Whole-slide imaging: where does the future lie?

Whole-slide imaging (WSI), also known as virtual microscopy, involves the use of scanners to image the entire tissue slide. The resulting high-resolution digital images are available for pathologists to review or share with collaborators around the globe. WSI scanners and technologies for whole-slide image analysis are evolving rapidly. WSI is becoming an integral part of digital pathology, and emerging WSI technologies promise to shape the future of diagnostic pathology.^1,2

Employing multispectral WSI

The popularity of multispectral imaging amongst pathologists is increasing rapidly. By facilitating the acquisition of tissue images at different wavelengths, multispectral imaging allows for the in-depth characterization of tissues stained simultaneously for different markers.³ In contrast to RGB images acquired by a standard digital camera, images acquired by multispectral imaging can be used to detect the expression and distribution of multiple molecules (up to 10 spectral bands at 16-bit per pixel resolution) in the same tissue section.^4,5

This feature renders multispectral imaging particularly useful for the characterization of the complex microenvironment of tumors and the classification of tumor-infiltrating immune cells;⁶ elucidating the extent of tumor infiltration by different immune cell subsets can predict patient response to immunotherapy.⁷ Additionally, multispectral imaging can improve marker-based tumor classification, maximizing the diagnostic and prognostic potential of histological examination.⁸

Despite the many advantages of multispectral imaging, its implementation in WSI can be challenging. In multispectral imaging, images are typically acquired and stored at each wavelength examined, increasing significantly the scanning time and size of WSI files. With the continuing decrease in the costs associated with data storage and the development of faster whole-slide scanners, the use of multispectral imaging in WSI is becoming more feasible.⁹ Additionally, the development of systems to acquire multispectral images without requiring multiple passes is gaining increasing attention and may provide a solution to many of the challenges limiting the use of multispectral imaging in WSI.^10,11 For instance, Liao et al.¹⁰ developed a WSI multichannel microscopy system based on a dual light‐emitting diode. The system allows the simultaneous acquisition of six different focal planes of thick specimens; by relaying these focal planes to the same focal position, information at six spectral bands can be obtained simultaneously.¹⁰

Adopting three-dimensional image reconstruction

Three-dimensional (3D) imaging entails the use of sensors and detectors to acquire image stacks—also known as z-stacks—that are then converted into a 3D representation.¹² As two-dimensional WSI is not sufficient to provide an accurate representation of heterogeneous 3D structures, such as tumors, 3D reconstruction of whole-slide images can improve clinical diagnosis.^12,13

The generation of 3D whole-slide images starts with the mounting of serial ultrathin (4 to 6 μm) tissue sections to glass slides. Automated preparation of serial sections (e.g., using robotic microtomes) is typically preferred to facilitate the 3D reconstruction process, as automated sectioning minimizes sectioning artifacts and variations in section thickness, as well as improves section alignment—all of which are paramount for generating high-quality 3D whole-slide reconstructions.¹²

Serial sections are then stained (H&E or immunohistochemical staining) and scanned using a WSI scanner, generating a z-stack of digital images for each level of the tissue block. Finally, specialized software is used to generate 3D reconstructions based on the serial whole-slide images.¹²

Emerging technologies for whole-slide image analysis

The analysis of whole-slide images is challenging because of the high heterogeneity of cell phenotypes and morphology, tissue architecture, and staining colors in tissue sections. The heterogeneity of pathological phenotypes further complicates the interpretation of WSI data, which is also affected—at least to some extend—by observer bias.¹⁴ Therefore, the development of novel, automated, and robust whole-slide image analysis methods has come to the forefront, and many new image analysis technologies have been introduced in the last few years.

A common feature of these technologies is their ability to accurately delineate the region of interest. Typically, whole-slide images contain various types of tissues and cells, ranging from immune cells, pathological cells (e.g., tumor cells), stromal cells, connective tissue cells, and vasculature. Thus, the ability of WSI analysis algorithms to accurately identify the target region and segmentation of cells within complex tissues is of paramount importance.¹⁵ Most algorithms allow users to adjust image parameters and thresholds to ensure optimal cell and nuclear segmentation in digital pathology images.^16,17

Currently, image analysis algorithms exist to automate the evaluation of digital pathology images of tissues stained for specific biomarkers (membrane, cytoplasmic, or nuclear). However, appropriate quality control measures are required to eliminate image artifacts (e.g., segmentation and classification errors) and minimize the risk of misdiagnosis.¹⁷

Integrating artificial intelligence in whole-slide image analysis

Artificial intelligence (AI) has undoubtedly transformed various aspects of healthcare, including pathology and medical diagnosis. AI involves the use of systems that possess human intelligence; hence, AI technologies have emerged as a powerful tool to tackle complex problems and make “smart decisions” based on available data.¹⁸

Machine learning (ML) entails the use of AI-assisted computer algorithms that can be trained to perform automated tasks. ML algorithms can build models based on information from training datasets. These models are then used by computers to make predictions or make decisions. ML can be particularly useful in whole-slide image analysis and precision pathology owing to their exceptional ability to identify clinically relevant patterns in images.¹⁹

Numerous studies have confirmed the diagnostic and prognostic value of ML algorithms, and commercially available ML-assisted platforms are gaining increasing popularity amongst pathologists and diagnostic laboratories. ML has proved particularly useful for the automated evaluation of H&E-stained pathology images, cellular characterization of the tumor microenvironment, and morphologic analysis of tissues.²⁰

Furthermore, the recent development of deep ML technologies has enabled the automatization of image segmentation and object classification. Analysis of whole slide images of tumors using deep ML algorithms can detect metastatic lesions, classify tumors, and predict response to treatment.²¹ Hence, the wide adoption of ML technologies in pathology is expected to enhance pathology workflow, accelerate diagnosis, and improve the prognostic accuracy of pathological analyses.

 Conclusions

The implementation of WSI in routine pathology is increasing rapidly. With more and more WSI applications gaining regulatory approval and the rapid evolution of WSI technologies, it is imperative that pathologists and researchers remain up to date with the latest technological advances in this field but also with emerging technologies. In addition, pathologists and researchers should ensure that they are aware of not only the advantages of WSI technologies but also their pitfalls.

References

1. Ghaznavi F, Evans A, Madabhushi A, Feldman M. Digital imaging in pathology: Whole-slide imaging and beyond. Annu Rev Pathol Mech Dis. 2013;8(October):331-359. doi:10.1146/annurev-pathol-011811-120902
2. Kumar N, Gupta R, Gupta S. Whole Slide Imaging (WSI) in Pathology: Current Perspectives and Future Directions. J Digit Imaging. 2020;33(4):1034-1040. doi:10.1007/s10278-020-00351-z
3. Levenson RM, Fornari A, Loda M. Multispectral imaging and pathology: seeing and doing more. Expert Opin Med Diagn. 2008;2(9):1067-1081. doi:10.1517/17530059.2.9.1067
4. Mansfield JR. Multispectral Imaging: A Review of Its Technical Aspects and Applications in Anatomic Pathology. Vet Pathol. 2014;51(1):185-210. doi:10.1177/0300985813506918
5. Ortega S, Halicek M, Fabelo H, Callico GM, Fei B. Hyperspectral and multispectral imaging in digital and computational pathology: a systematic review [Invited]. Biomed Opt Express. 2020;11(6):3195-3233. doi:10.1364/BOE.386338
6. Wickenhauser C, Bethmann D, Feng Z, et al. Multispectral fluorescence imaging allows for distinctive topographic assessment and subclassification of tumor-infiltrating and surrounding immune cells. Methods Mol Biol. 2019;1913:13-31. doi:10.1007/978-1-4939-8979-9_2
7. Maibach F, Sadozai H, Seyed Jafari SM, Hunger RE, Schenk M. Tumor-Infiltrating Lymphocytes and Their Prognostic Value in Cutaneous Melanoma. Front Immunol. 2020;11(September):1-20. doi:10.3389/fimmu.2020.02105
8. Maggioni M, Davis GL, Warner FJ, et al. Hyperspectral microscopic analysis of normal, benign and carcinoma microarray tissue sections. Opt Biopsy VI. 2006;6091(d):60910I. doi:10.1117/12.646078
9. Guzman M, Judkins AR. Digital pathology: A tool for 21st century neuropathology. Brain Pathol. 2009;19(2):305-316. doi:10.1111/j.1750-3639.2009.00264.x
10. Liao J, Wang Z, Zhang Z, et al. Dual light-emitting diode-based multichannel microscopy for whole-slide multiplane, multispectral and phase imaging. J Biophotonics. 2018;11(2):e201700075. doi:https://doi.org/10.1002/jbio.201700075
11. Ono S. Snapshot multispectral imaging using a pixel-wise polarization color image sensor. Opt Express. 2020;28(23):34536. doi:10.1364/oe.402947
12. Farahani N, Braun A, Jutt D, et al. Three-dimensional Imaging and Scanning: Current and Future Applications for Pathology. J Pathol Inform. 2017;8:36. doi:10.4103/jpi.jpi_32_17
13. Tanaka N, Kanatani S, Tomer R, et al. Whole-tissue biopsy phenotyping of three-dimensional tumours reveals patterns of cancer heterogeneity. Nat Biomed Eng. 2017;1(10):796-806. doi:10.1038/s41551-017-0139-0
14. Shirinifard A, Thiagarajan S, Vogel P, Sablauer A. Detection of Phenotypic Alterations Using High-Content Analysis of Whole-Slide Images. J Histochem Cytochem. 2016;64(5):301-310. doi:10.1369/0022155416639884
15. Gurcan MN, Boucheron LE, Can A, Madabhushi A, Rajpoot NM, Yener B. Histopathological Image Analysis: A Review. IEEE Rev Biomed Eng. 2009;2:147-171. doi:10.1109/RBME.2009.2034865
16. Zarella MD, Bowman D, Aeffner F, et al. A practical guide to whole slide imaging a white paper from the digital pathology association. Arch Pathol Lab Med. 2019;143(2):222-234. doi:10.5858/arpa.2018-0343-RA
17. Aeffner F, Zarella MD, Buchbinder N, et al. Introduction to Digital Image Analysis in Whole-slide Imaging: A White Paper from the Digital Pathology Association. J Pathol Inform. 2019;10:9. doi:10.4103/jpi.jpi_82_18
18. Amisha, Malik P, Pathania M, Rathaur VK. Overview of artificial intelligence in medicine. J Fam Med Prim care. 2019;8(7):2328-2331. doi:10.4103/jfmpc.jfmpc_440_19
19. Acs B, Rantalainen M, Hartman J. Artificial intelligence as the next step towards precision pathology. J Intern Med. 2020;288(1):62-81. doi:10.1111/joim.13030
20. Sakamoto T, Furukawa T, Lami K, et al. A narrative review of digital pathology and artificial intelligence: Focusing on lung cancer. Transl Lung Cancer Res. 2020;9(5):2255-2276. doi:10.21037/tlcr-20-591
21. Serag A, Ion-Margineanu A, Qureshi H, et al. Translational AI and Deep Learning in Diagnostic Pathology. Front Med. 2019;6(October):1-15. doi:10.3389/fmed.2019.00185