Currently, the classification system of NETs is based on the evaluation of proliferation with incremental intervals of mitosis and on the expression of Ki-67. Counts are generally expressed as the number of mitotic cells per mm2, ideally counted up to 10 mm2 to increase accuracy. Several studies have indicated that, due to the large differences between observers in the count, the validity and reproducibility of the Ki-67 index are clearly superior to those of the mitotic count . In theory, any cancer cell with detectable Ki-67 is proliferating and therefore Ki-67 has been accepted as a very sensitive proliferation marker. Many studies over the past decade have shown that the Ki-67 index is a valuable biomarker for the diagnosis, classification and stratification of cancer prognosis [2–5]
However, problems remain about how to accurately measure Ki-67 expression. Over the years, various methods have been developed to quantify Ki-67: manual estimation by a pathologist who counts in real time through the eyepiece of the microscope ; manual counting using a printed image acquired using a camera mounted on a microscope with × 20  magnification; computer-assisted quantification by digital image analysis (DIA) . A precise assessment of the Ki-67 proliferation index is required for a rigorous classification of NETs. Furthermore, the reproducibility of the Ki-67 evaluation can be influenced by several technical factors such as the type of sample (biopsy or fine needle aspiration cytology), the staining technique and the type of antibody used . Regardless of the method, there is a lack of consistency in the evaluation of Ki-67 expression when evaluating borderline values at the threshold for a classification change, both between G1 and G2 (range 2 to 5%) and between G2 and G3 (from 15% to > 20%). At first glance, DIA appears to be more reliable than "eyeballing" because it can significantly reduce intra- and inter-observer variability [9, 10]. In fact, several studies have indicated that computational methods such as "eyeball" estimates turned out to be the least reliable. However, DIA has other drawbacks, such as setting subjective thresholds for positive counts and a significant increase in the cost of the test, which often means a possible reduction in its broad applicability due to a lack of specialized machinery and trained personnel.
The threshold for positive count can change due to both the heterogeneity of Ki-67 expression within a tumor and the intensity of the staining that indicates that expression. In fact, it is known that the expression of Ki-67 constantly increases during the cell proliferation cycle, reaching a peak at G2 / M. Current standards suggest to include any discernible staining (nuclear or diffuse, weak or strong) in the evaluation of the expression of Ki-67 [11, 12], in accordance with the recommendations of the World Health Organization (WHO). As a consequence of these recommendations, the grading of the classification using Ki-67 is significantly higher than using the mitotic count. Indeed, up to one third of tumors have higher grades based on Ki-67 than classifications based on mitotic counts .
Another problem could be the possible inclusion, in the Ki-67 count, of non-tumor proliferating cells residing within the tumor sample such as intratumoral endothelial cells, underlying epithelium (e.g. glands, crypts, etc.) and lymphocytes. To minimize the number of false positives produced by the evaluation of Ki-67 expression, it is also important to distinguish negative tumor cells from stromal cells, which do not need to be counted. Stromal cells are typically smaller, spindle-shaped, and generally surround the clusters and nests of cancer cells. The inclusion of these Ki-67 positive non-neoplastic cells (false positives regarding neoplastic activity) can significantly increase the overall count and, on average, are sufficient to increase the classification of most turms by at least one degree. Considering that, in most cases, false positives are characterized by lighter coloration, it is important to reconsider the current guidelines on the evaluation of Ki-67. Some researchers have raised the idea that a more accurate assessment of cell proliferation can be obtained by using the intensity distribution degrees in the immunohistochemical stains of Ki-67 , rather than simply considering the expression of Ki-67 as an "on-off" binary switch: "on" during cell proliferation is "turned off" during quiescence and senescence.
Finally, it should be considered that the empirical observation of paraffin sections subjected to immunohistochemistry (IHC) for the determination of Ki-67 shows that typical / atypical mitoses are easily recognizable. Other positive nuclei are routinely identified without considering differences in Ki-67 immunoreactivity which are also easily distinguishable. Researchers have described signs other than Ki-67 patterns in normal and cancerous cells: spots or spots, loose or agglomerated grains, highly developed and homogeneous mitotic nuclei and chromosomes [14–21].
These studies have shown that the position of Ki-67 differs in the different phases of the cell cycle, in fact, during the interphase, Ki-67 is involved in the organization of heterochromatin and in the nucleolar periphery [22–24]. Its expression is already low at the beginning of the G1 phase represented by homogeneously distributed points that form increasingly larger and more irregular aggregates until they are found only in the nucleolus. These features have been observed with immunofluorescence techniques in HeLa cells and are indicative of the existence of two localization patterns of Ki-67, a homogeneous patch in the Early G1 phase (EG1) and a late phase nodule pattern (LG1). [19, 25]. In the S phase, the size of the Ki-67 point increases until it becomes dense granules which are distributed throughout the nucleus [20, 22, 26]. Ki-67 expression increases from the S phase onwards with a progressive increase during G2 and reaches its maximum in mitosis [27–30].
In the present work, therefore, we have decided to evaluate the expression of Ki-67 in GEP-NETs not as a binary state, but on the basis of the intensity of the coloring. We based our hypothesis on the fact that the Ki-67 protein is continuously degraded during G0 and G1 and the production steadily increases from the beginning of the S phase until the mitotic output, and thus the expression of Ki-67 in a single cell it is a function of the stage of the cell cycle in which the biopsy was performed, as well as its development trajectory: the longer a cell has spent dormant, the lower its Ki-67 level will be when it re-enters the cell cycle. Considering these biological mechanisms of cell differentiation, we argue that a simple Ki-67 score as positive or negative in a tumor biopsy can be oversimplification . Therefore, the main objective of this study was to develop and propose a new method for evaluating Ki-67 expression by focusing on the intensity of staining provided by DIA. Since we used both computer-aided quantification and researcher assessment of the characteristics of the stained area (both intensity and classical neuroendocrine morphology), we have referred to this method as semiquant DIA. To determine the proliferative activity of GEP-NETs, only the highly colored (near black) nuclei in the neoplastic tissue were considered actively proliferating, hence neoplastic cells in the G2/M phase of the cell cycle. We compared the tumor classification of GEP-NETs using currently used methods (binary state of Ki-67 expression) and our semiquant DIA assessment to assess the significance of grade changes, with the aim of improving the accuracy of the current classification and therefore to offer patients suffering from GEP-NETs more accurate therapeutic solutions. Our hypothesis was that the proposed semiquant assessment reduced the overall number of higher grade classifications (G2 and G3) by better reflecting the actual clinical data.