Continuous representation of tumor microvessel density and detection of angiogenic hotspots in histological whole-slide images - PubMed (original) (raw)
Continuous representation of tumor microvessel density and detection of angiogenic hotspots in histological whole-slide images
Jakob Nikolas Kather et al. Oncotarget. 2015.
Abstract
Blood vessels in solid tumors are not randomly distributed, but are clustered in angiogenic hotspots. Tumor microvessel density (MVD) within these hotspots correlates with patient survival and is widely used both in diagnostic routine and in clinical trials. Still, these hotspots are usually subjectively defined. There is no unbiased, continuous and explicit representation of tumor vessel distribution in histological whole slide images. This shortcoming distorts angiogenesis measurements and may account for ambiguous results in the literature. In the present study, we describe and evaluate a new method that eliminates this bias and makes angiogenesis quantification more objective and more efficient. Our approach involves automatic slide scanning, automatic image analysis and spatial statistical analysis. By comparing a continuous MVD function of the actual sample to random point patterns, we introduce an objective criterion for hotspot detection: An angiogenic hotspot is defined as a clustering of blood vessels that is very unlikely to occur randomly. We evaluate the proposed method in N=11 images of human colorectal carcinoma samples and compare the results to a blinded human observer. For the first time, we demonstrate the existence of statistically significant hotspots in tumor images and provide a tool to accurately detect these hotspots.
Keywords: digital pathology; spatial statistics; tumor angiogenesis; vessel density.
Conflict of interest statement
CONFLICTS OF INTEREST
The authors declare the following conflicts of interest. JNK: none, AM: none, CCRA: none, LRS: none, FGZ: none, CAW: none
Figures
Figure 1. Immunostained blood vessels are automatically segmented in whole-slide images
In this figure, image tiles of 1600×1600 px are shown and an image detail is enlarged. A. Original image region, B. result of color deconvolution, C. result of thresholding and morphological post-processing.
Figure 2. Whole slide image of a colorectal tumor sample
Panel A. shows a whole slide image of a colorectal cancer sample, two tumor regions are delineated by hand. Analysis of the right-hand region is subsequently shown in Figure. 4. Below, a detail is enlarged to demonstrate the staining quality. In panel B. a fat tissue sample is shown. The delineated fat tissue region served as a negative control for validation of the new method. Below the main panel, a detail is enlarged. The corresponding blood vessel map can be found in Figure 3.
Figure 3. Blood vessels in fat tissue are distributed randomly and do not show significant clustering
A point map of vessels in fat tissue from Figure 2B is shown in panel A.1. Point map of random pattern A.2. in the same region; microvessel density function of fat tissue B.1.; density function of the complete spatial randomness (CSR) model B.2.; color-coded units of the density functions are arbitrary. C. Probability map (units: standard deviations of CSR), Bonferroni corrected level of significance is at F = 5.25 (marked by *). No significant angiogenic hotspot can be detected in this tissue region.
Figure 4. Blood vessels in colorectal tumor tissue show highly significant clustering
A point map of vessels in colorectal tumor tissue from Figure 2A is shown in A.1. Point map of random pattern A.2. in the same region; microvessel density function of tumor tissue B.1.; density function of the complete spatial randomness (CSR) model B.2.; color-coded units of the density functions are arbitrary. C. Probability map (units: standard deviations of CSR), Bonferroni corrected level of significance is at F = 5.27 (marked by *). Five statistically significant tumor angiogenic hotspots emerge in this region (indicated by arrows).
Figure 5. Comparison of manual and automatic hotspot detection
In this figure, a whole slide image is shown in low magnification. The blue line shows the contour of the tumor. The black lines show angiogenic hotspots as delineated by a blinded human observer (# marks the primary hotspot as defined by the observer). The hotspot probability map is overlaid (red/yellow; level of significance is indicated by *). It can be seen that all manually detected hotspot areas were also detected by the automatic method.
Figure 6. Spatial statistics for fat and tumor
Empty space function F(r) and nearest neighbour distance distribution function G(r) for vessels in fat A. and tumor B. tissue. The observed functions are plotted against the functions of a corresponding random pattern. While F and G for fat do not differ from the random functions, tumor vessel distribution markedly differs from a random pattern. km = Kaplan-Meier estimate, cs = Chiu-Stoyan estimate, bord = border corrected estimate, han = Hanisch estimate, pois = theoretical Poisson distribution (CSR).
Similar articles
- Development and Validation of a Histological Method to Measure Microvessel Density in Whole-Slide Images of Cancer Tissue.
Marien KM, Croons V, Waumans Y, Sluydts E, De Schepper S, Andries L, Waelput W, Fransen E, Vermeulen PB, Kockx MM, De Meyer GR. Marien KM, et al. PLoS One. 2016 Sep 1;11(9):e0161496. doi: 10.1371/journal.pone.0161496. eCollection 2016. PLoS One. 2016. PMID: 27583442 Free PMC article. - Morphometric study of microvessels in primary CNS tumors and its correlation with tumor types and grade.
Deb P, Boruah D, Dutta V. Deb P, et al. Microvasc Res. 2012 Jul;84(1):34-43. doi: 10.1016/j.mvr.2012.03.004. Epub 2012 Mar 23. Microvasc Res. 2012. PMID: 22465694 - Tumor microvasculature characteristics studied by image analysis: histologically-driven angiogenic profile.
Fakhrejahani E, Asao Y, Toi M. Fakhrejahani E, et al. Int J Biol Markers. 2014 Sep 30;29(3):e204-7. doi: 10.5301/jbm.5000098. Int J Biol Markers. 2014. PMID: 25041785 Review.
Cited by
- Deep learning-based automated pipeline for blood vessel detection and distribution analysis in multiplexed prostate cancer images.
Karageorgos GM, Cho S, McDonough E, Chadwick C, Ghose S, Owens J, Jung KJ, Machiraju R, West R, Brooks JD, Mallick P, Ginty F. Karageorgos GM, et al. Front Bioinform. 2024 Jan 23;3:1296667. doi: 10.3389/fbinf.2023.1296667. eCollection 2023. Front Bioinform. 2024. PMID: 38323039 Free PMC article. - Immunohistochemical Assessment of Microvessel Density in OSCC: Spatial Heterogeneity of Angiogenesis and Its Impact on Survival.
Mamilos A, Lein A, Winter L, Haas M, Reichert TE, Ettl T, Künzel J, Spanier G, Brochhausen C. Mamilos A, et al. Biomedicines. 2023 Oct 8;11(10):2724. doi: 10.3390/biomedicines11102724. Biomedicines. 2023. PMID: 37893098 Free PMC article. - Artificial Intelligence Assists in the Detection of Blood Vessels in Whole Slide Images: Practical Benefits for Oncological Pathology.
Timakova A, Ananev V, Fayzullin A, Makarov V, Ivanova E, Shekhter A, Timashev P. Timakova A, et al. Biomolecules. 2023 Aug 29;13(9):1327. doi: 10.3390/biom13091327. Biomolecules. 2023. PMID: 37759727 Free PMC article. Review. - Oral squamous cell carcinomas: state of the field and emerging directions.
Tan Y, Wang Z, Xu M, Li B, Huang Z, Qin S, Nice EC, Tang J, Huang C. Tan Y, et al. Int J Oral Sci. 2023 Sep 22;15(1):44. doi: 10.1038/s41368-023-00249-w. Int J Oral Sci. 2023. PMID: 37736748 Free PMC article. Review. - PAK1 overexpression promotes myxofibrosarcoma angiogenesis through STAT5B-mediated CSF2 transactivation: clinical and therapeutic relevance of amplification and nuclear entry.
Li CF, Chan TC, Fang FM, Yu SC, Huang HY. Li CF, et al. Int J Biol Sci. 2023 Jul 31;19(12):3920-3936. doi: 10.7150/ijbs.83467. eCollection 2023. Int J Biol Sci. 2023. PMID: 37564209 Free PMC article.
References
- Carmeliet P. Angiogenesis in life, disease and medicine. Nature. 2005;438:932–6. - PubMed
- Folkman J. Angiogenesis in cancer, vascular, rheumatoid and other disease. Nat Med. 1995;1:27–31. - PubMed
- Pusztaszeri MP, Seelentag W, Bosman FT. Immunohistochemical expression of endothelial markers CD31, CD34, von Willebrand factor, and Fli-1 in normal human tissues. J Histochem Cytochem. 2006;54:385–95. - PubMed
- Tomlinson J, Barsky SH, Nelson S, Singer S, Pezeshki B, Lee MC, Eilber F, Nguyen M. Different patterns of angiogenesis in sarcomas and carcinomas. Clin Cancer Res. 1999;5:3516–22. - PubMed
- Eberhard A, Kahlert S, Goede V, Hemmerlein B, Plate KH, Augustin HG. Heterogeneity of angiogenesis and blood vessel maturation in human tumors: implications for antiangiogenic tumor therapies. Cancer Res. 2000;60:1388–93. - PubMed
Publication types
MeSH terms
LinkOut - more resources
Full Text Sources
Other Literature Sources