Study Of The Breast Cancer Ki67 Interpretation Consistence By One Observer Using Different Assessment Methods

Objectives:The Ki67 immunohistochemistry index is a key number for classifying the breast cancer.It is also an important base for determining whether a cytotoxic drugs targeted treatment is required or not.Due to limitations of conventional assessment methods that are currently used,there exist inconsistences between different observers.Such an inconsistence can also be found even one observer performs his/her assessments at different time.This paper has made improvements of existing methods,and established a new method of assessing the Ki67 immunohistochemistry results based on the Digital Image Analysis.The paper has further studied the consistence of the interpretations by same observer with this new method when assessing the Ki67 immunohistochemistry results,and has provided a scientific rationale for classifying the breast cancer and selecting tailor-made treatment process.Methods:1.Set up new method of assessing the Ki67 immunohistochemistry results based on the Digital Image Analysis:Demarcating-square Automatic Counting In Hot Spot,(DACHS)and Demarcating-square Automatic Counting In Multiple Spot,(DACMS)?Demarcating-square Manual Counting In Hot Spot,(DMCHS)?2.Stain 30 pieces of breast cancer experimental specimens and scanned into digital images.3.One observer exercised assessment and analysis of these 30 slices with above method at 10 different time.4.Same observer exercised the assessment of these 30 slices but respectively with the method of Manual Counting Under Microscope In Hot Spot(MCUMHS).5.The results that were obtained from above different methods were processed and analyzed with R-Language of statistics for their consistence and dispersion in order to compare the repeatability between different methods.Results:1.There are totally 30 slides of the breast cancer Ki67 immunohistochemistry prepared for the experiment.The slides of the hot spots accounts for 83%.Of the hot spot slides,19 slides represent hot spot dispersed distribution and 6 slides the hot spot concentrated distribution.11slides of hot spots came from the invasive edges of the tumor organization and 14 slides from the tumor organization proper.The maximum diameter of the hot spot is 9,980?m vs.the minimum diameter 47?m of the hot spot.The hot spot diameters of 4 pathologic slides are smaller than 100?m;the hot spot diameters of other 10 pathologic slides are in range of 100-200?m.The hot spot diameters of 11 pathologic slides are greater than 200?m.2.The number of the tumor cells in a 100×100?m2 assessed area is app.57 and the number of the tumor cells in a 200x200?m2 assessed area app.180.There are about 640 tumor cells in a 300×300?m2 assessed area.3.The normal differences of Ki67 evaluated values obtained by one observer with method of the hot spot demarcating-square automatically counting is smaller than that obtained by same observer with method of the multi-spots demarcating-square automatically counting,which reveals that the dispersive degree of method of the hot spot demarcating-square automatically counting is better than that of method of multi-spots demarcating-square automatically counting.4.Method of the single hot spot demarcating-square automatically counting achieved ICC values of 0.964 and 0.977 respectively for set areas 200×200 ?m2 and 300×300?m2.The ICC values of the multi-spots demarcating-square automatically counting are 0.853 and 0.880 respectively for set areas 200×200?m2 and 300×300?m2.The method of hot spot demarcating-square manually counting achieved ICC values of 0.926 and 0.973 respectively for set areas 200×200?m2 and 300×300?m2.Conventional visual evaluation under microscopes could result in ICC values 0.881 and 0.815 respectively for hot spot demarcating-square and multi-spots demarcating-square.The conventional demarcating-square manually counting under microscopes could only produce ICC value 0.832.5.It can be obtained from the dispersion figure that the dispersive degree of the manually counting under microscope is greater than demarcating-square automatically counting.6.From the comparisons of the median values and normal differences,the dispersions of different methods are in such a descending order:Conventional hot spot manually counting under microscope,demarcating-square manually counting and demarcating-square automatically counting.Conclusions:1.The repeatabilities of all methods that were used to assess the breast cancer Ki67 immunohistochemistry results meet the requirements.The best repeatability was achieved with the method of the digital image analysis of the defined area 300×300 um2 of hot spots.2.The method of manually counting of defined areas of the hot spots can prevent the identifying error and counting error made by the computer with the method of digital image analysis.3.Because enormous quantities of the images have to be counted manually with human eyeballs under microscope,the process of this method is very tedious and exhaust.Error of repeatedly counting or miss-counting are likely to occur and unpreventable.Therefore,it is not recommended to the clinical pathologists or practitioners.