Journal of Pathology Informatics Journal of Pathology Informatics
Contact us | Home | Login   |  Users Online: 1832  Print this pageEmail this pageSmall font sizeDefault font sizeIncrease font size 




 
Table of Contents    
SYMPOSIUM INTERNATIONAL ACADEMY OF DIGITAL PATHOLOGY (IADP)
J Pathol Inform 2015,  6:29

A perspective on digital and computational pathology


Center for Biophysical Pathology, Rutgers-NJMS, Newark, NJ 07103, USA

Date of Submission12-Mar-2015
Date of Acceptance19-Mar-2015
Date of Web Publication03-Jun-2015

Correspondence Address:
Stanley Cohen
Center for Biophysical Pathology, Rutgers-NJMS, Newark, NJ 07103
USA
Login to access the Email id

Source of Support: None, Conflict of Interest: None


DOI: 10.4103/2153-3539.158059

Rights and Permissions
   Abstract 

The digitization of images has not only led to increasingly sophisticated methods of quantitating information from those images themselves, but also to the development of new physics-based techniques for extracting information from the original specimen and presenting this as visual data in both two and three-dimensional (3D) forms. This evolution of an image-based discipline has reached maturity in Radiology, but it is only just beginning in Pathology. An historical perspective is provided both on the current state of computational imaging in pathology and of the factors that are impeding further progress in the development and application of these approaches. Emphasis is placed on barriers to the dissemination of information in this area. The value of computational imaging in basic and translational research is clear. However, while there are many examples of "virtual diagnostics" in Radiology, there are only relatively few in Pathology. Nevertheless, we can do cellular level analysis of lesions accessible by endoscopic or catheterization procedures, and a number of steps have been taken toward real-time imaging as adjuncts to traditional biopsies. Progress in computational imaging will greatly expand the role of pathologists in clinical medicine as well as research.

Keywords: Computational imaging, confocal-based imaging, pathology-radiology convergence, super-resolution, virtual biopsy


How to cite this article:
Ramamurthy B, Coffman FD, Cohen S. A perspective on digital and computational pathology. J Pathol Inform 2015;6:29

How to cite this URL:
Ramamurthy B, Coffman FD, Cohen S. A perspective on digital and computational pathology. J Pathol Inform [serial online] 2015 [cited 2019 Nov 18];6:29. Available from: http://www.jpathinformatics.org/text.asp?2015/6/1/29/158059


   Introduction Top


This discussion will examine the evolution of digital pathology to embrace all forms of computational imaging, with a consideration of possible impediments for the availability of this technology for diagnostic and experimental pathology. One such impediment is a lack of awareness by pathologists of many of the newer modalities that are emerging in this field. Another is a lack of integration among the many relatively small societies that serve as conduits for association, dissemination, and advocacy in the field. A third is the relative lack of journals that embrace sophistication both in the development of new technologies and the use of those modalities to solve experimental problems, with content accessible to both biomedical investigators and the engineers, physicists, and mathematicians that are essential for synergistic progress. Many authors gravitate toward engineering journals, which are not usually read by pathologists or, for that matter, most biomedical investigators. Often, the pathobiological application shown is a proof-of-concept demonstration rather than a sophisticated investigative study. Papers that appear in pathology journals have the opposite problem; they are sketchy about the physical/engineering side of the work, and do not attract many readers from the engineering and physics community. Perhaps the simplest way of summarizing these interrelated issues is by considering four stages of scholarly activity; the creation, communication and dissemination, storage, and novel utilization of knowledge. The choke point for our current endeavors seems to be at the stage of communication and dissemination, which limits collaboration and cross-pollination.

Digital pathology has been defined as "a dynamic image-based environment that enables the acquisition, management, and interpretation of pathology information obtained from a glass slide" (Digital Pathology Association). The slowly evolving change to this modality is comparable to the replacement of X-ray films by digital images. Progress in pathology has been much slower for many reasons, such as the storage problems involved with color images and the additional computation and the storage required to provide depth imaging so that the pathologist can focus up and down on the virtual side. In other words, information density is much higher in traditional pathology than in traditional radiology. This distinction will no longer be valid not only because of increased computational power and storage but also by the fact that the (no longer) new technologies in radiology are as computationally intensive as those that are emerging for pathology. There are two other practical considerations. First, in radiology images are now captured directly digitally, whereas pathologists currently scan glass slides prepared and stained in the traditional manner. This burden, however, is counterbalanced by the fact that advances in both computers and computational methods allow the performance of increasingly sophisticated advanced image analysis so that the computer becomes the pathologist's "partner" in the diagnostic and investigational milieu. The second factor is a financial one. New technologies in radiology have directly resulted in new revenue for hospitals, thereby justifying the expenses involved, whereas the return on investment is less clear for pathology at present. We must demonstrate that there is a secondary gain due to better diagnostic and prognostic information we can obtain, but also that there can be a direct return on investment. With respect to grant availability, the current paradigm of hypothesis testing as a major criterion for publication is an impediment for both publication and grants for a field in which the hypotheses relate to the technological, and where the biological applications are currently observational. Observation is not a dirty word. Progress in science depends on an interplay between observation, hypothesis, and theory. Without observation, there could be no astronomy and little theoretical physics. In other words, hypothesis generation (the acquisition of new knowledge used to construct hypotheses) is as important as hypothesis testing. The arrow between the two should point in both directions. When this is fully understood, computational pathology will flourish.


   Emergence of Computational Imaging in Radiology and Pathology Top


As seen above, digitization of images opens up a whole new world of computational analysis. The converse is also true, namely that computational analysis opens up a whole new world of digital imaging. Photonic interactions with a biological specimen involve a small number of parameters, namely transmission, absorption, reflection, secondary emission, diffraction and scatter. The power of computational imaging is that it can optimize the imaging potential of these parameters while minimizing deleterious effects (for example diffraction).

From an historical perspective, it is clear that the emergence of magnetic resonance imaging (MRI) and computer-assisted tomography (CAT) have revolutionized radiology. Conventional tomography, or planography was developed in the 1930s and was used extensively in clinical medicine through the early 1970s. In this technique, a sectional image is made by moving an X-ray source and film in opposite directions during the exposure. This makes subjects in the focal plane appear sharper while structures in other planes appear blurred. This approach was of limited utility at the time. However, with the advent of powerful computers, it became possible to obtain data from multiple directions and use reconstruction software to create extremely sharp images in multiple planes, which may be viewed as two-dimensional slices or 3D reconstructions. [1] In many ways, this use of computer processing to sum multiple images was the ancestor of computational imaging. MRI, in contrast (pun intended), makes use of a strong magnetic field to align proton spin (primarily but not always) in the hydrogen atoms of water molecules. A right angle radiofrequency (RF) pulse is then used to perturb this state. The return to equilibrium, when the RF stops (relaxation) is accompanied by an induced RF signal from the nuclei. Measurement of these signals at multiple points in the subject is used to reconstruct an image. In this technique, then, the computer is used both to convert an induced RF to a visible signal and to analyze and manipulate the multiple signals to construct an image. [2] This represents a higher level of computational analysis than CAT.

In pathology, one of the earliest forms of computational imaging was confocal microscopy, and in particular, confocal laser scanning microscopy (CSLM) reviewed in. [3] Traditional widefield epifluorescent microscopy is hampered in its ability to provide fine detail, because of limits imposed by diffraction, and is also hampered in its ability to provide 3D information because of light scattering. In a CSLM, the specimen is excited by laser light focused to a diffraction-limited spot within that specimen. The emitted light from the spot is separated from the exciting laser light and then passed through a confocal pinhole that rejects light generated outside the focal plane as well as scattered light. The image is created by scanning the diffraction-limited spot in three dimensions. The emitted light is detected, and the resulting dataset is reconstructed to create an image by computer, utilizing similar reconstruction algorithms to those described above.

Another well-established technique is optical coherence tomography (OCT) which saw initial use in ophthalmology and is now a tool for radiologists and pathologists, and directly by clinical interventionalists. [4] OCT uses light to capture images from within optically scattering media such as tissue. It uses relatively long wavelength (near-infrared) light to allow penetration and utilizes backscatter to obtain information. Micrometer resolution is obtained but only for a depth of a few millimeters into the tissue.


   Recent Advances in Confocal Laser Scanning Microscopy and Optical Coherence Tomography Top


0Just as confocal microscopy evolved from traditional fluorescence microscopy and computed tomography evolved from crude superimposition of image planes, these modalities themselves have rapidly evolved. As this communication is meant to provide a broad overview and perspective on this rapidly emerging field rather than a scholarly review per se, with a few exceptions, no primary references are cited. Instead, reference is made to monographs and/or websites where appropriate. Some examples of advances in confocal microscopy include spinning disk, multi-photon, total internal reflectance, and lateral sheet illumination microscopy. Confocal based techniques allow super-resolution microscopy via various techniques (APLM STORM, STED, etc.) Light sheet illumination microscopy illuminates the tissue with a laser beam that is projected on the sample as a thin sheet of light parallel to the focal plane. This allows for high-resolution reconstruction of the sample in three dimensions. These techniques are described in. [3] Another recent technique is spatial light interference microscopy that combines phase contrast and holography with topographic accuracy comparable to that of atomic force microscopy. [5] There have also been advances in OCT such as Angle Resolved Low Coherence Interferometry, which allow quantitative measurements of size and texture of subcellular structures [6] as well as ultra-high resolution spectral domain OCT and Doppler tomography. OCT has also been combined with other modalities, such as photoacoustic tomography.


   Raman Imaging Top


Although we usually consider light scatter the enemy of visual observation and resolution, light scattering can be our friend as well. This is best seen in Raman spectroscopy and Raman imaging. When a beam of photons transverses a specimen, the bulk of light scattering is via Rayleigh scattering where the scattered photons have the same energy as the incident photons. Raman scattering, in contrast, results from inelastic interactions with vibrating molecules (most intensely in the region of double bonds) and results in a small number of scattered photons with a frequency different (usually lower) than the incident photons. The resulting Raman spectra provide characteristic molecular fingerprints and allow chemical analysis of intact cells. Recently, Raman information has been used to construct images through computational approaches analogous to those used in the modalities described above, [7] the most useful being intensity maps of specific molecular species in a cell in three dimensions enhanced by mathematical deconvolution. [8]


   Applications in Diagnostic Pathology Top


The power of computational imaging in basic and translational research is obvious. Additionally, a number of laboratories are making great strides in the application of these methods to assist in diagnostic decision-making, with an ultimate evolution toward a complete virtual biopsy. Current examples include OCT and confocal microscopic analysis via endoscopy using either tethered capsules, or optical probes linked to external instrumentation. These approaches are dependent on advances in miniaturization and resolution of existing techniques as well as the emergence of novel techniques. Tearney have developed micro-OCT and utilized this for studies of airway microanatomay via bronchosopy, coronary atherosclerosis via cardiac catheterization, [9] as well as endoscopic studies of the gastrointestinal tract. This lab has also pioneered in the development of multiple micro-confocal techniques.

While the above techniques are feasible where probes can be placed in proximity to lesions, it is not yet possible to obtain cellular resolution at diagnostically useful depths within tissue for true virtual biopsies. Nevertheless, advanced imaging techniques have already been shown to be useful in computationally assisted biopsies. For example, the use of micro-computed tomography for tumor margins in breast resections is being developed at the MGH. [10] Additionally, there have been recent developments in high-resolution MRI, where the limitations to resolution include beam strength and magnetic strength. With new developments in both pulse generation and relaxation detection, these limitations may be overcome in clinical settings. In the laboratory, MRI has been able to achieve nanometer resolution (!) using ATF probes as antennas and doped diamond lattices as detectors. The tricorder used by the crew of the Enterprise in Star Trek may not be that far out of reach in the near future. However, the question remains as to whose hand will hold the tricorder. It seems clear, from the various examples presented above that the disciplines of radiology and pathology are converging toward a new specialty, with a head start for radiology. Unless pathologists accelerate the pace of embracing new imaging modalities to provide better diagnostic and prognostic data for clinical use, this new entity will be called "Radiology" and anatomic pathologists will become part of a radiological subspecialty called "cellular imaging." This may or may not be a bad thing in terms of hospital support. However, we would rather see an entire new discipline, possibly called "diagnostic imaging" emerge, which would allow radiologists and pathologists to retain their professional identities within it. In either event "proof of concept" for convergence can already be seen in the increasing ability of radiologists to perform molecular imaging and the pathologist's ability to image and detect molecular interactions.

 
   References Top

1.
Suetens P. X-ray computed tomography. In: Fundamentals of Medical Imaging. London: Cambridge University Press; 2002. p. 66-98.  Back to cited text no. 1
    
2.
Suetens P. Magnetic resonance imaging, Ibid: 99-144.  Back to cited text no. 2
    
3.
Fritzky L, Lagunoff D. Advanced methods in fluorescence microscopy. In: Biophotonics in Pathology. Amsterdam:S. Cohen, IOS Press; 2013. p. 23-42.  Back to cited text no. 3
    
4.
Jung W, Boppart SA. Optical coherence tomography for rapid tissue screening and directed histological sectioning. Ibid, 109-128.  Back to cited text no. 4
    
5.
Wang Z, Millet L, Chan V, Ding H, Gillette MU, Bashir R, et al. Label-free intracellular transport measured by spatial light interference microscopy. J Biomed Opt 2011;16:026019.  Back to cited text no. 5
    
6.
Wax A, Chalut KJ. Nuclear morphology measurements with angle-resolved low coherence interferometry for application to cell biology and early cancer detection. In: Biophotonics in Pathology. Amsterdam: S Cohen, IOS Press; 2013. p. 129-52.  Back to cited text no. 6
    
7.
Smith CJ, Huser TR, Wachsmann-Hogiu S. Raman scattering in pathology, Ibid: 207-234.  Back to cited text no. 7
    
8.
Lau R, Ramamurthy B, Coffman F. A label-free investigation of cellular changes in MG-63 osteosarcoma cells undergoing autophagy using Raman spectroscopic imaging. June 30-July 3, 2014; Microscience Microscopy Congress, Manchester UK, abstract.  Back to cited text no. 8
    
9.
Liu L, Gardecki JA, Nadkarni SK, Toussaint JD, Yagi Y, Bouma BE, et al. Imaging the subcellular structure of human coronary atherosclerosis using micro-optical coherence tomography. Nat Med 2011;17;1010-4.  Back to cited text no. 9
    
10.
R, Buckley JM, Fernandez L, Coopey S, Aftreth O, Michaelson J, et al. Micro-computed tomography (Micro-CT): A novel approach for intraoperative breast cancer specimen imaging. Breast Cancer Res Treat 2013;139:311-6.  Back to cited text no. 10
[PUBMED]    




 

 
Top
  

    

 
  Search
 
   Browse articles
  
    Similar in PUBMED
   Search Pubmed for
   Search in Google Scholar for
 Related articles
    Access Statistics
    Email Alert *
    Add to My List *
* Registration required (free)  

 
  In this article
    Abstract
   Introduction
    Emergence of Com...
    Recent Advances ...
   Raman Imaging
    Applications in ...
    References

 Article Access Statistics
    Viewed1488    
    Printed28    
    Emailed1    
    PDF Downloaded365    
    Comments [Add]    

Recommend this journal