New Healthcare Inventions on Breast Cancer

Abstract

Background: The Ki67 labeling index (LI) for breast carcinoma is essential for therapy. It is determined by visual assessment under a microscope which is subjective, thus has limitations due to inter-observer variability. A standardized method for evaluating Ki67 LI is necessary to reduce subjectivity and improve precision. Therefore, automated Digital Image Analysis (DIA) has been attempted as a potential method for evaluating the Ki67 index.

Materials and Method: We included 48 cases of invasive breast carcinoma in this study. Ki67 immunostaining was carried out on paraffin-embedded sections of all cases.

The Visual Assessment (VA) of the Ki67 index was carried out, and images of Ki67 stained slides were obtained under a magnification of 40X. We considered the mean value of two observers as the final visual assessment scores. Images were processed using Tissue Quant software to identify positively stained areas. We obtained the area and count of the positively stained regions. Statistical analysis by calculating the kappa value was performed to check the agreement between DIA and visual assessment.

Results: Out of 48 cases, 40 cases showed substantial agreement between DIA and visual analysis, which was established by calculating the kappa value (0.636).

Conclusion: DIA is an emerging and competitive alternative for biomarker assessment in breast cancer. It can effectively be put to use in clinical practice to obtain precise results.

Keywords: Ki67, Digital Image Analysis, breast cancer

Introduction

Breast cancer, among women, continues to be the most common cause of death due to cancer worldwide. Hence, the assessment of prognosis is an important aspect of clinical practice. The hallmark of all tumors is uninhibited abnormal cell proliferation. Ki67, a nuclear antigen, is a cell proliferation marker expressed in all but the G0 phase of the cell cycle. With the advent of molecular profiling, Ki67 LI has emerged as a potential prognostic marker in breast cancer. Many of the studies have validated the importance of Ki67 LI correlation with breast cancer outcome, particularly in patients after hormonal and chemotherapy. Also, it has been suggested that Ki67 LI can aid in identifying patients who could benefit from chemotherapy.

Currently, the majority of institutions and laboratories follow the visual assessment method to quantify the Ki67 index owing to its economical nature. However, visual assessment is tedious and poses challenges such as intra- and inter-observer variability. Furthermore, no standardized scoring system has been established thus far to designate a tumor as highly proliferative or not. Hence, a standardized method is necessary to evaluate the Ki67 index due to its impact on clinical practice. Digital image analysis is emerging as a potential alternative to visual assessment as it is less time consuming, lacks manual errors, and is capable of processing images more accurately. Various recent studies have shown a high concordance rate between visual assessment and digital image analysis of Ki67 LI. The present study was conducted to compare the visual assessment results of Ki67 LI with TissueQuant image analysis software.

Materials and Method

The present study was a two-year study, from June 2016 to May 2018, conducted in a tertiary care hospital. A total of 48 cases, diagnosed as invasive breast carcinoma on histopathology, were included in the study. Formalin-fixed, paraffin-embedded tissues were used for Ki67 LI in all the cases.

Immunohistochemistry

Formalin-fixed, paraffin-embedded sections of 3µm thickness were cut for all 48 cases and processed for immunohistochemistry. Ki67 immunostaining was done using the mouse monoclonal primary antibody MIB-1 with a dilution of 1:100, and a secondary antibody kit PolyExcel HRP/DAB Detection system by PathnSitu Biotechnologies. The visual assessment of Ki67 LI was done by counting a total of 500 cells under 40X magnification and calculating the percentage of positively stained cells. Two observers independently calculated the values for all cases, and an average of both values was taken as the final Ki67 score by the manual method.

For digital image analysis, images of the same areas screened for the manual method were taken. Images were acquired at a magnification of 40X using a Lawrence brand microscope. The resolution of the images was 4608 x 3456 pixels. These images were subsequently processed for automated analysis using the software named TissueQuant, which picks the regions of a specific color of interest in an accurate manner. The brown shade corresponding to the positively stained nucleus was considered as the reference color while performing the analysis. TissueQuant software provides a batch process option, using which all the input images were analyzed. From each image, the count of the nuclei and the total area represented by all the positively stained nuclei were obtained through this analysis. As a second step, the area representing the field of view in each image was also obtained by using the background pixel color as the reference color for TissueQuant analysis. The ratio of the total area of the positively stained nuclei to the area of the field of view was considered as a measure to be correlated with the manual scoring.

The statistical analysis for the data was done by calculating the kappa value to analyze the agreement between the manual method and DIA. The cutoff value for categorizing the cases into low and high proliferation rate was 14.

Results

In the present study, 48 cases were evaluated by visual assessment of Ki67 by two observers, as well as by DIA, using TissueQuant software. The DIA calculated the percentage area of positivity in the given image. Out of 48 cases assessed, 40 cases showed considerable concordance, which was established by calculating the kappa value. The image taken manually (1) and the digital image of the same (2) are shown below. A graph was plotted with two readings of manual, the average of the two, and DIA values as depicted below (3).

Inter-observer variability between two readings, R1 and R2, was evaluated with a P value of 0.862. The kappa value was calculated between averages of these two readings and DIA readings. A value of 0.636 was obtained, which shows there is substantial agreement between the two methods. Out of 48 cases, 8 cases showed discordance between DIA and manual values. These cases showed values more than 14 by the manual method and less than 14 by DIA. This discordance is likely to be due to a smaller area in the image.

Discussion

The present study was carried out to evaluate the agreement between Ki67 values by visual assessment and DIA. Since Ki67 is an important prognostic marker in breast cancer, the values obtained were categorized into two groups, i.e., high proliferation and low proliferation. Presently, a Ki67 index above 14 is considered high proliferation for therapeutic purposes, hence, the cutoff value for this categorization was taken as 14. Values above 14 were taken as tumors having a high Ki67 index and values below 14 were taken as tumors having a low Ki67 index. In our study, we found that out of 48 cases, 40 cases showed agreement between the manual method and DIA values. In 8 cases, the DIA values were lower as compared to the manual values. This discrepancy could be due to variation in the image sizes, which in turn, leads to a lower area/pixel calculation, hence, lower values were obtained by DIA. Faded slides and those with excessive background staining were excluded from the present study.

A similar study done by Tuominen et al. using Immunoratio software showed a good correlation between manual and DIA values. Another study done by Stalhammer et al. showed that DIA values were more accurate as compared to manual values. Koopman T. et al. performed DIA using virtual dual staining which showed excellent inter-platform agreement between two independent DIA platforms. Likewise, the present study also showed good agreement between the two methods.

The TissueQuant software used here is a simple, cost-effective, and easy tool for DIA. In the present study, images were taken using a mobile camera with reasonably good resolution to obtain fairly accurate values. Furthermore, in institutions where automated DIA machines cannot be put into use due to financial constraints, this software proves to be economical. Visual assessment is subjective and has the possibility of higher inter-observer variability, though not recorded in our study possibly due to a smaller sample size. The imaging technique needs to be uniform and a standardized method can be practiced to avoid any such discrepancies, as encountered in 8 out of 48 cases in our study. Also, calibration of the software can be done to cater to specific-sized images. This method of DIA can further be extrapolated to other cancers to evaluate the Ki67 index, such as in oropharyngeal squamous cell carcinoma. Furthermore, this TissueQuant software can also be used to assess other immunohistochemistry markers and quantify them. TissueQuant was used to assess ER, PR, and Her 2 neu status and grade them with higher accuracy.

Conclusion

A highly studied biomarker, Ki67, has proven prognostic value in breast cancer. Tumors with a high Ki67 index have a poorer prognosis compared to low-grade tumors with a low Ki67 index¹¹. As it is a marker of the proliferation index, stratification of the tumors into one of the two categories is essential. Using TissueQuant software for DIA not only makes such stratification possible but can also yield specific Ki67 indexes. Further studies can be done using DIA to evaluate the effect on prognosis with specific Ki67 values. Thus, we conclude that DIA is a promising and competitive alternative for biomarker assessment in breast cancer and can be effectively put to use in clinical practice.

References

1. Stålhammar G. et al. “Digital image analysis outperforms manual biomarker assessment in breast cancer.” Modern Pathology, 2016, 29:318-329.
2. Zhong F., Bi R., Yu B., Yang F., Yang W., Shui R. “A Comparison of Visual Assessment and Automated Digital Image Analysis of Ki67 Labeling Index in Breast Cancer.” PLoS ONE, 2016, 11(2):e0150505.
3. Fasching et al. “Ki67, Chemotherapy Response, and Prognosis in Breast Cancer Patients Receiving Neoadjuvant Treatment.” BMC Cancer, 2011, 11:486.
4. Mikami Y. et al. “Interobserver Concordance of Ki67 Labeling Index in Breast Cancer: Japan Breast Cancer Research Group Ki67 Ring Study.” Cancer Sci, 2013, 104(11):1539-1543.
5. Konsti et al. “Development and Evaluation of a Virtual Microscopy Application for Automated Assessment of Ki-67 Expression in Breast Cancer.” BMC Clinical Pathology, 2011, 11:3.
6. Koopman T., Buikema H. J., Hollema H., Bock G. H., Vegt B. “Digital Image Analysis of Ki67 Proliferation Index in Breast Cancer Using Virtual Dual Staining on Whole Tissue Sections: Clinical Validation and Inter-platform Agreement.” Breast Cancer Res Treat, 2018, 169:33–42.
7. Tuominen V. J., Ruotoistenmäki S., Viitanen A., Viitanen M., Isola J. “ImmunoRatio: A Publicly Available Web Application for Quantitative Image Analysis of Estrogen Receptor (ER), Progesterone Receptor (PR), and Ki-67.” Breast Cancer Research, 2010, 12:R56.
8. Prasad K., Tiwari A., Ilanthodi S., Prabhu G., Pai M. “Automation of Immunohistochemical Evaluation in Breast Cancer Using Image Analysis.” World J Clin Oncol, 2011; 2(4): 187-194.
9. Laurinavicius et al. “A Methodology to Ensure and Improve Accuracy of Ki67 Labeling Index Estimation by Automated Digital Image Analysis in Breast Cancer Tissue.” Breast Cancer Research, 2014,16:R35.
10. Faratzis G., Tsiambas E., Rapidis A. D., Machaira A., Xeromeritis K., Parsouris E. “VEGF and Ki67 Expression in Squamous Cell Carcinoma of the Tongue: An Immunohistochemical and Computerized Image Analysis Study.” Oral Oncology, 2009; 45(7):584-588.
11. Soliman N. A., Yussif S. M. “Ki-67 as a Prognostic Marker According to Breast Cancer Molecular Subtype.” Cancer Biol Med, 2016;13: 496-504.

Did you like this example?

730

38