Digital pathology

File:Major topics of pathology informatics.png, and encompasses topics including slide scanning, digital imaging, image analysis and telepathology.]]

Digital pathology is a sub-field of pathology that focuses on managing and analyzing information generated from digitized specimen slides. It utilizes computer-based technology and virtual microscopy to view, manage, share, and analyze digital slides on computer monitors.{{cite journal | vauthors = Pantanowitz L | title = Twenty Years of Digital Pathology: An Overview of the Road Travelled, What is on the Horizon, and the Emergence of Vendor-Neutral Archives | journal = Journal of Pathology Informatics | year=2018 | volume = 9 | page = 40 | doi = 10.4103/jpi.jpi_69_18 | pmid = 30607307| pmc = 6289005 | doi-access = free }} This field has applications in diagnostic medicine and aims to achieve more efficient and cost-effective diagnoses, prognoses, and disease predictions through advancements in machine learning and artificial intelligence in healthcare.{{Cite web|url=https://www.mbfbioscience.com/whole-slide-imaging|title=Whole Slide Imaging {{!}} MBF Bioscience|website=www.mbfbioscience.com|access-date=2019-12-02}}{{Cite journal |last=Saillard |first=Charlie |last2=Dubois |first2=Rémy |last3=Tchita |first3=Oussama |last4=Loiseau |first4=Nicolas |last5=Garcia |first5=Thierry |last6=Adriansen |first6=Aurélie |last7=Carpentier |first7=Séverine |last8=Reyre |first8=Joelle |last9=Enea |first9=Diana |last10=von Loga |first10=Katharina |last11=Kamoun |first11=Aurélie |last12=Rossat |first12=Stéphane |last13=Wiscart |first13=Corentin |last14=Sefta |first14=Meriem |last15=Auffret |first15=Michaël |date=2023-11-06 |title=Validation of MSIntuit as an AI-based pre-screening tool for MSI detection from colorectal cancer histology slides |url=https://pubmed.ncbi.nlm.nih.gov/37932267/ |journal=Nature Communications |volume=14 |issue=1 |pages=6695 |doi=10.1038/s41467-023-42453-6 |issn=2041-1723 |pmid=37932267|pmc=10628260 }}

History

The roots of digital pathology trace back to the 1960s with early telepathology experiments. The concept of virtual microscopy emerged in the 1990s across various areas of life science research.{{cite journal|last1=Ferreira|first1=R|last2=Moon|first2=J|last3=Humphries|first3=J|last4=Sussman|first4=A|last5=Saltz|first5=J|last6=Miller|first6=R|last7=Demarzo|first7=A|year=1997|title=The virtual microscope|journal=Romanian Journal of Morphology and Embryology|volume=45|pages=449–453|pmc=2233368|pmid=9357666}} At the turn of the century the scientific community more and more agreed on the term "digital pathology" to denote digitization efforts in pathology. However, in 2000, the technical requirements (scanner, storage, network) were still a limiting factor for a broad dissemination of digital pathology concepts. This changed as new powerful and affordable scanner technology as well as mass / cloud storage technologies appeared on the market. The field of radiology has undergone the digital transformation almost 15 years ago, not because radiology is more advanced, but there are fundamental differences between digital images in radiology and digital pathology: The image source in radiology is the (alive) patient, and today in most cases, the image is even primarily captured in digital format. In pathology the scanning is done from preserved and processed specimens, for retrospective studies even from slides stored in a biobank. Besides this difference in pre-analytics and metadata content, the required storage in digital pathology is two to three orders of magnitude higher than in radiology. However, the advantages anticipated through digital pathology are similar to those in radiology:

Capability to transmit digital slides over distances quickly, which enables telepathology scenarios.
Capability to access past specimen from the same patients and similar cases for comparison and review, with much less effort than retrieving slides from the archive shelves.
Capability to compare different areas of multiple slides simultaneously (slide by slide mode) with the help of a virtual microscope.
Capability to annotate areas directly in the slide and share this for teaching and research.

Digital pathology is today widely used for educational purposes{{cite journal|last1=Hamilton|first1=Peter W.|last2=Wang|first2=Yinhai|last3=McCullough|first3=Stephen J.|last4=Sussman|s2cid=20599493|year=2012|title=Virtual microscopy and digital pathology in training and education|journal=APMIS|volume=120|issue=4|pages=305–315|doi=10.1111/j.1600-0463.2011.02869.x|pmid=22429213|doi-access=free}} in telepathology and teleconsultation as well as in research projects. Digital pathology allows to share and annotate slides in a much easier way and to download annotated lecture sets generates new opportunities for e-learning and knowledge sharing in pathology. Digital pathology in diagnostics is an emerging and upcoming field.

Environment

=Scan=

File:Leica microscopy slide scanner, annotated.jpg

File:Whole slide image quality comparison.png

Digital slides are created from glass slides using specialized scanning machines. All high quality scans must be free of dust, scratches, and other obstructions. There are two common methods for digital slide scanning, tile-based scanning and line-based scanning.{{cite web|title=Informatics, digital & computational pathology|url=https://www.pathologyoutlines.com/topic/informaticswholeslidefund.html}} Both technologies use an integrated camera and a motorized stage to move the slide around while parts of the tissue are imaged. Tile scanners capture square field-of-view images covering the entire tissue area on the slide, while line-scanners capture images of the tissue in long, uninterrupted stripes rather than tiles. In both cases, software associated with the scanner stitch the tiles or lines together into a single, seamless image.

Z-stacking is the scanning of a slide at multiple focal planes along the vertical z-axis.{{cite journal| vauthors=Evans AJ, Salama ME, Henricks WH, Pantanowitz L| title=Implementation of Whole Slide Imaging for Clinical Purposes: Issues to Consider From the Perspective of Early Adopters. | journal=Arch Pathol Lab Med | year= 2017 | volume= 141 | issue= 7 | pages= 944–959 | pmid=28440660 | doi=10.5858/arpa.2016-0074-OA | pmc= | doi-access=free }}

=View=

Digital slides are accessible for viewing via a computer monitor and viewing software either locally or remotely via the Internet. An example of an open-source, web-based viewer for this purpose implemented in pure JavaScript, for desktop and mobile, is the OpenSeadragon[https://openseadragon.github.io/ OpenSeadragon] on GitHub viewer. QuPath[https://qupath.github.io/ QuPath] on GitHub is another such open source software, which is often used for digital pathology applications because it offers a powerful set of tools for working with whole slide images. OpenSlide,[https://openslide.org/ OpenSlide website] on the other hand is a C library (Python and Java bindings are also available) that provides a simple interface to read and view whole-slide images.

=Manage=

Digital slides are maintained in an information management system that allows for archival and intelligent retrieval.

=Network=

Digital slides are often stored and delivered over the Internet or private networks, for viewing and consultation.

= Analyze =

Image analysis tools are used to derive objective quantification measures from digital slides. Image segmentation and classification algorithms, often implemented using deep neural networks, are used to identify medically significant regions and objects on digital slides. A GPU acceleration software for pathology imaging analysis, cross-comparing spatial boundaries of a huge amount of segmented micro-anatomic objects has been developed.{{Cite conference |author= Kaibo Wang|author2= Yin Huai|author3= Rubao Lee|author4=Fusheng Wang|author5=Xiaodong Zhang|author6= Joel H. Saltz| title="Accelerating Pathology Image Data Cross-Comparison on CPU-GPU Hybrid Systems" | journal=Proceedings of the VLDB Endowment. International Conference on Very Large Data Bases |conference=Proceedings VLDB Endowment| year=2012 | volume=5 | issue=11 |pages =1543–1554| pmid=23355955 | pmc=3553551 }} The core algorithm of PixelBox in this software has been adopted in Fixstars' Geometric Performance Primitives (GPP) library{{Cite web|url=https://developer.nvidia.com/geometric-performance-primitives-gpp |title=Geometric Performance Primitives (GPP)|website=NVIDIA Developer}} as a part of NVIDIA Developer, which is a production geometry engine for advanced graphical information systems, electronic design automation, computer vision and motion planning solutions.{{cite web|url=http://web.cse.ohio-state.edu/~zhang.574/Ohio_State_U_ack-letter.pdf|title=Letter to Kaibo Wang}}

File:Ki67 calculation by QuPath in a pure seminoma.png|Ki67 stain calculation by QuPath in a pure seminoma, which gives a measure of the proliferation rate of the tumor. The colors represent the intensity of expression: blue-no expression, yellow-low, orange-moderate, and red-high expression.{{cite journal| vauthors=Lourenço BC, Guimarães-Teixeira C, ((Flores BCT)), Miranda-Gonçalves V, Guimarães R, Cantante M | display-authors=etal| title=Ki67 and LSD1 Expression in Testicular Germ Cell Tumors Is Not Associated with Patient Outcome: Investigation Using a Digital Pathology Algorithm. | journal=Life | year= 2022 | volume= 12 | issue= 2 | page=264| pmid=35207551 | doi=10.3390/life12020264 | pmc=8875543 | doi-access=free}}
- [https://www.researchgate.net/figure/Illustrative-examples-of-Ki67-and-LSD1-immunoexpression-across-different-testicular-germ_fig2_358555198 Figure 2 - available via license: Creative Commons Attribution 4.0 International]

File:Bone marrow cellularity in QuPath.png|Tissue segmentation for digital calculation of bone marrow cellularity in QuPath: The system is trained on the appearance of immune cells versus other tissue, and uses this to give an overall percentage of each type.

File:Histopathology of tumor identification by AI.png|Breast cancer prediction by AI.{{cite journal| vauthors=Cruz-Roa A, Gilmore H, Basavanhally A, Feldman M, Ganesan S, ((Shih NNC)) | display-authors=etal| title=Accurate and reproducible invasive breast cancer detection in whole-slide images: A Deep Learning approach for quantifying tumor extent. | journal=Sci Rep | year= 2017 | volume= 7 | issue= | pages= 46450 | pmid=28418027 | doi=10.1038/srep46450 | pmc=5394452 | url=https://www.ncbi.nlm.nih.gov/entrez/eutils/elink.fcgi?dbfrom=pubmed&tool=sumsearch.org/cite&retmode=ref&cmd=prlinks&id=28418027 }}
"This work is licensed under a Creative Commons Attribution 4.0 International License."

class=wikitable

|File:Simplified neural network training on cells.svg Simplified example of training a neural network in cytologic object detection: The network is trained by multiple images that are known to depict benign cells (upper left) and cancer cells (lower left), which are correlated with "nodes" that represent visual aspects, in this case nuclear size and chromatin pattern. The benign cells match with small nuclei and finely granular chromatin, whereas most cancer cells match with large nuclei and coarsely granular chromatin. However, the instance of a cancer cell with fine chromatin creates a weakly weighted association between them.

File:Simplified neural network on cells.svg Subsequent run of the network on an input image (left): The network correctly detects the benign cell. However, the weakly weighted association between fine chromatin and cancer cells also confers a weak signal to the latter from one of two intermediate nodes. In addition, a blood vessel (bottom left) that was not included in the training partially conforms to the patterns of large nuclei and coarse chromatin, and therefore results in weak signals for the cancer cell output. These weak signals may result in a false positive result for a cancer cell.

=Integrate=

Digital pathology workflow is integrated into the institution's overall operational environment. Slide digitization is expected to reduce the number of routine, manually reviewed slides, maximizing workload efficiency.

=Sharing=

Digital pathology also allows internet information sharing for education, diagnostics, publication and research. This may take the form of publicly available datasets or open source access to machine learning algorithms.

Challenges

File:Folding artifact on whole slide imaging of bone.png tissue is particularly prone to folding artifacts. In this micrograph, the automatic camera is focused on a fold (left in image), resulting in defocus aberration (blur) of the surrounding tissue (right in image).]]

File:CD117 (c-kit) stain of mixed malignant germ cell tumor - crop.png

Digital pathology has been approved by the FDA for primary diagnosis.{{cite press release|title=FDA allows marketing of first whole slide imaging system for digital pathology|url=https://www.fda.gov/news-events/press-announcements/fda-allows-marketing-first-whole-slide-imaging-system-digital-pathology|publisher=FDA|date=April 12, 2017|access-date=May 24, 2017}} The approval was based on a multi-center study of 1,992 cases in which whole-slide imaging (WSI) was shown to be non-inferior to microscopy across a wide range of surgical pathology specimens, sample types and stains.{{cite journal|last1=Mukhopadhyay|first1=Sanjay|last2=Feldman|first2=Michael|last3=Abels|first3=Esther|year=2017|title=Whole slide imaging versus microscopy for primary diagnosis in surgical pathology: a multicenter randomized blinded noninferiority study of 1992 cases (pivotal study).|journal=American Journal of Surgical Pathology|volume=42|issue=1|pages=39–52|doi=10.1097/PAS.0000000000000948|pmc=5737464|pmid=28961557}} While there are advantages to WSI when creating digital data from glass slides, when it comes to real-time telepathology applications, WSI is not a strong choice for discussion and collaboration between multiple remote pathologists.{{cite journal|last1=Siegel|first1=Gabriel|last2=Regelman|first2=Dan|last3=Maronpot|first3=Robert|last4=Rosenstock|first4=Moti|last5=Hayashi|first5=Shim-mo|last6=Nyska|first6=Abraham|date=Oct 2018|title=Utilizing novel telepathology system in preclinical studies and peer review|journal=Journal of Toxicologic Pathology|volume=31|issue=4|pages=315–319|doi=10.1293/tox.2018-0032|pmc=6206289|pmid=30393436}} Furthermore, unlike digital radiology where the elimination of film made return on investment (ROI) clear, the ROI on digital pathology equipment is less obvious. The strongest ROI justification includes improved quality of healthcare, increased efficiency for pathologists, and reduced costs in handling glass slides.{{cite web|url=http://www.sectra.com/medical/pathology/resources/articles/How_to_build_digital_pathology_business_case.html|title=How to Build a Business Case to Justify the Investment in Digital Pathology|publisher=Sectra Medical Systems|access-date=April 26, 2015}}

= Validation =

Validation of a digital microscopy workflow in a specific environment (see above) is important to ensure high diagnostic performance of pathologists when evaluating digital whole-slide images. There are different methods that can be used for this validation process.{{cite journal|last1=Bertram|first1=Christof A|last2=Stathonikos|first2=Nikolas|last3=Donovan|first3=Taryn A|last4=Bartel|first4=Alexander|last5=Fuchs-Baumgartinger|first5=Andrea|last6=Lipnik|first6=Karoline|last7=can Diest|first7=Paul J|last8=Bonsembiante|first8=Federico|last9=Klopfleisch|first9=Robert|date=September 2021|title=Validation of digital microscopy: Review of validation methods and sources of bias|journal=Veterinary Pathology|volume=59|issue=1|pages=26–38|doi=10.1177/03009858211040476|pmid=34433345|pmc=8761960}} The College of American Pathologists has published a guideline with minimal requirements for validation of whole slide imaging systems for diagnostic purposes in human pathology.{{cite journal|last1=Evans|first1=Andrew J|last2=Brown|first2=Richard|last3=Bui|first3=Marilyn|title=Validating whole slide imaging systems for diagnostic purposes in pathology: guideline update from the College of American Pathologists in collaboration with the American Society for Clinical Pathology and the Association for Pathology Informatics|journal=Arch Pathol Lab Med|year=2021|volume=146 |issue=4 |pages=440–450 |doi=10.5858/arpa.2020-0723-CP|pmid=34003251|doi-access=free}}

Potential

Trained pathologists traditionally view tissue slides under a microscope. These tissue slides may be stained to highlight cellular structures. When slides are digitized, they are able to be shared through tele-pathology and are numerically analyzed using computer algorithms. Algorithms can be used to automate the manual counting of structures, or for classifying the condition of tissue such as is used in grading tumors. They can additionally be used for feature detection of mitotic figures, epithelial cells, or tissue specific structures such as lung cancer nodules, glomeruli, or vessels, or estimation of molecular biomarkers such as mutated genes, tumor mutational burden, or transcriptional changes.{{Cite journal|last1=Aeffner|first1=Famke|last2=Zarella|first2=Mark D.|last3=Buchbinder|first3=Nathan|last4=Bui|first4=Marilyn M.|last5=Goodman|first5=Matthew R.|last6=Hartman|first6=Douglas J.|last7=Lujan|first7=Giovanni M.|last8=Molani|first8=Mariam A.|last9=Parwani|first9=Anil V.|last10=Lillard|first10=Kate|last11=Turner|first11=Oliver C.|date=2019-03-08|title=Introduction to Digital Image Analysis in Whole-slide Imaging: A White Paper from the Digital Pathology Association|journal=Journal of Pathology Informatics|volume=10|pages=9|doi=10.4103/jpi.jpi_82_18|issn=2229-5089|pmc=6437786|pmid=30984469 |doi-access=free }}{{Cite journal|last1=Jain|first1=Mika S.|last2=Massoud|first2=Tarik F.|date=2020|title=Predicting tumour mutational burden from histopathological images using multiscale deep learning|url=https://www.nature.com/articles/s42256-020-0190-5|journal=Nature Machine Intelligence|language=en|volume=2|issue=6|pages=356–362|doi=10.1038/s42256-020-0190-5|s2cid=220510782|issn=2522-5839}}{{Cite journal|last1=Serag|first1=Ahmed|last2=Ion-Margineanu|first2=Adrian|last3=Qureshi|first3=Hammad|last4=McMillan|first4=Ryan|last5=Saint Martin|first5=Marie-Judith|last6=Diamond|first6=Jim|last7=O'Reilly|first7=Paul|last8=Hamilton|first8=Peter|date=2019|title=Translational AI and Deep Learning in Diagnostic Pathology|journal=Frontiers in Medicine|volume=6|pages=185|doi=10.3389/fmed.2019.00185|pmid=31632973|pmc=6779702|issn=2296-858X|doi-access=free}} This has the potential to reduce human error and improve accuracy of diagnoses. Digital slides can be easily shared, increasing the potential for data usage in education as well as in consultations between expert pathologists. Multiplexed imaging (staining multiple markers on the same slide) allows pathologists to understand finer distribution of cell-types and their relative locations.{{Cite journal |last1=Nirmal |first1=Ajit J. |last2=Maliga |first2=Zoltan |last3=Vallius |first3=Tuulia |last4=Quattrochi |first4=Brian |last5=Chen |first5=Alyce A. |last6=Jacobson |first6=Connor A. |last7=Pelletier |first7=Roxanne J. |last8=Yapp |first8=Clarence |last9=Arias-Camison |first9=Raquel |last10=Chen |first10=Yu-An |last11=Lian |first11=Christine G. |date=2022-04-11 |title=The spatial landscape of progression and immunoediting in primary melanoma at single cell resolution |journal=Cancer Discovery |volume=12 |issue=6 |pages=1518–1541 |doi=10.1158/2159-8290.CD-21-1357 |issn=2159-8290 |pmid=35404441|pmc=9167783 }} An understanding of the spatial distribution of cell-types or markers and pathways they express, can allow for prescription of targeted drugs or build combinational therapies in a personalized manner.