Skip to main content

Comparation of the quantification of the proliferative index KI67 between eyeball and semi-automated digital analysis in gastro-intestinal neuroendrocrine tumors



Neuroendocrine tumors (NETs) constitute tumors widely distributed and with heterogeneous biological behavior. For gastrointestinal neuroendocrine tumors (GI-NETs) the following prognostic factors have been identified: location, production of hormones, size and proliferative grade. The latter must be calculated using proliferation index by the number of mitosis or the proportion of tumor cells positive for Ki67 immunostaining. The objective of this study was to use a quantitative tool to calculate the Ki67 index in GI-NETs.

Material and methods

We reviewed 40 cases of GI-NETs diagnosed at the Department of Pathological Sciences, Santa Casa de Misericórdia de São Paulo Hospital between 2004 and 2014 and compared the Ki67 index by manual count using scanned photomicrographs with semi-automated digital analysis (MC) and eyeball estimation (EE) of the histological slide.

After Ki67 immunostaining, the slides were scanned with 3DHistech Pannoramic Scanners. Hot spots were selected and exported in a high-resolution image format and the Ki67 index was calculated with semi-automated image analysis software (AxioVision 3.0). Ki67 immunoreactivity was expressed as the percentage of tumor cells with nuclear staining (number of positive tumor cells/a minimum of 500 total tumor cells).


We compared the classification of the neuroendocrine tumor by using the two methods in the semi-automated method 26 maintained the same grade, while 14 were re-classified, 4 being upgraded and 10 downgraded.


In the EE method there was a larger estimate of the percentage of positivity for KI67. As the Ki67 values are the criteria for the classification of neuroendocrine tumors, the semi-automated method can have less error.


Neuroendocrine tumors (NETs) constitute a wide range of tumors derived from neuroendocrine cells (NE) that are widely distributed throughout the human body, with heterogeneous clinical behavior in terms of local growth and the presence of metastasis. The incidence of the disease is 1–2/100.000 inhabitants, and NETs represent 0.5% of all malignant neoplasms. This group of neoplasms predominantly affects females, and the most prevalent age group is above 50 years of age (Modlin et al. 2003; Caldarella et al. 2011; Taal and Visser 2004).

Since 2004, the Ki67 (MIB-1) cell proliferation marker has been used as the grading parameter for prognostic in gastroenteropancreatic neuroendocrine tumors (GEP-NET) as a method of distinguishing low-grade, intermediate-grade and high-grade tumors and motivated the European Neuroendocrine Tumor Society (ENETS) to propose a 3-tiered grading system (Strosberg et al. 2009; Rindi et al. 2006; Rindi et al. 2007; Klimstra et al. 2010a). The WHO also recommends using either mitotic rate or the Ki67 labeling index for histological grading of the inherent biologic aggressiveness of gastroenteropancreatic neuroendocrine tumors (GEP-NETs) (Kulke et al. 2010; Klimstra et al. 2010b).

Thus, NETs are divided into 3 grades: G1 or low grade (mitoses < 2/10HPF or 2mm2, or Ki67-labeling index of < 3%); G2 or intermediate grade (mitoses of 2 to 20/HPF, or Ki67-labeling index of 3 to 20%); and G3 or high grade (mitoses > 20/10HPF or Ki67-labeling index > 20%). The objective of this classification system was to introduce standardized and yet practical prognostic categories that can predict the prognosis of NETs (Bosman et al. 2010). Although the treatment between G1 and G2 is similar, some cases may present unexpected evolution, especially G2 cases with a proliferative index greater than 10% for Ki67 (Nuñez-Valdovinos et al. 2018).

Despite the use of this grading classification, the behavior of neuroendocrine tumors is uncertain; aggressiveness is variable, with low grade tumors presenting aggressive manifestations such as metastasis, and sometimes there is discordance between the mitotic rate and Ki67 index (Basturk et al. 2015).

The literature describes different methods to evaluate Ki67 expression. One of the most practical and most commonly used in pathological anatomy laboratories is semiquantitative counting (eyeball estimation). Although it is low-cost and does not need a lot of time (less than one minute), it has low accuracy. A practical and reliable alternative demonstrated by Reid et al. is a manual count from printed photomicrographs of areas with high positivity for Ki67 (hot spots) (Reid et al. 2015). This method has been assessed by a number of other studies (Tang et al. 2012; Cottenden et al. 2018; Kroneman et al. 2015; Young et al. 2013).

Thus, we propose to use an alternative method to count the number of positive Ki67 cells in gastrointestinal neuroendocrine tumors, where the histological sections will be scanned and the proliferative index will be manually counted using scanned slides with semi-automated digital analysis (MC) and compared with the results of eyeball estimation (EE).

Materials and methods

The study was approved by the local ethics committee, number 006253/2015. This work does not represent a clinical trial and was therefore not registered as such.

We retrospectively identified 74 cases with confirmed diagnoses of gastrointestinal neuroendocrine tumors (GI-NET) between 2004 and 2014 from the computer records of the Department of Pathological Sciences, Santa Casa de Misericórdia de São Paulo Hospital, included well differentiated NETs and poorly differentiated neuroendocrine carcinoma (NECs). A total of 40 of these cases were retrieved with their corresponding paraffin blocks from the files of the Hospital Pathology Division. Unfortunately, 34 cases without paraffin-embedded were not available for review and Ki67 imunohistochemical staining.

Paraffin-embedded tissue blocks with presence of tumor were cut into 4 μm-thick sections for immunohistochemistry. These sections were deparaffinized and subjected to antigen retrieval in a pressure cooker in a sodium citrate buffer (pH 6.0). Endogenous peroxidase was blocked using 5% hydrogen peroxide. Nonspecific staining was blocked in 2% normal swine serum. The slides were incubated with the primary antibodies (Ki67 – Mib-1 clone, 1:160; Dako Corporation, Carpinteria, CA, USA), counterstaining was performed using Mayer’s hematoxylin. Positive control tissue (lymph node) was stained in parallel with all the study cases. All cases were ki67 immunostaining at the same time.

Immunohistochemically stained slides were scanned with 3DHistechPannoramic Scanners (3Dhistech, Budapest, Hungry), using a 20x objective. The areas of highest density of Ki67 immunoreactivity (hot spots) were manually selected by an experienced pathologist for quantification in magnification of 400x (Fig. 1). These areas were exported in a high-resolution image format and analyzed through semi-automated digital analysis software (AxioVision 3.0 – ZeissGMb, Germany) (Figs. 2 and 3). To each case selected 2 or 4 fields hot spots. Ki67 immunoreactivity for all assessment methods was expressed as the percentage of tumor cells with nuclear staining, considering the number of positive tumor cells divided by a minimum of 500 total tumor cells (Figs. 2 and 3). Were considered positive cells when a brown color was observed in the nucleus and/or nucleolus. The cells positive were marked and numbered with green colour and the negatives cells with red colour in magnification of 400x until 500 cells in the total (negatives and positives). Thus, the cases were divided into 3 grades: G1 or low grade (Ki67-labeling index of < 3%); G2 or intermediate grade (Ki67-labeling index of 3 to 20%); and G3 or poorly differentiated neuroendocrine carcinomas, (Ki67-labeling index > 20%) (Bosman et al. 2010). Table 2 shows a comparative analysis between the MC and EE methods in respect of identification of the three groups of GI-NET.

Fig. 1

Image scanning for quantitative immunohistochemical analysis and areas of "hot-spot selected for count

Fig. 2

Ki67 positive nuclear immunoreactivity for the count (brown colour)

Fig. 3

Manual count of the digital analysis, the red dots are the positive cells for Ki67

The estimation EE by pathologist of the Ki67-labeling index was based in the inspection of the images, without performing formal counting of individual cells.

Table 2 shows a comparative analysis between the MC and EE methods in respect of identification of the three groups of GI-NET.

The mitotic index was not included, due to the limitation of the amount of tumor tissue in 23 of the biopsies, making it impossible to count 50 high power fields, as recommended in the literature (Kim and Hong 2016) most cases were 0 mitoses/10HPF.


After selection, we studied 40 cases of GI-NET. The distribution by sex, age, organ and Ki67 are shown in Table 1. The cases included 22 biopsies and 18 surgical resections.

Table 1 Distribution of the 40 cases by sex, age (median) and Ki67 (mean and median) in relation to the different organs

To compare the methods of quantification of the proliferative index, divided in 3 groups according to index proliferative. (Table 2).

Table 2 Distribution by proliferative grade according to eyeball estimation (EE) and manual count (MC)

When we compared the classification of neuroendocrine tumors based on MC and EE, there was a mismatch in classification in 35% (14/40) of the cases. 10 cases were downgraded and 4 cases upgraded when we used MC and compared with EE.


The Ki67 proliferative index has been incorporated into the WHO classification (Bosman et al. 2010) for GI-NET grading and is one of the most reliable prognostic factors. However, Ki67 assessment has limitations due to lack of accuracy, uniformity, reproducibility and consistency in quantification, especially for intermediate-grade tumors with equivocal “gray zones”.

Most papers show variabilities in interlaboratory and interobserver reproducibility in interpretation of Ki67 immunoreactivity and differences in the method of counting such as eyeball estimation, manual count with photomicrographs and digital analysis (Reid et al. 2015; Tang et al. 2012; Cottenden et al. 2018; Kroneman et al. 2015; Young et al. 2013). Therefore, there remain several challenges in determining the proliferation index in a given tumor, including the method of counting, tumor heterogeneity, and defining what constitutes a hot spot and what constitutes positive staining. However, Kroneman et al. (Kroneman et al. 2015) showed that the three quantification methods for the Ki-67 index had almost perfect agreement and correlated with patient survival. EE had higher predictive ability for survival and recurrence, although the values were significantly less than those of the other two methods. Other papers have shown that although EE is quick and low-cost it is error prone and produces different estimates depending on the individual observer (Tang et al. 2012).

Manual counting of printed photographs is increasingly considered to be the most accurate approach to determine the Ki67 index and NET grading (Reid et al. 2015). However, since manual counting is more expensive and time consuming it is not adopted in diagnostic routine, even if it is more effective (Cottenden et al. 2018).

Another form of counting described in literature is automatic digital image analysis (DIA), the time for this type of counting is unclear, however there is a need to standardize the programs for each case, and it is important to photograph areas with the best representation. In addition, it can induce errors such as counting stromal cells or lymphocytes.

We adopted the manual count method using scanned slides and semi-automated digital analysis rather than printing the photograph. First the slides are scanned, which does not need to be done by the pathologist. Then the most representative areas are selected by the pathologist and transferred as image to an analysis program for manually counting using a cell counting tool. These are high-resolution images, so we do not have to depend on a perfect print to visualize the nuclei staining. In addition, we can increase the magnification of the images for better visualization. In our results, there was a tendency for the grades to decrease, mainly from G2 to G1 (9 cases) and from G3 to G2 in (1 case); when we used digital analysis with a manual count our results were similar to Cottenden and et al. (Cottenden et al. 2018). Previous studies have also shown increases in this type of disagreement, mainly G1 to G2 (Tang et al. 2012; Kroneman et al. 2015).

Programs to count semi-automated digital analysis are easy to access and free download, as QuPath, has been developed for research applications at the Centre for Cancer Research & Cell Biology at Queen’s University Belfast, as part of research projects funded by Invest Northern Ireland and Cancer Research UK (Bankhead et al. 2017). Unfortunately, scanner equipment have an expensive price to private or public laboratory, but if increase the needs probably in the future this equipment will be more affordable.

Eyeball estimation is now strongly discouraged unless a tumor shows an unequivocally low or high index, but if fairly close to the established categorical cutoffs, there may be an error. EE may, therefore, be appropriate in cases with very low or very high Ki-67 indexes such as < 1 or > 20%, but not in cases with a low to intermediate Ki-67 index or with greater heterogenicities.

When we evaluate the percentage of positive cells by the EE, we usually do it by high increase fields and this type of evaluation can vary depending on the microscopic whose size is very variable, so it is important to count a number of tumor cells.

The main objective was to compare two methods to evaluate th Ki67 index in TNE-GI and with results to do grading classification of the GI-TNE. As there were 23 biopsies, these cases should not often show the actual classification, but we compared the results between two methods with same material that may in the future be used in material suitable for analysis.


In view of the differences between the counting methods, it is worthwhile continuing studies aimed at developing a new methodology for the quantification of Ki67. Further studies are needed to identify accurate, clinically appropriate ways of making estimations which will help to assess the evolution of the disease and the prognosis for patients.

Availability of data and materials

Not applicable.



Automatic digital image analysis


Eyeball estimation


European Neuroendocrine Tumor Society


Gastrointestinal neuroendocrine tumors


High power fields

Ki67 (MIB-1):

Anti-Ki67 antibody


Semi-automated digital analysis


Neuroendocrine cells


Poorly differentiated neuroendocrine carcinoma


Neuroendocrine tumors


World Health Organization


  1. Bankhead P, Loughrey MB, Fernández JA, Dombrowski Y, McArt DG, Dunne PD et al (2017) QuPath: open source software for digital pathology image analysis. Sci Rep 7:16878

    Article  Google Scholar 

  2. Basturk O, Yang Z, Tang LH, Hruban RH, Adsay V, McCall CM et al (2015) The high-grade (WHO G3) pancreatic neuroendocrine tumor category is morphologically and biologically heterogenous and includes both well differentiated and poorly differentiated neoplasms. Am J Surg Pathol 39:683–690

    Article  Google Scholar 

  3. Bosman F, Carneiro F, Hruban R, Theise N (2010) WHO Classification of Tumours of the digestive system. IARC Press, Lyon

    Google Scholar 

  4. Caldarella A, Crocetti E, Paci E (2011) Distribution, incidence, and prognosis in neuroendocrine tumors: a population based study from a cancer registry. Pathol Oncol Res 17:759–763

    CAS  Article  Google Scholar 

  5. Cottenden J, Filter ER, Cottreau J, Moore D, Bullock M, Huang WY et al (2018) Validation of a cytotechnologist manual counting service for the Ki67 index in neuroendocrine tumors of the pancreas and gastrointestinal tract. Arch Pathol Lab Med. 142:402–407

    CAS  Article  Google Scholar 

  6. Kim JY, Hong SM (2016) Recent updates on neuroendocrine tumors from the gastrointestinal and Pancreatobiliary tracts. Arch Pathol Lab Med 140:437–448

    CAS  Article  Google Scholar 

  7. Klimstra DS, Modlin IR, Adsay NV, Chetty R, Deshpande V, Gönen M et al (2010b) Pathology reporting of neuroendocrine tumors: application of the Delphic consensus process to the development of a minimum pathology data set. Am J Surg Pathol 34:300–313

    Article  Google Scholar 

  8. Klimstra DS, Modlin IR, Coppola D, Lloyd RV, Suster S (2010a) The pathologic classification of neuroendocrine tumors: a review of nomenclature, grading, and staging systems. Pancreas. 9:707–712

    Article  Google Scholar 

  9. Kroneman TN, Voss JS, Lohse CM, Wu TT, Smyrk TC, Zhang L (2015) Comparison of three Ki-67 index quantification methods and clinical significance in pancreatic neuroendocrine tumors. Endocr Pathol 26:255–262

    CAS  Article  Google Scholar 

  10. Kulke MH, Anthony LB, Bushnell DL, de Herder WW, Goldsmith SJ, Klimstra DS et al (2010) NANETS treatment guidelines: well-differentiated neuroendocrine tumors of the stomach and pancreas. Pancreas. 39:735–752

    Article  Google Scholar 

  11. Modlin IM, Lye KD, Kidd M (2003) A 5-decade analysis of 13,715 carcinoid tumors. Cancer. 15(97):934–959

    Article  Google Scholar 

  12. Nuñez-Valdovinos B, Carmona-Bayonas A, Jimenez-Fonseca P, Capdevila J, Castaño-Pascual A, Benavent M et al (2018) Neuroendocrine tumor heterogeneity adds uncertainty to the World Health Organization 2010 classification: real-world data from the Spanish tumor registry (R-GETNE). Oncologist. 23:422–432

    Article  Google Scholar 

  13. Reid MD, Bagci P, Ohike N, Saka B, Erbarut Seven I, Dursun N et al (2015) Calculation of the Ki67 index in pancreatic neuroendocrine tumors: a comparative analysis of four counting methodologies. Mod Pathol 28:686–694

    Article  Google Scholar 

  14. Rindi G, Klöppel G, Alhman H, Caplin M, Couvelard A, de Herder WW et al (2006) TNM staging of foregut (neuro)endocrine tumors: a consensus proposal including a grading system. Virchows Arch 449:395–401

    CAS  Article  Google Scholar 

  15. Rindi G, Klöppel G, Couvelard A, Komminoth P, Körner M, Lopes JM et al (2007) TNM staging of midgut and hindgut (neuro) endocrine tumors: a consensus proposal including a grading system. Virchows Arch 451:757–762

    CAS  Article  Google Scholar 

  16. Strosberg J, Nasir A, Coppola D, Wick M, Kvols L (2009) Correlation between grade and prognosis in metastatic gastroenteropancreatic neuroendocrine tumors. Hum Pathol 40:1262–1268

    Article  Google Scholar 

  17. Taal BG, Visser O (2004) Epidemiology of neuroendocrine tumours. Neuroendocrinology 80(suppl1):3–7

    CAS  Article  Google Scholar 

  18. Tang LH, Gonen M, Hedvat C, Modlin IM, Klimstra DS (2012) Objective quantification of the Ki67 proliferative index in neuroendocrine tumors of the gastroenteropancreatic system: a comparison of digital image analysis with manual methods. Am J Surg Pathol 36:1761–1770

    Article  Google Scholar 

  19. Young HT, Carr NJ, Green B, Tilley C, Bhargava V, Pearce N (2013) Accuracy of visual assessments of proliferation indices in gastroenteropancreatic neuroendocrine tumours. J Clin Pathol 66:700–704

    CAS  Article  Google Scholar 

Download references


Not applicable.


Fundação Amparo à Pesquisa – Santa Casa 005/2015.

Scientific initiation from Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq).

Author information




MLB: Participated in the acquisition of clinical and pathological data. FSK: Participated in the acquisition of clinical and pathological data. FDCB: Responsible for the conception and design of the study, participated in the analysis and interpretation of data, and drafted the manuscript. All authors read and approved the final manuscript.

Corresponding author

Correspondence to Fabíola Del Carlo Bernardi.

Ethics declarations

Ethics approval and consent to participate

The study was approved by the local ethics committee, number 006253/2015. This work does not represent a clinical trial and was therefore not registered as such.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (, which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver ( applies to the data made available in this article, unless otherwise stated.

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Basile, M.L., Kuga, F.S. & Del Carlo Bernardi, F. Comparation of the quantification of the proliferative index KI67 between eyeball and semi-automated digital analysis in gastro-intestinal neuroendrocrine tumors. Surg Exp Pathol 2, 21 (2019).

Download citation


  • Neoplasms
  • Neuroendocrine tumors
  • Digital imaging analysis
  • Ki-67 antigen, accuracy