Sage Journals: Discover world-class research

Abstract

Counting mitotic figures (MF) in hematoxylin and eosin–stained histologic sections is an integral part of the diagnostic pathologist’s tumor evaluation. The mitotic count (MC) is used alone or as part of a grading scheme for assessment of prognosis and clinical decisions. Determining MCs is subjective, somewhat laborious, and has interobserver variation. Proposals for standardizing this parameter in the veterinary field are limited to terminology (use of the term MC) and area (MC is counted in an area measuring 2.37 mm²). Digital imaging techniques are now commonplace and widely used among veterinary pathologists, and field of view area can be easily calculated with digital imaging software. In addition to standardizing the methods of counting MF, the morphologic characteristics of MF and distinguishing atypical mitotic figures (AMF) versus mitotic-like figures (MLF) need to be defined. This article provides morphologic criteria for MF identification and for distinguishing normal phases of MF from AMF and MLF. Pertinent features of digital microscopy and application of computational pathology (CPATH) methods are discussed. Correct identification of MF will improve MC consistency, reproducibility, and accuracy obtained from manual (glass slide or whole-slide imaging) and CPATH approaches.

Keywords

mitotic figure mitotic count MC prophase prometaphase metaphase anaphase telophase computational pathology CPATH artificial intelligence AI whole slide image WSI pathology oncology

Counting mitotic figures (MF) in histologic preparations is the oldest and most widely used method to estimate cell proliferation in tumors. Although possibly not as precise and reproducible as other proliferation indices, the mitotic count (MC) is a rapid, inexpensive test that can be performed by any pathologist with a microscope and glass slide or with a computer and whole-slide image (WSI) technology. The results can be used alone or as part of a grading scheme to aid clinical decisions about prognosis and treatment. As indispensable as the MC is in the assessment of tumor behavior for many tumor types, this test has never been standardized in veterinary oncologic pathology. A standard area of 2.37 mm² has recently been proposed to resolve the inherent variability in utilizing 10 high power fields for the MC.⁴⁴ However, multiple studies have demonstrated disparity in reproducibility of the MC with interobserver agreement (κ-values) ranging between 0.08 to 0.77 even if standard area is used.^{13,20,22,27,28,45,51} Additional potential causes of inconsistency in the MC include variable identification of MF,^11,40,45,62 an inability to evaluate the complete tumor area,^7,11,34,57 and potentially pathologist fatigue from performing repetitive tasks. Disagreements of labeling structures as MF or mitotic-like figures (MLF) occurred with frequencies of 6% to 35%,⁴⁰ 10% to 46%,⁴⁵ 17%,^11,21,58,59 and 68%⁶² and can be improved by establishing clear criteria for MF morphology.

In order to advance our understanding of the relationship between tumor morphology and behavior and to ensure progress in oncologic pathology, it is essential to delineate how to perform a MC, which starts with clearly defining the morphology of MF. This article therefore provides standards for the identification of MF and AMF to help distinguish them from MLF (Table 1). Additionally, we present comparisons of the general features of WSI as compared to light microscopy for identification of MF as well as a discussion of the role that computational pathology (CPATH) may play in MF identification and MC determination.

Table 1.

Structure	Characteristics	Helpful features
Prometaphase MF	Dark aggregate (cluster or ring shape) with spikes/projections	May be difficult to differentiate from MLF; look for spikes
Metaphase MF	Dark aggregate (linear or ring shape) with spikes/projections	Common; ring form represents end-on (parallel) view of the chromosomes in relation to spindle; normal MF
Anaphase MF	Two separated aggregates variable distances apart; linear with spikes/projections	Common, can have morphologic overlap with telophase
Telophase MF	Two separated aggregates at opposite ends of the cell; cleavage furrow	Uncommon, may be very thin cytoplasm connection between 2 daughter cells; cell borders are less than 1 cell width apart
Tripolar AMF	More than 2 chromosome clusters (spindle poles) during any phase, appears as 3 or more linear plates	Common, AMF has been correlated with poorer prognosis in some human cancers
Asymmetric AMF	Unequal sizes of 2 chromosome clusters in anaphase/unequal sizes of metaphase plate	Less common form of AMF; unequal clusters may be influenced by sectioning artifacts
Chromosome bridging AMF	Chromosomes stretching from one anaphase pole to the other (touching both), occurs in anaphase or telophase	Morphologic overlap with chromosome lagging
Chromosome lagging AMF	Chromosomes not in contact with larger aggregate/linear cord; occurs in anaphase and telophase	Morphologic overlap with chromosome bridging
MLF/hard negative	One or more hyperchromatic bodies, smooth surfaces, no spikes/projections; may have eosinophilic cytoplasm or nuclear membrane	Morphologic overlap with some MF (prometaphase); 3 helpful features: surface contour, context, cytoplasm color

Abbreviations: MF, mitotic figure; AMF, atypical mitotic figure; MLF, mitotic-like figure.

Mitosis

Mitosis (karyokinesis) occurs during the M phase of the cell cycle, in which duplicated DNA is organized for equipartitioning.^21,42 The DNA of any complete genome is exceedingly longer than the cell diameter (in humans, almost 2 meters, with cells less than 1/100 000th of that size), translating to a herculean task in accurately segregating these long strands.⁴² Chromatin, a complex of DNA and protein, has 2 forms visible with light microscopy: heterochromatin (densely compact, staining as dark granules) and euchromatin (active, loosely coiled, and light staining). Nuclei with greater replicative activity have more euchromatin and open space between granules of heterochromatin. Heterochromatin along the nuclear membrane is what allows this membrane to be visualized microscopically. Although pairs of duplicated, compacted DNA are referred to as chromatids, we will refer to this nuclear material visible in cytologic and histologic preparations as chromosomes, since there is no definitive way to differentiate chromosomes from chromatids with light microscopy. The phases of normal mitosis include prophase, prometaphase, metaphase, anaphase, and telophase.^{17, 21,42,50,58,59} It is important to note that mitosis is a continuum, and that cells observed in histologic preparations may be arrested in transition between phases. The duration of each phase may vary between animal species and cell types. Both prophase and metaphase have been reported as the longest phases of mitosis; however, it is unclear whether this plays a role in how commonly identified these phases are in tumors viewed by light microscopy.^17,26,31,39

Morphologies of MF in Histology and Cytology

Mitosis is the process of karyokinesis while MF are the structures that can be identified with light microscopy. MF are defined as nuclear aggregates containing short rods or spikes of chromosomes (first described by van Diest as hairy projections),⁵⁹ unbound by a nuclear membrane after prophase (the first phase of mitosis).^34,59 Depending on the phase of mitosis, the morphologies of MF vary. While identification of the phases of mitosis is not necessary to perform MCs, general knowledge of the phases is useful when differentiating normal MF and AMF from MLF.

Prophase

At the first stage of mitosis, the chromosomes are compacted via reduced length and increased thickness, facilitated by proteins called condensins.⁴² In prophase, a microtubule cytoskeletal spindle (rarely visible with light microscopy) forms, and attaches to and guides the chromosomes during the subsequent phases.⁴² Prophase has 3 subphases (early, middle, and late). Histologically, early to mid-prophase is difficult to identify with light microscopy, as it is composed of chromosomes that are loosely compacted (not densely aggregated) into rod-shaped forms best seen in cytologic samples with preservation of the nuclear shape and membrane^42,50,58 (Fig. 1). These early subphases are difficult or impossible to reliably distinguish from MLF due to the similarity of the condensed nuclear chromosomes (which lack rods or spikes in prophase) to condensed, hyperchromatic, or pyknotic nuclei (Figs. 1, 18, 20, 21, 24, 26).^11,40,45,57 A pathologist or CPATH is not likely to reliably and consistently recognize early to mid-prophase because the characteristics that define this portion of prophase are too subtle when evaluating histologic preparations with light microscopy. The nuclear membrane is present and chromosomes are condensed uniformly, appearing as dark aggregates but without the easily visible dark clusters of chromosomes with rods or spikes seen in later phases. Therefore, we recommend that early to mid-prophase should not be enumerated in the MC as it cannot be reliably distinguished from MLF. During late prophase to prometaphase, condensed aggregating chromosomes are compacted^57,59 and chromosome spikes start to project from the cluster.

Figures 1–16.

Cytologic appearance (Figs. 1–4 and 9–12; Romanowsky) and histologic appearance (Figs. 5–8 and 13–16; hematoxylin and eosin) of mitotic figures (MF) from various canine tumors. Figures 1–8. Normal mitotic figures. Figure 1. Prophase: Condensed and duplicated chromosomes with no projections of individual chromosomes. This phase can be recognized cytologically but is difficult to recognize in histologic sections due to the lack of a dark cluster of chromosomes with rods/spikes. Figure 5. Prometaphase: Dark, typically round cluster of chromosomes with short rods/spikes or projections. A ring or rosette may also be visible. Differentiation from mitotic-like figures (MLF; ie, dark, hyperchromatic nuclei) in histologic specimens is necessary. Figures 2, 6, 53. Metaphase: A plate, band, or ring of chromosomes with easily visible rods or spikes (one of the most common and easily recognizable phases). Insets (Figures 2,6) : The ring form or rosette/starburst MF, when the metaphase plate of chromosomes is parallel to the spindle. This normal form has been misinterpreted as atypical. See also Figures 53 and 54. Figures 3, 7. Anaphase: The duplicated chromosomes formed during mitosis are separated by variable distances. The aggregates have visible irregular rods, spikes, or projections. Figures 4, 8. Telophase: Two genomes at opposite ends of the cell with formation of a nuclear envelope and cleavage furrows (separating the 2 daughter cells). This is uncommon in cytologic and histologic preparations. Figures 9–16. Atypical mitotic figures. Figures 9, 13. Multipolar MF: More than 2 plates of chromosomes during any stage of mitosis. Figure 9. Tripolar MF with 3 plates of chromosomes and 4 to 5 lagging chromosomes. Figure 13. Tripolar AMF. Figures 10, 14. Polar asymmetry in anaphase: Unequal chromosomes in the 2 anaphase clusters. Figures 11, 12, 15, 16. Bridging chromosomes stretch from one anaphase pole to the other; lagging chromosomes or chromosome fragments are short segments of chromosomes not in contact with the dark aggregate of chromosomes, seen in anaphase or telophase. Figure 15. Interpreted as anaphase bridging, or possibly a MLF. Figures 12, 16. Anaphase with lagging chromosome. These are seen easily in the cytologic preparation but in the histology image the lagging chromosomes are inconspicuous (arrow). Figures 1–4 and 9–12. cytology (Romanowsky), reprinted with permission.⁵⁸

Prometaphase and Metaphase

Prometaphase is characterized by a dark cluster of chromosomes with visibly protruding chromosome rods and spikes. During late prometaphase, the chromosomes are briefly arranged into a single, wreath-shaped ring known as the chromosome rosette.⁴⁶ Metaphase is denoted by alignment of the chromosomes at the “equator” and can be visualized in histologic and cytologic samples as a linear plate, band or ring shape.^17,21,42 This dark cluster has rods or spikes of chromosomes projecting from the surface that are seen easily with light microscopy at 400× final magnification and which are the most definitive way to identify MF and distinguish them from MLF (Figs. 2, 6). MLF such as apoptotic nuclei consist of a homogenous mass of degraded DNA with a smooth outer contour (Figs. 18, 19, 20, 21, 22, 24 –26).³⁴

Figures 17–26.

Histologic appearance of mitotic figures (MF), atypical mitotic figures (AMF), and mitotic-like figures (MLF) from various canine tumors. Hematoxylin and eosin. Figure 17. Multipolar AMF with more than 3 metaphase plates. Figure 18. Prometaphase/metaphase MF (upper left) has distinct chromosome rods or spikes, but a pyknotic nucleus (arrow, apoptosis, or necrosis) has a smooth contour and eosinophilic cytoplasm. Figure 19. MLF (apoptosis or necrosis) with condensed nuclear chromatin at opposite poles of the nucleus (anaphase imposter). Figure 20. The pyknotic lobes of neutrophil nuclei (2 horizontal arrows) appear as dark spheres (MLF) that resemble a multipolar AMF. An apoptotic or necrotic cell has condensed chromatin and a smooth contour (vertical arrow). Figure 21. Multiple pyknotic nuclei (arrows). A metaphase MF (lower right) has a plate of rod-shaped chromosome projections. Figure 22. Karyomegalic nucleus with vesicular chromatin (left) beside an oval pyknotic nucleus with a smooth contour (metaphase imposter, arrow). Figure 23. Eosinophil (arrow) with a pyknotic, lobulated nucleus in a subcutaneous mast cell tumor. Figure 24. Cell with karyorrhexis. Multiple, pyknotic nuclear fragments have smooth contours (arrow). Figure 25. Two MLF appear as pyknotic, elongated, hyperchromatic nuclei (vertical arrows). A MF (horizontal arrow) could be a normal ring-form (starburst) metaphase MF viewed in a slightly unusual plane, or it may represent the start of anaphase or an AMF. Regardless, it is a definitive MF or AMF because of its density and spikes. Figure 26. Large cell with a MLF appearing as multiple smooth, round bodies of condensed chromatin (probable apoptosis/pyknosis). MLF nuclear remnants have smooth contours with no distinct chromosome spikes. Note the adjacent cells and nuclei are poorly preserved because this area of the tumor was necrotic. The interpretation of structures is influenced by the appearance of the adjacent tissue.

This contour of smooth vs irregular surfaces with short spikes should be used by pathologists to aid identification of MF with manual (glass slides, WSI) and CPATH approaches. Depending on the orientation of the cell, these stacks of chromosomes appear either perpendicular or parallel to the spindle apparatus, which alters their appearance in histologic sections.⁵⁸ If the metaphase plate is perpendicular to the spindle, the dark aggregate of chromosomes is linear with projecting spikes (Figs. 2, 6, 53). If metaphase is viewed as “head on” or parallel to the spindle apparatus, the dark cluster of chromosomes has a circular ring shape and is referred to as “starburst” or “rosette,” which has been incorrectly interpreted as AMF (insets, Figs. 2, 6, Fig. 54).⁵⁸

Figures 27–50.

Detection and classification results of a dual-stage deep learning-based algorithm for mitotic figures in digital images of canine cutaneous mast cell tumors.⁷ Detections (first stage) are labeled by boxes. Classification (second stage) into MF or MLF is indicated by color: Numbers below the boxes (model scores) as well as colors of the boxes indicate how likely a structure is interpreted to be MF (dark green, score ≥0.5, Figs. 27–38) or MLF (light green, score <0.5, Figs. 39–50). The pathologist-defined ground truth labels (MF, as defined by 5 board-certified veterinary pathologists) are indicated by blue circles. Undetected structures do not have boxes. Figures are obtained from WSI (resolution 0.25 µm per pixel or approximately 400× magnification). Figures 27–32. True-positive algorithmic detection (MF). The majority of MF in this row are easily identifiable dark clusters with chromatin spikes and have model scores >0.5. MF morphology is consistent with prometaphase (Fig. 27), anaphase (Fig. 28), metaphase (Fig. 29), ring form of prometaphase/metaphase (Fig. 30), telophase (Fig. 31), and a tripolar AMF (Fig. 32). Note that the telophase MF (Fig. 31) was detected twice by the algorithm, and annotated once by the pathologist, thereby necessitating correction via pathologist review. Figures 33–38. False-positive algorithmic detection (MLF) with scores > 0.5. None were annotated as MF by pathologists. MLF in this row are generally blurred and hyperchromatic with rounded margins. Figure 37. The dispersed chromatin was detected as 2 MF. Figures 39–44. True-negative algorithmic detection (MLF) with scores <0.5 or no scores. MLF in this row are composed of hyperchromatic nuclear material with rounded margins. Figure 43. Eosinophilic cytoplasm and pyknotic nuclear material suggests cell death (MLF) rather than a MF. Figure 44. is a MLF that appears similar to a tripolar AMF. Figure 33 (false-positive MLF) and Figure 39 (true-negative MLF) have a similar appearance with a blurred, dark chromatin aggregate in the center of the cell. Given that the algorithm is derived from a deep learning CNN network, the reason for the disparate scores between these 2 similar images is unknown. Figures 45–50. False negative algorithmic detection (MF) with scores <0.5 or no scores. These MF were misidentified as MLF, with model scores close to the center of the range (0.5) or were undetected. MF morphology is consistent with prometaphase (Fig. 45), anaphase (Fig. 46), metaphase (Fig. 47), ring or rosette form of prometaphase/metaphase (Fig. 48), anaphase/telophase (Fig. 49), and tripolar AMF (Fig. 50). Note that the columns are arranged by shared morphology.

Anaphase

Anaphase, also seen easily in histologic and cytologic preparations, occurs when the single plate or ring of chromosomes from metaphase splits into 2 equal clusters of sister chromosomes. Each cluster separates and is pulled apart along the microtubules of the spindle apparatus^17,21,50,58 (Figs. 3, 7). The 2 clusters of chromosomes will be variable distances from one another until they reach the opposite ends of the cell. It is important to note that there is a continuum between anaphase and telophase as the chromosome clusters reach the ends of the cell. These differences (distance between chromosome clusters) are relatively inconsequential for the pathologist but may have importance for CPATH applications when delineating phases of mitosis or counting the clusters as 1 or 2 MF.

Telophase

Telophase is characterized by movement of 2 chromosome clusters to opposite ends of the cell, formation of cleavage furrows (separation of the 2 daughter cells) and nuclear envelopes, and the emergence of nucleoli^50,58 (Figs. 4, 8). Formation of the nuclear membranes and nucleoli cannot be seen consistently with histopathology or cytopathology in glass slides or WSI (Figs. 4, 8, 31). In some cases, cells in telophase may be connected by a thin strand of cytoplasm, which may or may not be in a plane visible to the pathologist (examples in Supplemental Figures S49, S54, S55). Cytokinesis is the physical separation of cytoplasmic components into 2 daughter cells and occurs after telophase. The lack of cytokinesis can produce multinucleate cells.²¹

Consensus of MF Morphology

If a pathologist is confident that a structure represents a MF or an AMF, it is counted regardless of the exact phase of mitosis or type of AMF. Fortunately for pathologists, it is not required that the normal or atypical phases of mitosis are specifically identified (or agreed upon!) in order to obtain a MC. However, the characteristics of and therefore the differentiation between MF, AMF, and MLF are important so that use of microscopes, WSI, and CPATH provide the same MC. If the morphologic criteria for these structures are adhered to as described above, highly reproducible MF identification is possible. Herein, we provide expert consensus on aspects that have been performed inconsistently in previous studies and require standardization.

Prophase

In order to provide consensus on the identification of MF, the authors reviewed hundreds of photomicrographs of potential prophase MF in different tumor types. Most structures were MF with a circular cluster of chromatin and inconspicuous rods suggestive of late prophase, prometaphase, or a transition between prometaphase and metaphase^{8,34,46,57
–59} (Figs. 5, 18). Other examples had a clear space within the central nuclear aggregate which likely represents a nonperpendicular view of the prometaphase/metaphase plate (rosette, starburst;⁵⁸ Figs. 2, 6, 54). Some structures displayed a condensed, round basophilic nucleus which could not be reliably differentiated from a pyknotic nucleus, autolyzed nucleus, or other types of MLF (Figs. 18, 20, 21, 33, 39). Due to discrepancies in how pathologists interpret prophase, Meyer et al recommended that prophase be excluded from mitotic counts performed for grading of human breast cancer, and the authors of the current manuscript agree with this recommendation.⁴⁵ These images and comments exemplify the fact that there will be structures that must be interpreted carefully. Fine focusing and examination of the immediately adjacent tissue may influence interpretation. For example, if the doubtful structure is close to an area of necrosis, MLF may be a more correct interpretation. If the structure in question were close to a “hot spot” (area of high mitotic activity within a tumor) and there is good tissue preservation, then a MF may be more likely. The final interpretation when using adjacent tissues is an example of contextual reasoning, which is an area of development in CPATH. The problems posed by attempting to accurately identify early stages of prophase warrant restricting MF enumeration to the stages of mitosis which can be reliably distinguished on the basis of identification of chromosome clusters with short rods/spikes. These include prometaphase, metaphase, anaphase, and telophase (Table 1).

Figure 5 is our best attempt at finding an example of prometaphase in histologic sections, with additional examples in Table 1 and Supplemental Figures S1–S16. Despite reviewing hundreds of examples of MF in histologic sections and images, we could not find a suitable histologic image to represent prophase. MF previously characterized as prophase are likely to be more accurately identified as prometaphase or MLF.

Early studies recommended that the MC includes only MF at the metaphase stage or later (ie, definitive MF), presumably due to the difficulty of recognizing prophase MF.^43,59 Given the paucity of information on confirming early to mid-prophase in histologic sections, future studies may reveal an alternate method of recognizing prophase. For example, immunohistochemistry (IHC) for the phosphorylated form of histone 3 (PHH3) is considered to be a reliable marker of mitosis and allows differentiation of MF from MLF.^4,9,55 This immunohistochemical marker is present from early prophase to telophase and must highlight a visible MF in order to be counted as positive.⁴ The increased contrast facilitates detection and thus may further aid pathologists and CPATH in recognizing MF,^9,55 but it would have to be correlated with outcome assessments in order to determine if the resulting data are predictive of tumor behavior.

Telophase

Telophase MF should be counted as one MF (Table 1, Figs. 4, 8, 31, Supplemental Figures S45–S56). Although it was originally recommended to count the 2 chromosome clusters of telophase as 2 separate MF,⁵⁹ we propose to count the 2 chromosome aggregates of telophase as a single MF. The original intent of counting telophase as 2 MF was to correlate manual counts with those made by future automated systems, as exemplified in Figure 31. However, CPATH systems can be trained to recognize a telophase figure as a single MF and this should represent the standard for telophase enumeration. A pathologist can look for cleavage between the 2 aggregates, and how close these aggregates are to one another, and thereby deduce that they were derived from one cell (ie, one MF). Alternatively, a pathologist may assess the 2 chromosome clusters as dissimilar and too far apart to be from the same MF, therefore counting them as two, perhaps both in metaphase. Rarely, cells in telophase appear to have undergone cytokinesis (although they may be still attached in a plane not visible to the pathologist). If these MF arise from one cell, they should be counted as one. We propose that the cell borders be a minimum distance of at least one tumor cell width apart (therefore the center of the cells are at least 2 tumor cell widths apart) in order to count 2 MF. Distances less than one tumor cell width apart should be counted as 1 MF, even if a clear connection between cells cannot be appreciated for reasons stated above. CPATH can produce similar results if the initial training and ongoing validation process adheres to these morphologic definitions for telophase. CPATH programs may use color (or color depth), shape (morphology), surface irregularities (texture), intensity (darkness), size, distance between clusters (µm or pixels), or other unknown characteristics (“black box”)^1,63 to determine if a structure represents a MF and if it is 1 or 2 MF. In a computer-assisted approach for obtaining MCs, the pathologist can modify the pixel size or diameter of the annotation, review candidate MF, and discard doubtful structures.

Morphologies of AMF in Histology and Cytology

Chromosome segregation errors during cell division can cause genetic abnormalities in neoplastic cells.^24,30 These abnormalities can be observed morphologically as AMF and include 2 broad categories: (1) mitotic/polar asymmetry and (2) abnormal segregation of chromosomes. More complex aberrations are derived from combinations of these categories.³⁰

AMF in cytologic preparations were described by Tvedten as a reference for veterinary clinical pathologists to improve accuracy and consistency in identification and classification of AMF (Figs. 9–12).⁵⁸ These abnormalities are also appreciated in histologic preparations but can be less obvious due to lower magnification and greater thickness of preparations. Cytologic samples are often evaluated with oil immersion at 1000× magnification, while histopathology samples are thin slices (3–6 µm thickness) of a formalin-fixed, paraffin-embedded block and are typically evaluated or scanned at 200× to 400× magnification. Compare the similar AMF in cytologic preparations (Figs. 9–12) and histologic preparations (Figs. 13–16). For instance, Figure 11 is clearly anaphase but the separation is not distinct and there are chromosomes bridging from one cluster to the other. These features are easier to see in the cytologic preparation but are not as well visualized histologically (see Fig. 15, interpreted as anaphase bridging). In Figure 12, lagging chromosomes can be easily seen in the center of the cell in the cytologic preparation. This is less conspicuous in the histologic preparation, appearing as a small, thin purple streak in the center of the cell (Fig. 16, interpreted as lagging chromosomes). Nonetheless, AMF are readily identifiable in histologic sections and if possible, should be classified according to the morphologic features described herein.

AMF With Mitotic/Polar Asymmetry

Abnormalities of mitotic/polar symmetry are composed of multipolar MF and asymmetrical bipolar MF.³⁰ Multipolar MF (frequently tripolar MF) are defined by the presence of more than 2 spindle poles during any stage of mitosis,³⁰ resulting in movement of the chromosomal content in three or more directions at anaphase. This abnormality is easily appreciable from metaphase through telophase (Figs. 9, 13, 17, 32, 50 Supplemental Figures S57-S59, S62, S64, S65, S67). If these cells survive through anaphase, the daughter cells are likely to contain different numbers of chromosomes as compared to the mother cell due to random redistribution.^30,36 Toxins that affect the mitotic spindle have been shown to create tripolar mitoses, but it is unclear whether multipolar mitoses are the cause or consequence of cancer.³⁶

Asymmetrical bipolar MF are described as unequal sizes of the metaphase axes or anaphase poles³⁰ (Figs. 10, 14). Oblique sectioning artifacts in normal or multipolar cells can sometimes artifactually create the appearance of asymmetry.³⁰

AMF With Abnormalities of Chromosome Segregation

The second broad category, abnormalities of chromosome segregation include anaphase bridging, lagging chromosomes, and lagging chromosome fragments.³⁰ Anaphase bridges are defined as the stretching of chromosomes connecting one anaphase pole cluster to the other³⁰ (Figs. 11, 15). Lagging chromosomes or chromosome fragments are left behind the other chromosomes at anaphase and telophase, but there is no standard criterion for this distance (Figs. 9, 12, 16). Differentiation between a lagging chromosome and a chromosome fragment is difficult or impossible in histologic samples. Chromosome lagging has been associated with multiple abnormalities in cellular signaling systems.³⁰ Chromosome segregation disorders have been linked to chromosomal instability and telomere dysfunction in some human cancers, which contributes to carcinogenesis via genomic alterations including oncogene amplification and tumor suppressor gene deletion.³⁵

Significance of AMF

AMF have been correlated with poorer prognosis and outcome for some tumors in humans.^35,36,41,47 In those studies, AMF included MF with multipolarity, lagging chromosomes, ring morphology, asymmetry, and anaphase bridging. In the authors’ opinion, ring MF should not be included in the category of AMF as they are more likely to reflect a prometaphase/metaphase rosette in which the chromosomes are seen parallel to the spindle⁵⁸ (Figs. 2, 6 inset, Fig. 54). Future studies that have accurate and robust outcome assessments should consider evaluating if AMF counts and types of AMF have any prognostic significance for tumors of veterinary patients. Enumerating AMF will be possible but carefully identifying the specific types of AMF may require evaluation at high magnifications (greater than 400×).

Figures 51–52.

Possible computer-assisted system for determination of mitotic count in whole-slide images of canine cutaneous mast cell tumors.⁷ Figure 51. In a complete tissue section, the black grid shows the tumor area (2.37 mm²) with the highest mitotic density. Figure 52. Algorithmic detections are classified as MF (dark green boxes; 7 in the Figure) or MLF (light green box; 1 in the lower right corner), and definitive confirmation requires verification by a pathologist.

Figures 53–54.

Three-dimensional representation of a cell undergoing mitosis at metaphase. Figure 53. The chromosomes are viewed perpendicular to the mitotic spindle, such that the aggregate with rod-like projections is linear. Figure 54. The chromosome cluster is viewed parallel to the mitotic spindle, such that the aggregate with rod-like projections is circular (rosette or starburst).

Mitotic-Like Figures

MLF are structures or processes that appear similar to MF, and include apoptotic bodies, hyperchromatic nuclei, deformed nuclei, karyorrhectic debris, inflammatory cells, and tissue artifacts^3,8 (Figs. 18–26). MLF are characterized by smooth contours and are generally circular or oval and dark (hyperchromatic) bodies that can be single or multiple in an area where a nucleus should be located (Figs. 18–26). Light microscopic criteria useful for identification of MF/AMF and distinguishing them from MLF include the absence of a nuclear membrane, an irregular surface (presence of short rods, spikes or projections), and amphophilic or basophilic cytoplasm.⁸

Distinguishing MF and AMF from MLF is one of the most critical and error-prone steps in determining the MC, for both humans and computers. Figures 18 to 26 and Figures 33 to 44 contain examples of MLF (imposter MF), some of which can be difficult to distinguish from MF. Making this distinction requires thorough examination via fine focusing, and more importantly, considering the context of the adjacent tissue around these structures. For example, a structure adjacent to an area of necrosis (with multiple cells undergoing nuclear pyknosis or karyorrhexis) is more likely to be a MLF, whereas a single dark structure with “spikes” on the surface, especially in a region of mitotic activity favors a MF.^3,60,61 Cells undergoing degeneration, necrosis or apoptosis should ideally have an eosinophilic cytoplasm, whereas cells containing MF are expected to have more basophilic cytoplasm; however, exceptions to this rule are numerous and there is variability between different staining protocols. To summarize, 3 helpful features to aid in differentiation of MLF from MF are surface contour, evaluation of the adjacent tissue (i.e., the tissue context), and color of the cytoplasm. MF are incorrectly differentiated from MLF because of a lack of defined criteria or adherence to criteria, proceeding too rapidly, not using high enough magnification, and artifacts due to fixation, staining, and scanning.^3,8,9,62 Each of these can contribute to interpathologist variation in performing the MC. Careful and methodical evaluation is important for accurate classification of these structures. Candidate MF that cannot be reliably distinguished from MLF at high magnification (400× or higher) in a reasonable time frame should not be included in the MC.

It is inevitable that cells or structures may be difficult to classify when subjectively assessing biological specimens. This contention is well captured in a quotation by Dr Irwin Leav, “One man’s protozoan is another man’s necrotic nucleus.” Minimizing interpathologist (and intrapathologist) variation is important in reducing the frequency of false positives and false negatives. To mitigate these variables, only definitive MF and AMF should be enumerated as part of the MC and doubtful structures should not be included.

Factors That Affect MF Identification and the MC

Slide Preparation

MF may appear differently based on multiple factors involved in slide preparation, including staining technique, thickness of the section, and type of preparation being evaluated by the pathologist (glass slide or digitally scanned WSI).^{8,11,13,14,61,62} In cytologic preparations, the whole cell is visible in one focal plane. However, in histologic preparations the cells have been transected and only a portion of a MF may be seen, which can interfere with recognition. Distinguishing MF from MLF is thus easier with cytologic preparations than histologic sections or WSI due to the ability to appreciate greater cellular detail at 1000× magnification and the ability to evaluate the whole cell. Enumerating MF has become a part of grading dermal mast cell tumors cytologically.^19,33,52

Cytoplasmic basophilia or eosinophilia should be used carefully to differentiate MF versus MLF as variation in staining methods between laboratories can produce color changes in tissue sections. Eosinophilic cytoplasm may suggest necrotic cells because of denatured cytoplasmic proteins and loss of mRNA (Figs. 18, 24, 43).⁵⁶ However, variations in the concentrations of eosin and hematoxylin can affect these colors as can quality of fixation. Examples are provided of MF, AMF, and MLF with eosinophilic cytoplasm (Figs. 13, 18, 21, 24, 25 26, 43) and with amphophilic cytoplasm (Figs. 5, 8, 16, 17, 20). Amphophilia and basophilia are expected with increased mRNA on ribosomes of neoplastic or young cells.

Fixation Delay

Fixation delay has been reported to correlate with changes to the MC; however, studies have reached different conclusions and the significance is not clear.^{9,10,23,25,38,49} In larger tissues that have not undergone complete formalin fixation, cells remain capable of proliferation and can progress through the phases of mitosis until the tissues are fixed or cells exhaust their energy reserves, thereby artifactually increasing the numbers of observable MF.^25,38,49 One study comparing resected breast carcinomas to incisional biopsies of the same tumors found an increase in MF of up to 3-fold in the samples from the resected breast. This difference was so great that it resulted in increased tumor grades in approximately 25% of carcinomas, thus having an effect on prognosis and treatment plan.³⁸ This phenomenon has also been reported to occur in samples obtained from a canine hepatocellular adenoma. Fixation delay within a needle biopsy resulted in substantially increased numbers of mitotic figures as compared to the resected neoplasm that was immediately placed in formalin.⁴⁹ Alternatively, decreased numbers of MF was caused by reduced tissue quality from delayed fixation which led to an inability to differentiate between MF and MLF.^23,25 Other studies have shown no appreciable influence of fixation delay on the MC.^10,23,25 At this time, there is no consistent correlation between fixation delay and MC.

MF Identification in Whole-Slide Images (WSI)

Pros and Cons of WSI for MF Detection

WSI are widely used for histologic diagnoses and teaching in both veterinary and human pathology. WSI permits remote primary diagnosis, ergonomic benefits, economic advantages, annotation of structures, accurate measurement of areas/fields of interest (such as 2.37 mm²), and application of CPATH methodology.¹⁴ Additionally, duplicate counting of structures or fields is prevented, and doubtful structures can be reviewed with distant colleagues. Assessment of MF, AMF, and MLF in WSI poses some challenges due to the lack of fine focus capability and limited image resolution in routinely scanned images. Although multiple focal planes (z axis) can be captured to simulate fine focusing, and high scanning magnification is technically possible, the scan time and storage space required is impractical in the current typical diagnostic setting.^{3,11,13,14,54,61,64}

WSI Validation for Diagnosis and MC

There is generally good agreement between WSI and light microscopy (LM) with regard to cytologic and histologic diagnosis and MC in animal and human tumors.^{2,3,13,16,18,29,32,48,53,64} A better correlation was observed in cases with higher MCs as compared to those with lower MCs.³ Some studies reported a trend toward underestimating the MC with WSI as compared to LM and attributed this to difficulty of identifying MF via WSI.^3,32,53 Grade assignments via LM or WSI has not yielded consistent results.^13,32,54 One study found that the modality used (LM vs WSI) did not have significant effects on tumor grading derived from the MC for breast cancer in humans,³ while another study found that the grades obtained by using MC derived from WSI versus LM were different for sarcomas and neuroendocrine tumors in humans.³² Grading canine cutaneous MCT with a 2-tier system using WSI was inferior as compared to LM; however, pathologists lacked long-term experience with WSI.¹³

Counting MF with WSI as compared to LM has been reported to be more time consuming^3,32; however, as with any new task, this time factor eventually becomes negligible. Software modifications including grid applications (Figure 51), computerized preselection of regions with high mitotic activity, more efficient annotation, and field of view auto-sizing may further reduce the time required for obtaining the MC with WSI.^3,11,62 The diagonal size and resolution of the monitor, as well as the on-screen magnification at which the images are evaluated, influences how many screen areas must be counted in order to obtain a MC.^13,44 Comparisons of digital and microscopic high-power field areas revealed significant variations between different display resolutions and monitor sizes with minimal differences between WSI viewers and no differences between scanners.³⁷ Standardization of minimum monitor size, resolution, and use of standard color calibration may further reduce variability between pathologists when counting MF with WSI.^37,64

Limited resolution of the WSI does not seem to influence the overall MC significantly. Most studies comparing LM and WSI obtained MC using 400× magnification images (image resolution: 0.25 μm/pixel).^3,13,32,64 For MCs obtained from canine melanomas, WSI at 400× or 200× magnification had similar inter- and intraobserver concordance with one another and with LM.⁶⁴ Based on published information as well as personal experiences among the authors, diagnostic evaluation and MC can be readily performed with WSI.

Classifying Individual MF With WSI

Comparisons between WSI and LM in terms of identifying individual structures (including MF) has revealed conflicting results.^54,61 Comparing the ability to identify or classify individual structures as MF or MLF with LM and WSI is technically challenging.^54,61 Tabata et al⁵⁴ used a light microscope mounted with a camera and specific software to compare LM to WSI from 4 different scanners. Inter- and intraobserver agreement was substantial for all scanners; however, accuracy of MF identification compared to the pathologist-derived ground truth was significantly lower with WSI in 3 of the 4 scanners. WSI yielded lower numbers of MF compared to LM. In a study of human breast cancer that analyzed object-level interobserver agreement of MF classification using WSI, disagreements between pathologists occurred more frequently with smaller objects (average length of 8.7 μm).⁶¹ This trend was attributed to difficulties in spotting these smaller objects during scanning and more difficulty in appreciating distinguishing features like protruding chromosome rods or spikes in smaller objects.⁶¹ In order to counteract this limitation, using a minimum size for acceptance of candidate structures as MF may improve reproducibility, but would vary between tumor types.^59,61 Structures that are difficult to identify with LM will likely remain difficult to identify with WSI.

The Future: CPATH Recognition of MF, AMF, and MLF

CPATH is defined by the Digital Pathology Association as a “big-data” approach to pathology, in which a combination of image data and meta-data are used to develop digital and automatic analysis of features and patterns in WSI.¹ Machine learning and, more recently, deep learning (a subset of machine learning) are examples of artificial intelligence (AI) that encompass computational techniques and are key elements of CPATH.¹ Machine learning employs the relationship between input data (in the case of WSI, data from pixels that form the scanned image and patterns extracted from the image), and desired output information (such as an image label or detection of objects and shapes within the image).

Reasons to Use CPATH for MF Identification

Imperfect interobserver and intraobserver reproducibility when obtaining the MC can result from (1) difficulty of differentiating MF from MLF (human error); (2) an inability to efficiently locate the region of highest mitotic density; and (3) impaired reproducibility of MF identification or classification while performing repetitive tasks.^{7,11,13,20,22,27,28,34,35,40,45,51,57,65} With sufficient numbers of images, humans, and computer programs can be “trained” to differentiate MF, AMF, and MLF. Algorithms can assess the entire slide or several slides within a short period of time, are 100% reproducible, and (with good data sets and deep learning methods) they can be as accurate as pathologists. Therefore, it can be hypothesized that an indefatigable, fully reproducible computer algorithm may be a viable potential alternative or complementary method for identifying MF when performing the MC in histologic samples.

Recent CPATH Progress in Veterinary and Human Pathology

Recent progress in methods for the training of algorithms using big data may now open the possibility of using CPATH for automatically determining the MC in routine veterinary diagnostics. However, it is in the early stages of consideration in veterinary medicine^5
–7,12 and still requires further research and validation before being incorporated into conventional diagnostic services. CPATH approaches have been applied to automate MC calculation in digital images in order to compensate for human shortcomings and will likely increase pathologists’ efficiency, reproducibility, and accuracy. Current groundbreaking results of CPATH for MF identification have been achieved with convolutional neuronal networks (CNN; a type of deep neural network), which function autonomously to identify MF in digital histologic images. These networks consist of multiple layers of artificial neurons that are connected in a complex way. Through pathologist-defined labels that act as “ground truth” (eg, MF or not MF), the CNN autonomously extracts discriminative features from the images during supervised training of deep learning-based algorithms to produce user-desired outputs. During this learning process, the connections are modified with supervised cycles of positive and negative feedback input based on “correct” or “incorrect” as compared to the ground truth. The computer “sees” images differently from a human and has complex decision-making criteria (autonomous analytic and approach pathways), which are commonly too complicated to be understood by algorithm developers (the so-called “black box”). In some cases, the methods to better understand output information, such as visualization of a detected pattern, can be used in which case the black box becomes a “glass box.”¹ An analogy would be the comparison of decision making between an experienced pathologist and a trainee. The experienced pathologist may process the image, slide, or case by rapidly incorporating several steps simultaneously in order to make a diagnosis. When this decision tree is not explainable to the trainee, this is analogous to a “black box.” Alternatively, when the steps are able to be broken down and explained, this is analogous to a “glass box.” Deep learning methods, such as CNNs, have only recently become available due to the high computing capacity and large data sets required.^{5,6,12,15,60,62} These data sets contain numerous images (WSI or parts thereof) and thousands to tens of thousands of labels which include structures to be detected (MF) and structures that should not be detected (MLF). The latter (structures that should not be detected) are not required, but may aid algorithm performance.^{5,6,12,15,60,62}

Publicly available data sets of MF from human and animal tumors have tremendously fostered the development of deep learning methods.^12,15,63 Bertram and colleagues published a large-scale data set which contains an unprecedented 44 880 annotated MF and 27 965 annotated MLF (hard negatives) in canine mast cell tumors.¹² A deep learning–based algorithm for automatic MF detection yielded an F1 score of 0.820, indicating high recognition of MF on the test set (a combination of precision and recall, where a value of 1.00 indicates perfect recognition), which is close to the values found for human interobserver variability.¹² Similarly, another data set for canine mammary carcinoma, which included 13 907 annotated MF and 36 379 annotated MLF,^5,15 yielded an F1 score of 0.79 when compared to the ground truth based on a human observer. With this and other current deep learning approaches, algorithms have shown substantial agreement with pathologists.^5,15,61

CPATH Limitations

Major limitations in current CPATH solutions are the lack of “real-life” variability in the training datasets. This leads to a nonrobust algorithm that may perform almost perfectly in the training environment for a specific laboratory, species, and tumor type, but not when applied to other tumor types, species, and images from different laboratories. Variations in staining protocols of tissue sections, slide preparation, or even color representation of different slide scanners can significantly affect the algorithm’s ability to recognize MF versus MLF.⁵ The current veterinary pathology laboratory environment lacks global standardization for tissue processing, staining, and slide preparation.¹ This means that application of insular MF algorithms developed under one set of laboratory conditions requires adaptation and validation if introduced to other laboratory conditions. Thus, this variability must be included in the training data sets (composed of thousands or tens of thousands of image labels) to develop robust, globally applicable algorithms for usage in every laboratory.

Lack of standardization for staining was a major difficulty for analysis of breast cancer images.^61,63 Previous studies have attributed many false positives to darkly stained objects without rods or spikes of chromosomes (false positive; MLF) or light-staining nuclei (false negatives).^62,63 Careful definitions and annotation of MF, AMF, and MLF by several independent experts and the availability of numerous example image patches (thousands to tens of thousands) are needed so that the programs can learn and teach themselves how to identify MF, AMF, and MLF. Robustness of algorithms is still an unsolved problem and requires development of large representative data sets from different laboratories, scanners, and cell types. However, if developed carefully, the power of such algorithms to quantitate MC and other parameters may help predict tumor behavior better and faster than pathologists presently can. Standardization of the morphologic criteria of MF and AMF for development of consistent and accurate ground truth data sets is an important part of creating CNN algorithms.

Future Diagnostic Use of CPATH

Due to the availability of advanced deep learning–based pattern recognition pipelines and “big data,” we are now able to develop accurate CPATH software that has the potential to be incorporated into pathologists’ workflow for enumerating MF in diagnostic and research settings. Speed and accuracy of algorithm analysis allows the ability to identify the area of highest mitotic density more accurately than pathologists (Fig. 51).⁷ Computer-assisted systems that support the pathologists in these critical decision-making steps are considered promising. Figures 27 to 52 show examples of candidate MF and MLF proposed by CPATH that require verification by a pathologist. For this specific approach, the first stage (object detection) has a high sensitivity in detection of MF candidates but detects high numbers of false positives (MLF). The refining second stage (object classifier) has a higher specificity and is therefore designed to remove the false positive detections. Any computational approach proposed for use in veterinary pathology would need to be tested and validated before widespread use could occur.^12,13 Correlating the results from CPATH analysis with outcome assessments is essential in evaluating the utility of these tools to predict tumor behavior. Higher reproducibility of the MC in combination with CPATH approaches may reveal additional trends of prognostic importance to which we were previously unaware.

Conclusions

The purpose of the MC is to determine an accurate reflection of mitotic activity within a tumor. This article provides morphologic criteria for differentiating MF, AMF, and MLF with recommendations for counting and categorizing MF. Figures 1 to 26 are our best examples of MF, AMF, and MLF. They were selected because they contain definitive diagnostic features used for microscopic identification and should have near-unanimous agreement. However, they are at a higher magnification and resolution than most anatomic pathologists use while working with diagnostic material and many have a cytologic partner image to complement histologic identification. These idealistic examples do not recapitulate a diagnostic setting. We cannot encompass all the variables associated with tissue preservation, staining variability, tissue artifacts, magnification at scanning, or time allocated for slide/image review. Supplemental Figures S1 to S84 are closer to what is experienced in routine identification of MF in diagnostic pathology. Examples of problematic structures, ranges of magnification, and variably stained tissues are included to highlight the difference between publication-quality images and working-level diagnostic material, so that readers can identify candidate structures for themselves and compare their interpretations to ours. Identification of structures taken out of context, as done in this article, is not part of a pathologist’s decision tree. However, it is one conundrum that CPATH will address. There will always be doubtful candidate structures that the pathologist cannot confidently identify. What to do with these structures will likely remain an individual pathologist’s or laboratory’s choice and is difficult to standardize. The most important determination is simple: Should the candidate structure be counted as a MF? Since some structures are difficult to classify and interpathologist variation is inevitable, the most straightforward way to handle these and other confounders is to simply count definitive MF and AMF.

Future studies should evaluate the precision and prognostic utility of restricting the MC to MF which are the most easily identified (metaphase, anaphase, telophase). Should this prove useful, there will be decreased interobserver variation in MCs, increased accuracy of CPATH MF identification, and greater standardization of the MC to compare with outcome assessment. Accurate identification of MF, AMF, and MLF are central elements in each of these components.

Supplemental Material

Supplemental Material, Combined_supplemental_materials-Donovan_et_al - Mitotic Figures—Normal, Atypical, and Imposters: A Guide to Identification

Supplemental Material, Combined_supplemental_materials-Donovan_et_al for Mitotic Figures—Normal, Atypical, and Imposters: A Guide to Identification by Taryn A. Donovan, Frances M. Moore, Christof A. Bertram, Richard Luong, Pompei Bolfa, Robert Klopfleisch, Harold Tvedten, Elisa N. Salas, Derick B. Whitley, Marc Aubreville and Donald J. Meuten in Veterinary Pathology

Footnotes

Acknowledgements

The authors thank Mr. Joshua Blaker for creating and providing and Supplemental Images S117 to S119. The authors also thank Ms. Lilah Blaker for providing references 17 and 21.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: CAB gratefully acknowledges financial support received from the Dres. Jutta & Georg Bruns-Stiftung für innovative Veterinärmedizin.

ORCID iD

Taryn A. Donovan

Christof A. Bertram

Robert Klopfleisch

Supplemental material for this article is available online.

References

Abels

Pantanowitz

Aeffner

, et al. Computational pathology definitions, best practices, and recommendations for regulatory guidance: a white paper from the Digital Pathology Association. J Pathol. 2019;249(3):286–294.

Al-Janabi

Huisman

Willems

, et al. Digital slide images for primary diagnostics in breast pathology: a feasibility study. Hum Pathol. 2012;43(12):2318–2325.

Al-Janabi

van Slooten

Visser

, et al. Evaluation of mitotic activity index in breast cancer using whole slide digital images. PLoS One. 2013;8(12):e82576.

Alkhasawneh

Reith

Toro

, et al. Interobserver variability of mitotic index and utility of PHH3 for risk stratification in gastrointestinal stromal tumors. Am J Clin Pathol. 2015;143(3):385–392.

Aubreville

Bertram

Donovan

Marzahl

Maier

Klopfleisch

. A completely annotated whole slide image dataset of canine breast cancer to aid human breast cancer research. Sci Data. 2020;7(1):417

Aubreville

Bertram

Jabari

, et al. Inter-species, inter-tissue domain adaptation for mitotic figure assessment. In: Tolxdorff

Deserno

Handels

, et al, eds. Bildverarbeitung für die Medizin 2020. Springer; 2020:1–7.

Aubreville

Bertram

Marzahl

, et al. Deep learning algorithms out-perform veterinary pathologists in detecting the mitotically most active tumor region. Sci Rep. 2020;10(1):16447.

Baak

. Mitosis counting in tumors. Hum Pathol. 1990;21(7):683–685.

Baak

Gudlaugsson

Skaland

, et al. Proliferation is the strongest prognosticator in node-negative breast cancer: significance, error sources, alternatives and comparison with molecular prognostic markers. Breast Cancer Res Treat. 2009;115(2):241–254.

10.

Bergers

Jannink

van Diest

, et al. The influence of fixation delay on mitotic activity and flow cytometric cell cycle variables. Hum Pathol. 1997;28(1):95–100.

11.

Bertram

Aubreville

Gurtner

, et al. Computerized calculation of mitotic count distribution in canine cutaneous mast cell tumor sections: mitotic count is area dependent. Vet Pathol. 2020;57(2):214–226.

12.

Bertram

Aubreville

Marzahl

, et al. A large-scale dataset for mitotic figure assessment on whole slide images of canine cutaneous mast cell tumor. Sci Data. 2019;6(274):274.

13.

Bertram

Gurtner

Dettwiler

, et al. Validation of digital microscopy compared with light microscopy for the diagnosis of canine cutaneous tumors. Vet Pathol. 2018;55(4):490–500.

14.

Bertram

Klopfleisch

. The pathologist 2.0: an update on digital pathology in veterinary medicine. Vet Pathol. 2017;54(5):756–766.

15.

Bertram

Veta

Marzahl

, et al. Are pathologist-defined labels reproducible? Comparison of the TUPAC16 mitotic figure dataset with an alternative set of labels. In: Cardoso

, et al. eds. iMIMIC 2020/MIL3iD 2020/LABELS 2020, LNCS 12446, Springer Nature Switzerland AG; 2020:204–213.

16.

Bonsembiante

Bonfanti

Cian

, et al. Diagnostic validation of a whole-slide imaging scanner in cytological samples: diagnostic accuracy and comparison with light microscopy. Vet Pathol. 2019;56(3):429–434.

17.

Campbell

Reece

Urry

, et al. The cell cycle. In: Wilbur

, ed. Biology. 8th ed. Pearson; 2009:231–234.

18.

Campbell

Lele

West

, et al. Concordance between whole-slide imaging and light microscopy for routine surgical pathology. Hum Pathol. 2012;43(10):1739–1744.

19.

Camus

Priest

Koehler

, et al. Cytologic criteria for mast cell tumor grading in dogs with evaluation of clinical outcome. Vet Pathol. 2016;53(6):1117–1123.

20.

Chowdhury

Pai

Lobo

, et al. Interobserver variation in breast cancer grading: a statistical modeling approach. Anal Quant Cytol Histol. 2006;28(4):213–218.

21.

Clark

Choi

Douglas

. The cell cycle. In: Biology. 2nd ed. Openstax; 2018:279–283.

22.

Coindre

Trojani

Contesso

, et al. Reproducibility of a histopathologic grading system for adult soft tissue sarcoma. Cancer. 1986;58(2):306–309.

23.

Cross

Start

Smith

. Does delay in fixation affect the number of mitotic figures in processed tissue? J Clin Pathol. 1990;43(2):597–599.

24.

Cullen

Breen

. An overview of molecular cancer pathogenesis, prognosis, and diagnosis. In: Meuten

, ed. Tumors in Domestic Animals. Wiley Blackwell; 2016:1–26.

25.

Donhuijsen

Schmidt

Hirche

, et al. Changes in mitotic rate and cell cycle fractions caused by delayed fixation. Hum Pathol. 1990;21(7):709–714.

26.

El-Alfy

Leblond

. Long duration of mitosis and consequences for the cell cycle concept, as seen in the isthmal cells of the mouse pyloric antrum. II. Duration of mitotic phases and cycle stages, and their relation to one another. Cell Tissue Kinet. 1987;20(2):215–226.

27.

Frierson

Jr Wolber

Berean

, et al. Interobserver reproducibility of the Nottingham modification of the Bloom and Richardson histologic grading scheme for infiltrating ductal carcinoma. Am J Clin Pathol. 1995;103(2):195–198.

28.

Garbe

Eigentler

Bauer

, et al. Mitotic rate in primary melanoma: interobserver and intraobserver reliability, analyzed using H&E sections and immunohistochemistry. J Dtsch Dermatol Ges. 2016;14(9):910–915.

29.

García-Rojo

. International clinical guidelines for the adoption of digital pathology: a review of technical aspects. Pathobiology. 2016;83(2-3):99–109.

30.

Gisselsson

. Classification of chromosome segregation errors in cancer. Chromosoma. 2008;117(6):511–519.

31.

Grosset

Odartchenko

. Duration of mitosis and separate mitotic phases compared to nuclear DNA content in erythroblasts of four vertebrates. Cell Tissue Kinet. 1975;8(1):91–96.

32.

Hanna

Reuter

Hameed

, et al. Whole slide imaging equivalency and efficiency study: experience at a large academic center. Mod Pathol. 2019;32(7):916–928.

33.

Hergt

von Bomhard

Kent

, et al. Use of a 2-tier histologic grading system for canine cutaneous mast cell tumors on cytology specimens. Vet Clin Pathol. 2016;45(3):477–483.

34.

Jannink

Van Diest

Baak

. Comparison of the prognostic value of four methods to assess mitotic activity in 186 invasive breast cancer patients: classical and random mitotic activity assessments with correction for volume percentage of epithelium. Hum Pathol. 1995;26(10):1086–1092.

35.

Jin

Stewenius

Lindgren

, et al. Distinct mitotic segregation errors mediate chromosomal instability in aggressive urothelial cancers. Clin Cancer Res. 2007;13(6):1703–1712.

36.

Kalatova

Jesenska

Hlinka

, et al. Tripolar mitosis in human cells and embryos: occurrence, pathophysiology and medical implications. Acta Histochem. 2015;117(1):111–125.

37.

Kim

Pantanowitz

Schüffler

, et al. (Re)Defining the high-power field for digital pathology. J Pathol Inform. 2020;11(1):33.

38.

Lehr

Rochat

Schaper

, et al. Mitotic figure counts are significantly overestimated in resection specimens of invasive breast carcinomas. Mod Pathol. 2013;26(3):336–342.

39.

Lewis

. The duration of the various phases of mitosis in the mesenchyme cells of tissue. Anat Rec. 1917;13:359–367.

40.

Malon

Brachtel

Cosatto

, et al. Mitotic figure recognition: agreement among pathologists and computerized detector. Anal Cell Pathol (Amst). 2012;35(2):97–100.

41.

Matsuda

Yoshimura

Ishiwata

, et al. Mitotic index and multipolar mitosis in routine histologic sections as prognostic markers of pancreatic cancers: a clinicopathological study. Pancreatology. 2016;16(1):127–132.

42.

McIntosh

. Mitosis. Cold Spring Harb Perspect Biol. 2016;8(9):a023218.

43.

Medri

Volpi

Nanni

, et al. Prognostic relevance of mitotic activity in patients with node-negative breast cancer. Mod Pathol. 2003;16(11):1067–1075.

44.

Meuten

Moore

George

. Mitotic count and the field of view area: time to standardize. Vet Pathol. 2016;53(1):7–9.

45.

Meyer

Alvarez

Milikowski

, et al. Breast carcinoma malignancy grading by Bloom-Richardson system vs proliferation index: reproducibility of grade and advantages of proliferation index. Mod Pathol. 2005;18(8):1067–1078.

46.

Nagele

Freeman

McMorrow

, et al. Precise spatial positioning of chromosomes during prometaphase: evidence for chromosomal order. Science. 1995;270(5243):1831–1835.

47.

Ohashi

Namimatsu

Sakatani

, et al. Prognostic utility of atypical mitoses in patients with breast cancer: a comparative study with Ki67 and phosphohistone H3. J Surg Oncol. 2018;118(3):557–567.

48.

Pantanowitz

Sinard

Henricks

, et al. Validating whole slide imaging for diagnostic purposes in pathology: guideline from the College of American Pathologists Pathology and Laboratory Quality Center. Arch Pathol Lab Med. 2013;137(12):1710–1722.

49.

Ressel

. The biopsy “undead.” Vet Pathol. 2018;55(6):916–917.

50.

Robinow

. Mitosis in the yeast Lipomyces lipofer. J Biophys Biochem Cytol. 1961;9(4):879–892.

51.

Santos

Correia-Gomes

Santos

, et al. Interobserver reproducibility of histological grading of canine simple mammary carcinomas. J Comp Pathol. 2015;153(1):22–27.

52.

Scarpa

Sabattini

Bettini

. Cytological grading of canine cutaneous mast cell tumours. Vet Ccomp Ooncol. 2016;14(3):245–251.

53.

Stathonikos

Veta

Huisman

, et al. Going fully digital: perspective of a Dutch academic pathology lab. J Pathol Inform. 2013;4:15.

54.

Tabata

Uraoka

Benhamida

, et al. Validation of mitotic cell quantification via microscopy and multiple whole-slide scanners. Diagn Pathol. 2019;14(1):65.

55.

Tellez

Balkenhol

Otte-Höller

, et al. Whole-slide mitosis detection in H&E breast histology using PHH3 as a reference to train distilled stain-invariant convolutional networks. IEEE Ttrans Med Imaging. 2018;37(9):2126–2136.

56.

Thunnissen

Ambergen

Koss

, et al. Mitotic counting in surgical pathology: sampling bias, heterogeneity and statistical uncertainty. Histopathology. 2001;39(1):1–8.

57.

Tsuda

Akiyama

Kurosumi

, et al. Evaluation of the interobserver agreement in the number of mitotic figures of breast carcinoma as simulation of quality monitoring in the Japan National Surgical Adjuvant Study of Breast Cancer (NSAS-BC) protocol. Jpn J Cancer Res. 2000;91(4):451–457.

58.

Tvedten

. Atypical mitoses: morphology and classification. Vet Clin Pathol. 2009;38(4):418–420.

59.

van Diest

Baak

Matze-Cok

, et al. Reproducibility of mitosis counting in 2,469 breast cancer specimens: results from the Multicenter Morphometric Mammary Carcinoma Project. Hum Pathol. 1992;23(6):603–607.

60.

Veta

Heng

Stathonikos

, et al. Predicting breast tumor proliferation from whole-slide images: the TUPAC16 challenge. Med Image Anal. 2019;54:111–121.

61.

Veta

Van Diest

Jiwa

, et al. Mitosis counting in breast cancer: object-level interobserver agreement and comparison to an automatic method. PLoS One. 2016;11(8):e0161286.

62.

Veta

Van Diest

Willems

, et al. Assessment of algorithms for mitosis detection in breast cancer histopathology images. Med Image Anal. 2015;20:237–248.

63.

Wang

Cruz-Roa

Basavanhally

, et al. Mitosis detection in breast cancer pathology images by combining handcrafted and convolutional neural network features. J Med Imaging (Bellingham). 2014;1(3):034003.

64.

Wei

Halsey

Hoover

, et al. Agreement in histological assessment of mitotic activity between microscopy and digital whole slide images informs conversion for clinical diagnosis. Acad Pathol. 2019;6:2374289519859841.

65.

Yap

Rasotto

Priestnall

, et al. Intra- and inter-observer agreement in histological assessment of canine soft tissue sarcoma. Vet Comp Oncol. 2017;15(4):1553–1557.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

10.49 MB

Mitotic Figures—Normal,Atypical,and Imposters: A Guide to Identification

Abstract

Keywords

Mitosis

Morphologies of MF in Histology and Cytology

Prophase

Prometaphase and Metaphase

Anaphase

Telophase

Consensus of MF Morphology

Prophase

Telophase

Morphologies of AMF in Histology and Cytology

AMF With Mitotic/Polar Asymmetry

AMF With Abnormalities of Chromosome Segregation

Significance of AMF

Mitotic-Like Figures

Factors That Affect MF Identification and the MC

Slide Preparation

Fixation Delay

MF Identification in Whole-Slide Images (WSI)

Pros and Cons of WSI for MF Detection

WSI Validation for Diagnosis and MC

Classifying Individual MF With WSI

The Future: CPATH Recognition of MF, AMF, and MLF

Reasons to Use CPATH for MF Identification

Recent CPATH Progress in Veterinary and Human Pathology

CPATH Limitations

Future Diagnostic Use of CPATH

Conclusions

Supplemental Material

Supplemental Material, Combined_supplemental_materials-Donovan_et_al - Mitotic Figures—Normal, Atypical, and Imposters: A Guide to Identification

Footnotes

Acknowledgements

Declaration of Conflicting Interests

Funding

ORCID iD

References

Supplementary Material