Artificial Intelligence To Spot Breast Cancer. FutureUniverseTV Presents Tech News.
Digital pathology, an emerging method of diagnosis, can be used in conjunction with the Ki-67 standard comparison card for breast cancer.
An automatic artificial intelligence analysis has been proposed as a potential method for evaluating Ki-67 in a number of studies. Based on the results of this study, it appears that artificial intelligence counting and gold standard counting are highly consistent for Ki-67LI.
As a result of artificial intelligence, the International Breast Cancer Working Group’s standards for cell counting have been met.
As well as the range of cell numbers that the working group recommends have been met too. All automatic Ki-67 evaluation methods provide significantly greater sensitivity and specificity than manual methods.
Artificial intelligence software, however, has some disadvantages as well.
In studies, it has been demonstrated that the automatic evaluation method is less accurate for identifying tumor cells than the visual method. There is a possibility that some lymphocytes expressing Ki-67 may be identified as tumor cells, particularly in lymphocyte-rich tumors. The Ki-67LI is therefore overestimated as a result.
In this study, Ki-67LI was determined to be lower using AI than by the other three methods of interpretation.
Since breast cancer cells are heterogeneous, artificial intelligence cannot completely identify each tumor cell or can misidentify them, for example, by identifying interstitial cells as tumor cells, or ignoring positive tumor cells with blurred outlines and lighter staining, which results in low Ki-67LI levels. As reported in the Maeda study, the average Ki-67LI for visual assessment was 22 and the average Ki-67LI for AI count was 20.4. To address this problem, a study has proposed a semi-automatic method for evaluating Ki-67LI by manually labeling immunostained tumor cells and negative tumor cells in order to determine the accurate proliferation index value and then counting the cells automatically.
It is determined by dividing the total number of immunolabeled positive cells by the total number of tumor cells.
The gold standard in this study combined with breast pathology experts and artificial intelligence counting, to reduce the risk of Ki-67LI being overestimated or underestimated by avoiding computer errors in identifying tumor cells and human interpretation errors resulting from visual fatigue. Since breast cancer tumor cells are heterogeneous, most Ki-67 immunohistochemical sections have hot spots (high value-added areas) and cold spots (low value-added areas), and the hot spots include tumor margins and central hot spots.
It is inevitable that Ki-67LI results will vary depending on the interpretation area selected.
The International Breast Cancer Ki-67 Working Group recommended that hot spots in Ki-67 immunohistochemical staining sections be included in the interpretation area if there are hot spots. Based on the four methods used for interpreting Ki-67 in breast cancer, two hot spots and two non-hot spots were selected in this experiment. This method overcomes the effects of the selection of different evaluation areas on the repeatability of Ki-67 immunohistochemical interpretation of breast cancer caused by the selection of different evaluation areas. The results of this study show that artificial intelligence counting of Ki-67LI in heterogeneous tumors can minimize the impact of tumor heterogeneity on Ki-67LI and increase the consistency among pathologists significantly.
The average Ki-67LI value for breast cancer interpretation using the SRC, AI, and gold standard differed from the gold standard in the difference test.
In this case, the interpretation area may have been the cause of the difference. The experimental method selected three regions for Ki-67 immunohistochemistry; therefore, a multi-region analysis can still be performed to determine the differences between each of these regions and the gold standard. AI software for breast cancer Ki-67LI standardizes the interpretation area (multi-area average method), which is important for clinical applications of Ki-67.
Ki-67 assessment was updated by the International Breast Cancer Ki-67 Working Group (IKWG)
They talk about the analytical validity and clinical application status of Ki-67 immunohistochemical detection in breast cancer tissues, and recommends standardized visual assessments. To summarize, pathologists must determine a standardized method for interpreting Ki-67 in breast cancer cases.
The interpretation of Ki-67 immunohistochemistry by artificial intelligence software is highly accurate and repeatable.
The interpretation of breast cancer Ki-67 is performed in some pathology laboratories where artificial intelligence software has not yet become widely used, in order to ensure repeatability of interpretation results and to save time and energy. The breast cancer Ki-67 standard comparison card is used to ensure repeatability of interpretation results. Thus, it is expected that Ki-67 standard comparison cards will serve as a reference method for interpreting Ki-67 immunohistochemical results of breast cancer on a daily basis.
I hope you found this article useful to you. Thanks for reading Artificial Intelligence To Spot Breast Cancer.