Computer-Assisted Digital Image Analysis to Quantify the Mass and Purity of Isolated Human Islets Before Transplantation : Transplantation

Journal Logo

Original Articles: Experimental Transplantation

Computer-Assisted Digital Image Analysis to Quantify the Mass and Purity of Isolated Human Islets Before Transplantation

Niclauss, Nadja1,3; Sgroi, Antonino1; Morel, Philippe1; Baertschiger, Reto1; Armanet, Mathieu1; Wojtusciszyn, Anne1,2; Parnaud, Geraldine1; Muller, Yannick1; Berney, Thierry1; Bosco, Domenico1

Author Information
Transplantation 86(11):p 1603-1609, December 15, 2008. | DOI: 10.1097/TP.0b013e31818f671a
  • Free


Islet transplantation has become an established alternative to treat type 1 diabetes and relies on the mastering of the technically challenging process of islet isolation and purification. Product release criteria, consisting mostly of islet purity, viability and mass, must be met to be able to transplant an islet preparation into a human subject. Islet mass is currently assessed by the determination of the number of islet equivalents (IEQ), that is, islets normalized to an average size of 150 μm. In clinical trials conducted under the “Edmonton protocol” thresholds of 5000 IEQ/kg for the first islet infusion and 10,000 IEQ/kg in two infusions are required for the complete procedure (1).

The common method for counting human islet preparations is by direct microscopic analysis of dithizone-stained islet samples using an eyepiece reticle for islet diameter measurement. Most islets are not perfectly spherical and their approximate size is determined by mean diameter estimation. One IEQ is defined as an islet with a diameter of 150 μm. The established algorithm of Ricordi classifies islets into diameter classes using 50 μm diameter range increments. Relative conversion factors make it possible to convert the total islet number of each diameter class into IEQ number (2).

Direct microscopic analysis is always susceptible to operator-dependent variability and lacks objectivity. A reliable assessment of islet preparation is necessary to transplant the patient-targeted amount and purity of islets, based on an accurate, reproducible, and operator-independent count. In addition, the increasing multicenter collaborations for islet transplantation require a fine-tuned standardization of the procedure, including quantification of purified islets. Therefore, it is important to find a reliable and objective tool to assess IEQ number and islet purity. In this study, we present a computerized method based on digital image analysis developed to automatically quantify total islet number, IEQ number, and purity of islet preparations.


Islet Isolation

Pancreata were obtained from brain-dead multiorgan donors from Switzerland and France, and preserved organs were shipped to Geneva in standard conditions (3). Islet isolations were performed as previously described according to the Ricordi method with local adaptations (4, 5). Briefly, the pancreas was distended by intraductal infusion of a cold collagenase solution. After digestion at 37°C in a modified Ricordi chamber, separated exocrine and endocrine tissues were washed and purified in a continuous Biocoll gradient. After purification, islet fractions with three different purities were obtained (high, medium, and low). Two samples (dilution 1:2,500) of these fractions were prepared and transferred into 35-mm Petri dishes for staining with dithizone (1.5 mg/mL). One sample was analyzed by standard manual method. The other sample was digitally photographed for further dual manual and computerized analyses. Digital images were taken through a stereomicroscope equipped with a digital camera (Leica Microsystems, Renens, Switzerland).

Manual Islet Analysis

Standard manual analysis was carried out by microscopy using a calibrated grid for islet diameter measurement. The approximate size of nonspherical islets was determined by estimating their mean diameter. Purity was estimated considering the total dithizone-stained islets and surrounding unstained exocrine tissue. Multiple operators carried out standard manual analysis. The main steps of standard manual analysis are depicted in Figure 1 (method I). For manual analysis of digital images, a calibrated digital grid was generated using Adobe Photoshop and overlayed to digital images. Digital images were then analyzed manually following a procedure and criteria identical to those used for standard manual analysis (Fig. 1, method II). A unique operator carried out manual image analysis. To convert islet number into IEQ number, the Ricordi algorithm was used. This consists in classifying and counting islets according to their size using 50 μm-diameter range increments. Islets or islet fragments smaller than 50 μm diameter were not considered. Conversion factors for each number N of classified islets into IEQ are N/6, N/1.5, N×1.7, N×3.5, N×6.3, N×10.4, and N×15.8 for the 50 to 100, 100 to 150, 150 to 200, 200 to 250, 250 to 300, 300 to 350, and 350 to 400 μm islet diameter ranges, respectively. These factors make it possible to convert the total number of islets into IEQ number from any preparation (2).

Study flow chart of procedures to obtain IEQ number. Five different methods described in the figure were used to calculate IEQ number from each islet preparation and are identified by roman symbols (I–V).

Computerized Islet Analysis Using MetaMorph Software

For computerized analysis, digital images of all islet preparations were analyzed using the offline MetaMorph imaging software for microscopy (Universal Imaging Corporation, West Chester, PA) programmed to automatically quantify purity, number of islets, and IEQ number. A single operator carried out computerized analysis. Operations of these programs include separation, binarization, and thresholding of red and white colors of digital images corresponding to dithizone-stained islets and exocrine tissue, respectively (Fig. 2A–C). Real-size determination was performed using a dividing factor equal to image magnification. Total profile surface areas of islets and exocrine tissue were automatically calculated. The size of each individual islet was defined as its equivalent sphere volume, calculated as the volume of a sphere that would have an equatorial cross-sectional area equal to that of the profile area of the islet (Fig. 1, methods III and IV). Alternatively, islet size was calculated using equivalent prolate volumes (Fig. 1, method V). The equivalent prolate volume is defined as the volume of a prolate spheroid (cigar-shaped object) produced by revolving an ellipse around its major axis matching that of the object. Considering that one spherical IEQ with a diameter of 150 μm has a volume of 1,767,146 μm3 (volume of a sphere: V=4/3πr3, where r is the radius of the sphere), it is possible to calculate an IEQ value for each islet by the following formula: equivalent spherical volume/1,767,146 μm3. IEQ number was then calculated by classifying islets according to their diameter and calculating IEQ using the Ricordi algorithm as in the standard analysis method (Fig. 1, method III) or by considering individual IEQ of each islet (Fig. 1, method IV).

Digital image analysis of dithizone-stained human islets. One representative sample of a purified human islet preparation stained with dithizone is shown (A). Operations of the MetaMorph Imaging Software for Microscopy include separation, binarization, and thresholding (orange in B and C) of white and red colors corresponding to exocrine tissue (B) and dithizone-stained islets (C), respectively. Adjacent islets (green in C) are marked using a parting line (red in C) to allow them to be counted separately.

Results were transferred to a Microsoft Excel spreadsheet for calculations of purity, total islet number, and IEQ number. Purity was calculated by using the following formula: purity=islet area/[islet area+exocrine area]. Numbers calculated by the MetaMorph program in diluted samples were multiplied by the sample dilution factor (2500) to obtain total islet number and IEQ number of each preparation.

Computerized Analysis Using ImageJ Software

Twelve different digital images corresponding to 12 different islet preparations with varied islet numbers and purities were analyzed using both the MetaMorph software and the ImageJ free software (Rasband, W.S., US National Institutes of Health, Bethesda, MD, Operations performed with ImageJ were similar to that of MetaMorph and included separation, binarization, and thresholding of colors. Real-size determination was performed using a dividing factor equal to image magnification. The size of each individual islet was defined as its equivalent sphere volume, calculated as the volume of a sphere that would have an equatorial cross-sectional area equal to that of the profile area of the islet.

Glass Microsphere Analysis

To validate computerized analysis, two different lots of standardized red glass microspheres with known mean diameters of 220 and 370 μm, respectively (Corpuscular Inc., Cold Spring, NY) were analyzed. Microspheres were analyzed by direct microscopy and using the MetaMorph software in the same way as performed for islets.


All statistical analyses were performed using SPSS for Windows software (SPSS Inc., Chicago, IL). Results are expressed as means±SEM. Statistical analysis was performed by analysis of variance and least significant difference post hoc tests. Student’s t test was performed for comparison between two groups. Correlation was tested by determining the Pearson correlation coefficient r. Correlation was considered as good for a coefficient of determination r2 more than 0.8. Significance was taken at P less than 0.05.


Comparison of Two Different Softwares for Computerized Islet Analysis

Digital images of aliquots of 12 different islet preparations, comprising a total of 202 islets, were analyzed using both the ImageJ and MetaMorph softwares. IEQ of each individual islet (Fig. 3A) and, consequently, IEQ number of each islet preparation (Fig. 3B) calculated with ImageJ and MetaMorph were similar (r2=0.99). In addition, there was a good correlation between islet preparation purities calculated with ImageJ and MetaMorph (r2=0.91, Fig. 3C). These results indicate that IEQ and purity values automatically calculated by computer are not software dependent. Because the MetaMorph software is better established and we are more familiar with its routine use, this software was chosen for subsequent analyses aimed at validating computerized analysis in comparison with manual analysis.

Correlations between data obtained with ImageJ and MetaMorph softwares. Individual IEQ (A), IEQ number (B), and purity values (C) obtained with ImageJ are compared with those obtained with MetaMorph. (A and B) r2=0.99, n=202 (A), n=12 (B); (C) r 2=0.91, n=12.

Comparison of Standard Manual, Digital Manual, and Computerized Analyses

Analyses described in this paragraph are depicted as methods I, II, and III in Figure 1.

To ascertain comparability of standard manual islet counting and the various analyses carried out on digital images of the preparation, we first had to compare direct manual counting on the islet preparation (method I) and manual counting on its digital image (method II). As a second step, computerized analysis using MetaMorph (method III) was compared with manual analysis on the digital image (method II).

Total islet numbers were 152,901±12,650, 96,555±8,581, and 100,578±8,441 for standard manual (method I), digital manual (method II), and computerized (method III) analyses, respectively (Fig. 4A). The higher value for standard manual analysis was due to a higher count of small islets using direct microscopic analysis when compared with digital manual and computerized analyses (Fig. 4D). Many small islets with a diameter smaller than 50 μm were counted by standard manual analysis and not by digital manual and computerized analyses. IEQ numbers calculated with the Ricordi algorithm were 94,025±7,420, 94,426±7,863, and 93,280±8,110 for standard manual (method I), digital manual (method II), and computerized analyses (method III), respectively (Fig. 4B). These values are near identical and the small differences were not significant. Distributions of IEQ numbers according to islet diameters were also similar for standard manual analysis, and digital manual and computerized analyses (Fig. 4E). Purity values were 52%±3%, 48%±3%, and 43%±3% for standard manual (method I), digital manual (method II), and computerized analyses (method III), respectively (Fig. 4C). Standard manual analysis slightly but significantly overestimated (+9%, P<0.05) purity compared with computerized analysis. When individual values of IEQ number (Fig. 5A), total islet number, and purity were compared, a good correlation between digital manual and computerized analyses was observed with coefficients of determination r2=0.97, 0.96, and 0.89, respectively. Coefficients of determination r2 between standard manual and computerized analyses were 0.68, 0.62, and 0.73 for IEQ number (Fig. 5B), total islet number, and purity, respectively.

Standard manual, digital manual, and computerized analyses. Total islet number (A), IEQ number calculated with the Ricordi algorithm (B), and islet purity (C) are compared using standard manual (method I, white columns, n=107), digital manual (method II, black columns, n=115), and computerized analyses (method III, hatched columns, n=115). Islet (D) and IEQ numbers (E) are also categorized by islet diameter using 50 μm increments. Values are shown as mean±SEM. *P<0.001 when compared with digital manual or computerized analysis. **P<0.05 when compared with computerized analysis.
Correlation between digital manual and computerized analyses. (A) Individual values of IEQ number calculated with the Ricordi algorithm are compared using digital manual (method II) and computerized analyses (method III) (r 2=0.97, n=115). (B) Individual values of IEQ number calculated with the Ricordi algorithm are compared between standard manual (method I) and computerized analyses (method III) (r 2=0.68, n=107).

To explore the impact on interoperator reproducibility of the method, eight different preparations were counted independently by two different operators, using standard manual (method I) and computerized (method III) analyses. As anticipated, the computerized method had better reproducibility than the standard manual method, as shown by better interoperator correlation of IEQ numbers and lower interoperator variability. Coefficients of determination r2 were 0.760 and 0.996, and coefficients of variability were 13.3% and 2.9% (P<0.01), for the standard manual and computerized methods, respectively.

Comparison of the Ricordi Algorithm and Individual Islet Volume Measurement to Calculate IEQ Number

Analyses described in this paragraph are depicted as methods III and IV in Figure 1. Computerized analysis of digital images makes it possible to calculate IEQ values for individual islets. This was carried out by dividing the IEQ sphere volume by the volume of one 150 μm-diameter islet (1,767,146 μm3). From the same digital image, IEQ number was calculated by using the Ricordi algorithm (method III) or from the arithmetic sum of individual IEQ values (method IV). As shown in Figure 6(A), IEQ number was approximately 16% higher when calculated with the Ricordi algorithm, when compared with the arithmetic sum of individual IEQ (93,280±8,110 vs. 80,337±7,176; P<0.0001). There was an excellent correlation (r2=0.98) between IEQ number calculated with the Ricordi algorithm and by the arithmetic sum of individual IEQ (Fig. 6B).

Ricordi algorithm and individual islet volume measurement to calculate IEQ number. (A) Columns show means±SEM of IEQ numbers (n=115) calculated using Ricordi algorithm (method III) or individual islet volumes (method IV). (B) Individual values of IEQ number calculated with the Ricordi algorithm (method III) are compared with those calculated considering individual islet volumes (method IV) (r 2= 0.98, n=115). *P<0.0001.

Comparison Between Equivalent Sphere and Prolate Volumes to Calculate IEQ Number

Analyses described in this paragraph are depicted as methods IV and V in Figure 1. As an alternative to equivalent sphere volumes (method IV) used to calculate IEQ values for individual islets, we used equivalent prolate volumes (method V). IEQ numbers calculated by using equivalent prolate volumes were approximately 40% higher (P<0.0001) when compared with those calculated by using equivalent sphere volumes (120,225±19,854 vs. 86,352±14,789; n=35), and correlation of individual IEQ values was high with a coefficient of determination r2=0.98.

Analysis of Glass Microspheres of Known Size

To validate our computer-assisted digital image analysis, two batches of glass microspheres whose diameters ranged between 190 and 250 and 325 and 410 μm were analyzed by standard (method I) and computerized (method IV) procedures. We obtained near-identical values for diameter, total sphere number, and IEQ number (Fig. 7A–C), and correlation between individual diameter values was high (r2=0.98) (Fig. 7D).

Analysis of glass microspheres. Two sets of glass microspheres with diameters ranging from 190 to 250 (n=26) and 325 to 410 μm (n=12) were analyzed. Diameters (means±SEM) (A) and total numbers (B) and IEQ numbers (computerized spheric, C) were evaluated by standard manual (white columns) and computerized analyses (hatched columns). Diameters of individual glass microspheres are compared using standard manual and computerized analyses (D, r 2=0.98, n=38).


We have developed, assessed, and validated a computer- assisted digital image analysis method designed to determine the numbers of islets and IEQ and the purity of an islet preparation.

This tool should allow to eliminate interindividual variability and to gain time in the counting and assessment process. All personnel involved in the islet isolation procedure are familiar with the level of subjectivity encountered during manual evaluation of individual islet size and purity of the overall preparation. Consequently, islet size and purity are subject to inter- and intraindividual variability and errors by standard microscopic procedure. Concerning the islet size, the estimation of mean diameter of nonspherical islets is the major problem. This could result in an inaccurate estimation of IEQ number for each size category of islets.

Individual values of total islet number, IEQ number, and purity were much better correlated between digital manual and computerized analyses than between standard manual and computerized analyses. The dispersion around the regression line in the standard manual approach (Fig. 5) illustrates the inter- and intraoperator variabilities of this method, and is a well-known problem in the assessment process of human islets.

The accuracy and reliability of this technique were provided by the analysis of glass microspheres of known diameter and the demonstration that counts were identical by comparing results obtained by the computerized and manual methods with the actual size of the microspheres.

A higher number of small islets (50–100 μm) was found by standard manual analysis when compared with computerized analysis. The explanation is that many small islets have a real diameter size smaller than 50 μm and are not considered by computerized analysis. Nevertheless, the above-mentioned difference in small islet numbers has only a small impact on total IEQ numbers, because IEQ volume is proportional to its cubed diameter. This is reflected by the nearly identical mean IEQ numbers obtained for standard manual, digital manual, and computerized analyses. It is also always difficult to differentiate intact small islets from islet fragments, which should not be taken into account anyway. On the other hand, it was recently suggested that small size islets (<50 μm) may have superior function (6). If this finding is confirmed, the computerized method could underestimate the true value of an islet preparation by disregarding the smallest islets.

The Ricordi algorithm has been developed to calculate IEQ numbers easily and quickly. The result is obviously approximate, because islets differing up to 50 μm in diameter are pooled into a same category and corrected by the same factor. For instance 151 μm and 200 μm diameter islets are both multiplied by 1.7, but the difference in volume between these two islets is 230%. Using MetaMorph, it is possible to extrapolate the equivalent sphere volume and an exact IEQ value for each individual islet. When we calculated the total IEQ numbers from the same digital image using the Ricordi algorithm or equivalent sphere volumes, we observed that total IEQ numbers were 16% higher with the Ricordi algorithm. This indicates that the Ricordi algorithm slightly overestimates total IEQ numbers compared with the more precise calculation based on equivalent sphere volumes.

Islets are in fact spheroids than true spheres. Because spheroids will naturally lie on their long axis, we also evaluated total IEQ numbers using a formula for prolate spheroid calculation. Prolate spheroid formula overestimated total IEQ numbers by 40% when compared with sphere volume of individual islets, and by 20% when compared with the Ricordi algorithm. It is reasonable to assume that for some islets with a true prolate morphology the corresponding formula gives more exact IEQ values. By contrast, for most islets with a more irregular or asymmetric shape the use of a prolate spheroid formula certainly results in overestimated IEQ values.

In addition to relieve the problem of objectivity, the computerized method has the advantage of speed compared with the standard procedure. Depending on the skill of the operator, 5 to 15 min is required to complete standard manual microscopic analysis. The entire process to analyze the digital image with the MetaMorph program runs in less than 10 sec, including automatic data transfer and calculation in the Excel file. The more time-consuming step is the capture of the digital image. Indeed, the image must have adequate brightness and contrast to be processed by the MetaMorph program. Accurate image exposition will depend on sample preparation including dithizone staining and on stereo microscope and digital camera settings. Altogether and depending on personal skills the image settings and capture takes between 15 and 60 sec. This procedure also allows us to constitute a library of images that can be useful for supplementary quality control and retrospective analysis of islet preparations.

Although determination of IEQ numbers remains part of the standard pretransplant islet graft quality control in most clinical islet transplantation programs, previous reports have shown that IEQ number is a less adequate index for β-cell mass than β-cell number itself (7). Furthermore, assessment of β-cell number and function was shown to be a better tool to predict metabolic outcome of human islet transplantations than assessment of IEQ number alone (8, 9). Future developments of digital image analysis will include the assessment of β-cell numbers in human islet preparations.

In conclusion, we have presented a computer-assisted digital image analysis method and shown that it is a simple, reliable, and time-saving tool for assessing purity and islet numbers of human islet preparations before transplantation. Additionally, it eliminates operator-dependent variability, which is one major problem of standard microscopic assessment of islet preparations.


1. Badet L, Benhamou PY, Wojtusciszyn A, et al. Expectations and strategies regarding islet transplantation: Metabolic data from the GRAGIL 2 trial. Transplantation 2007; 84: 89.
2. Berney T, Pileggi A, Molano RD, et al. Pancreatic islet cells. In: Atala A, Lanza R, eds. Methods in tissue engineering. San Diego: Academic Press 2002, p 203.
3. Kempf MC, Andres A, Morel P, et al. Logistics and transplant coordination activity in the GRAGIL Swiss-French multicenter network of islet transplantation. Transplantation 2005; 79: 1200.
4. Ricordi C, Lacy PE, Finke EH, et al. Automated method for isolation of human pancreatic islets. Diabetes 1988; 37: 413.
5. Bucher P, Mathe Z, Morel P, et al. Assessment of a novel two-component enzyme preparation for human islet isolation and transplantation. Transplantation 2005; 79: 91.
6. Lehmann R, Zuellig RA, Kugelmeier P, et al. Superiority of small islets in human islet transplantation. Diabetes 2007; 56: 594.
7. Keymeulen B, Ling Z, Gorus FK, et al. Implantation of standardized beta-cell grafts in a liver segment of IDDM patients: Graft and recipients characteristics in two cases of insulin-independence under maintenance immunosuppression for prior kidney graft. Diabetologia 1998; 41: 452.
8. Ichii H, Inverardi L, Pileggi A, et al. A novel method for the assessment of cellular composition and beta-cell viability in human islet preparations. Am J Transplant 2005; 5: 1635.
9. Keymeulen B, Gillard P, Mathieu C, et al. Correlation between β cell mass and glycemic control in type 1 diabetic recipients of islet cell graft. Proc Natl Acad Sci USA 2006; 103: 17444.

Human islet; Islet isolation; Transplantation; Quality control

© 2008 Lippincott Williams & Wilkins, Inc.