01.12.2017 | Methodology | Ausgabe 1/2017 Open Access

# Computer-assisted stereology and automated image analysis for quantification of tumor infiltrating lymphocytes in colon cancer

- Zeitschrift:
- Diagnostic Pathology > Ausgabe 1/2017

## Electronic supplementary material

## Background

^{2}) of TILs in immunohistochemically stained sections, however, only few studies have validated this method [14]. To enable the use of TILs as a clinical biomarker in colorectal cancer, an Immunoscore has been proposed [6–8, 15]. The Immunoscore is based on estimates of two different populations of T-lymphocytes counted in two different areas of the tumor defined as the center and invasive front of the tumor, respectively. For prognostic purposes, CD3+ and CD8+ TILs appeared to be most informative, and a multinational investigation of the practical implementation and clinical impact of the Immunoscore is in progress [16]. The combined analysis of tumor center and invasive margin has been suggested to overcome sampling bias caused by heterogeneity. Intra-tumoral heterogeneity is an inherent characteristic of malignant tumors, and it is well known that the micro-environment varies throughout the tumor with effects on growth, proliferation and metastatic potential [17, 18]. This is also the case in CC, which is known to be architecturally, molecularly and biologically heterogeneous [19]. It is well known that tumor heterogeneity may have significant impact on the interpretation of biomarkers [20], but the overall understanding of intra- and inter-tumoral heterogeneity of the micro-environment is limited [21], and to our knowledge the heterogeneity of TILs has only been sparsely investigated in CC. Thus, heterogeneity is of crucial importance in biomarker research, especially in the perspective of clinical application [22].

## Methods

### Patients and tissue

### Immunohistochemistry

_{4}in TBS buffer pH 7.6 for 10 min. Meyer’s hematoxylin (Merck, Damstadt, Germany) was used as counterstain, and finally, the histological slides were coverslipped with Tissue-Tek PERTEX (Histolab Products AB, Göteborg, Sweden).

### Scanning of histological slides and identification of regions of interest

### Stereological analysis

^{2}and 2.925 μm

^{2}for CD3 and CD8 TILs estimation, respectively, taking into account the different densities of the two lymphocytic populations (Fig. 4).

_{A}= numerical density) was calculated as the number of positive lymphocytic profiles divided by the total, investigated counting frame area [24]:

^{2}. The area fraction was calculated dimensionless as the sum of points hitting the CD3+ or CD8+ lymphocytes divided by the sum of the points hitting the vital tumor tissue within the ROI [24]:

### Image analysis

### Statistical analysis

## Results

^{2}(mean 20.7 mm

^{2}) for the CA, and from 1.7 to 43.4 mm

^{2}(mean 12.2 mm

^{2}) for the IA.

### Correlation between image analysis and stereology

#### Density

^{2}, when counted stereologically, while image analysis ranged from 53 to 1680 cells/mm

^{2}. In the IA the CD3+ numerical T-cell densities varied from 60 to 2302 cells/mm

^{2}by stereology and from 57 to 1927 cells/mm

^{2}by image analysis. In the CA, the CD8+ numerical T-cell densities varied from 8 to 2043 cells/mm

^{2}by stereology, while image analysis ranged from 18 to 2195 cells/mm

^{2}. In the IA the CD8+ numerical T-cell densities varied from 17 to 1852 cells/mm

^{2}and from 24 to 1852 cells/mm

^{2}by stereology and image analysis, respectively (Table 1).

Mean | Min | Max | |
---|---|---|---|

CD3-Central Image analysis | 557 | 53 | 1680 |

CD3-Central Stereology | 533 | 50 | 1786 |

CD3-Invasive Image analysis | 373 | 57 | 1927 |

CD3-Invasive Stereology | 402 | 60 | 2302 |

CD8-Central Image analysis | 375 | 18 | 2195 |

CD8-Central Stereology | 285 | 8 | 2043 |

CD8-Invasive Image analysis | 398 | 24 | 1695 |

CD8- Invasive Stereology | 360 | 17 | 1852 |

Spearman correlation All sections (n = 129) | Spearman correlation Deepest section (n = 43) | Spearman correlation Random section A (n = 43) | Spearman correlation Random section B (n = 43) | |
---|---|---|---|---|

CD3-Central | 0.9457 | 0.9623 | 0.9328 | 0.9080 |

CD8-Central | 0.9638 | 0.9609 | 0.9745 | 0.9431 |

CD3-Invasive | 0.9496 | 0.9678 | 0.9420 | 0.9644 |

CD8-Invasive | 0.9552 | 0.9782 | 0.9367 | 0.9381 |

#### Area fraction

Mean | Min | Max | |
---|---|---|---|

CD3- Central Image analysis | 2.96 | 0.28 | 9.44 |

CD3-Central Stereology | 1.90 | 0.15 | 6.67 |

CD3-Invasive Image analysis | 3.11 | 0.27 | 9.43 |

CD3-Invasive Stereology | 2.12 | 0.16 | 6.77 |

CD8-Central Image analysis | 2.62 | 0.11 | 15.01 |

CD8-Central Stereology | 1.83 | 0.03 | 13.59 |

CD8-Invasive Image analysis | 2.84 | 0.17 | 14.07 |

CD8-Invasive Stereology | 2.32 | 0.08 | 13.33 |

Spearman correlation All sections (n = 129) | Spearman correlation Deepest section (n = 43) | Spearman correlation Random section A (n = 43) | Spearman correlation Random section B (n = 43) | |
---|---|---|---|---|

CD3-Central | 0.9404 | 0.9434 | 0.9322 | 0.9244 |

CD8-Central | 0.9603 | 0.9499 | 0.9671 | 0.9269 |

CD3-Invasive | 0.9400 | 0.9406 | 0.9394 | 0.9710 |

CD8-Invasive | 0.9497 | 0.9665 | 0.9080 | 0.9557 |

#### Intra-observer / intra-technical reliability

Spearman correlation All sections (n = 129) | Spearman correlation Deepest section (n = 43) | Spearman correlation Random section A (n = 43) | Spearman correlation Random section B (n = 43) | |
---|---|---|---|---|

CD3-Central Stereology | 0.9575 | 0.9503 | 0.9688 | 0.9448 |

CD3-Central Image analysis | 0.9926 | 0.9872 | 0.9878 | 0.9899 |

CD8-Central Stereology | 0.9796 | 0.9838 | 0.9578 | 0.9872 |

CD8-Central Image analysis | 0.9932 | 0.9869 | 0.9941 | 0.9917 |

CD3-Invasive Stereology | 0.9519 | 0.9457 | 0.9443 | 0.9438 |

CD3-Invasive Image analysis | 0.9890 | 0.9905 | 0.9787 | 0.9962 |

CD8-Invasive Stereology | 0.9676 | 0.9665 | 0.9713 | 0.9565 |

CD8-Invasive Image analysis | 0.9907 | 0.9917 | 0.9912 | 0.9878 |

#### Evaluation of efficiency

### Heterogeneity

ICC (CD3) | ICC (CD8) | |
---|---|---|

Image analysis Central | 0.712 (0.613–0.810) | 0.749 (0.661–0.838) |

Stereology Central | 0.665 (0.555–0.776) | 0.775 (0.693–0.856) |

Image analysis Invasive | 0.686 (0.581–0.792) | 0.765 (0.681–0.849) |

Stereology Invasive | 0.707 (0.607–0.807) | 0.765 (0.682–0.849) |

ICC (CD3) | ICC (CD8) | |
---|---|---|

Image analysis Central | 0.704 (0.603–0.804) | 0.746 (0.657–0.836) |

Stereology Central | 0.615 (0.493–0.737) | 0.724 (0.628–0.819) |

Image analysis Invasive | 0.702 (0.601–0.804) | 0.763 (0.678–0.847) |

Stereology Invasive | 0.746 (0.657–0.836) | 0.746 (0.657–0.835) |

## Discussion

### Correlation between image analysis and stereology

^{2}(median 471 cells/mm

^{2}). They compared with computer-assisted image analysis and found correlation coefficients very similar to our results, varying from 0.960 to 0.987. Their manual counts were not based on stereology, and moreover, they were performed in a limited number of fields of vision without a clearly stated sampling approach. The counts were, however, performed by four different observers. We only had one observer, but using strict stereological counting rules our results were reproducible by both techniques. We investigated the correlation between the numerical density estimates and the area fractions, as obtained separately by the stereological technique or image analysis, and found excellent correlation coefficients for these “intra-technical reproducibility tests”.

### Heterogeneity

^{2}[8] to 3.14 mm

^{2}[9]. We used three whole sections from each tumor and analyzed a considerably larger tumor bulk than TMA-based studies.