Rich IN.  1996.  Erythropoietin – the first 90 years.  Clin Lab Haem.  18:137-145.


Luna LG.  1968.  Manual of Histological Staining Methods of the Armed Forces Institutes of Pathology.  McGraw-Hill Book Company, New York-Toronto-London-Sydney:179.


Rasband WS.  Research Service Branch of National Institute of Mental Health.  ImageJ [Online].  [March 29, 2003]


Research Services Branch of National Institute of Mental Health.  About NIH image [Online]. [Feb.15, 2003].


Roy CN, and Enns CA.  2000.  Iron homeostasis: new tales from the crypt.  Blood.  96:4020-4027.


Ruifrok AC.  1997.  Quantification of immunohistochemical staining by color translation and automated thresholding.  Anal & Quant Cytol & Histol.  19(2):107-112.


Samanci N, Ovali F, and Dagoglu T.  1996.  Effects of recombinant human erythropoietin in infants with very low birth weights.  J. of Int. Med. Research.  24:190-198.


Statview.  1999.  Statview Reference.  SAS, USA:  79-80, 84-86.
Table 1.  Hand Counting Classifications




0% of cells blue





1-25% of cells blue





25-75% of cells blue





75-100% of cells blue





Blue Lumen









the lumen contains visible iron



the lumen may or may not contain visible iron



the lumen may or may not contain visible iron



the lumen may or may not contain visible iron



the endothelial cells may or may not contain visible iron


















Positive or Negative for Iron



Table 1.  The five different classifications used when hand counting to quantifying the degree of iron absorption in small intestinal villi.







Table 2.  Coefficients of Variations



# Villi

% Any Blue

0% Blue

1-25% Blue

25-75% Blue

75-100% Blue

Positive Intravillous Space










































































Table 2.  Most coefficients of variation for each slide A through I are sufficiently low, indicating that there was acceptable consistency when each slide was blindly hand counted three times.  For example, a value of zero indicates that there was no variation in any of the three trials and a value of .155 is 15.5% coefficient of variation.