Journal of Pathology Informatics Journal of Pathology Informatics
Contact us | Home | Login   |  Users Online: 2265  Print this pageEmail this pageSmall font sizeDefault font sizeIncrease font size 

Table of Contents    
J Pathol Inform 2011,  2:26

Why a pathology image should not be considered as a radiology image

1 Department of Pathology,University of Michigan, University of Michigan Health System, M4233A Medical Science I, 1301 Catherine, Ann Arbor, Michigan 48109-0602, USA
2 Health - Bioinformatics, Booz Allen Hamilton, One Preserve Parkway, Suite 200, Rockville, MD 20852, USA
3 Office of Biorepositories and Biospecimen Research, National Cancer Institute, NIH, 31 Center Dr., Suite 10 A03, Bethesda, MD20892, USA

Date of Submission22-Mar-2011
Date of Acceptance26-Apr-2011
Date of Web Publication14-Jun-2011

Correspondence Address:
Jason D Hipp
Department of Pathology,University of Michigan, University of Michigan Health System, M4233A Medical Science I, 1301 Catherine, Ann Arbor, Michigan 48109-0602
Login to access the Email id

Source of Support: None, Conflict of Interest: None

DOI: 10.4103/2153-3539.82051

Rights and Permissions

How to cite this article:
Hipp JD, Fernandez A, Compton CC, Balis UJ. Why a pathology image should not be considered as a radiology image. J Pathol Inform 2011;2:26

How to cite this URL:
Hipp JD, Fernandez A, Compton CC, Balis UJ. Why a pathology image should not be considered as a radiology image. J Pathol Inform [serial online] 2011 [cited 2022 Jul 2];2:26. Available from:

   A Pathology Image is not a Radiology Image Top

Often, the emergence of unified and seamless integration of digital images within contemporary radiology workflow models is held as the exemplar by which any possible future states of an all-digital workflow model in pathology should be compared. Indeed, there is strong evidence to suggest that pathology will ultimately transform into what could be properly termed as a 'digital diagnostic modality', but it is important to distinguish between the prior transformative process within radiology, with its own set of operational challenges (elimination of film, modality workflow management, Digital Imaging and Communications in Medicine (DICOM) harmonization, antitrust, and proprietary file format issues, to name a few), and those challenges that now face pathology, with there being only partial overlap.

As background, radiology did not transform into a digital specialty in a single, uninterrupted transition period, but rather, adopted incremental facets of its present all-digital workflow, based upon a continuously evolving mélange of maturing technologies (DICOM, computed radiography detector arrays, high speed networks, etc.). This process started in 1978, and only recently entered a stage that might be labeled as mature (but just barely).

Conversely, all digital workflow in pathology remains an ephemeral target, with only serious pilot attempts at attaining department-wide deployment being possible within the last two years. Underlying this transformation is the continuing evolution of Whole Slide Imaging (WSI) scanners, which allow for the imaging of entire tissue sections, thereby producing digital slide data sets. On account of recent improvements in the speed with which these appliances capture entire slides and similar improvements in their associated resolution and image quality, the use of WSI technology is making stepwise inroads into routine surgical pathology workflow. However, there remains persistent hesitation in the pathology community with respect to its adoption of total digital histopathology, owing to: expense associated with the storage of large digital images, user comfort level, and the potential time delay for diagnosis represented by the additional necessary step of scanning slides. Despite these factors, increasing numbers of pathology departments are willing to consider deploying all-digital solutions, recognizing the advances and improvements in workflow that digital radiology has already achieved, with the hope being that similar improvement opportunities await pathology. Although there are many parallels between the two specialties' use of digital technology, there are also many significant differences in their constitutive modalities, thus making direct comparison an imperfect proposition. Here, we will identify and discuss some of these differences centering on workflow issues and upon the diagnostic roles in the clinical care cycle. Towards the goal of better developing the rationale and implementation model appropriate for pathology, the focus of this communication necessarily centers on the potential role(s) of digital pathology, and its potential impact on both clinicians and investigators.

   Workflow Top

Radiology Workflow

Radiology's modality capture workflow differs significantly from that of Pathology, in that, Radiology's constitutive imaging modalities [Magnetic Resonance (MR), positron emission tomography (PET), computed tomography (CT), ultrasound, etc.] are intrinsically digital in nature and consequently generate images in native digital formats. They are obtained at one instance in time (or within brief time frames being less than an hour) while the patient is typically undergoing a procedure, or during contrast-enhancing imaging protocols, as well as other relatively static short-term image acquisition modes (e.g., digital mammography). The set of acquired diagnostic images are then interpreted by a radiologist with an understanding of the clinical information, and with the ultimate issuance of a report, and storage of all the associated digital data. Additional radiology studies may be requested, which would require the patient to be brought back to the imaging suite to undergo additional imaging protocols. The clinical care cycle then proceeds to the development of monitoring or treatment plans with the other clinical providers.

Pathology Workflow

The following anatomic pathology example underscores a typical workflow scenario, showing evidence of many more differences than similarities to Radiology. Initially, a patient may undergo an outpatient biopsy procedure (often performed in the morning or afternoon) that is utilized to obtain diagnostic tissue. This material, in turn, should ultimately result in a clinically-actionable diagnosis that may lead to the surgical resection of the primary lesion, and possibly, much greater amounts of tissue. In any given procedure, the total volume of tissue retrieved may range from a few microns or millimeters in greatest dimension (e.g., fine needle aspirate or biopsy) to greater than 80 cm in length (e.g., colon resection, etc.). Such retrieved tissue is initially grossly examined by a pathologist or pathologist's assistant (usually by no later than the afternoon of the day of procurement), who is tasked with documenting the available macroscopic findings. Based upon the gross examination alone, an experienced pathologist often can make a relatively accurate diagnosis. Also, at the time of gross examination, as an added measure for generating permanent documentation, photographs can be acquired as needed, via a handheld digital camera. Any resultant images are typically reviewed later, either at the time of microscopic examination or during clinical management conferences, such as tumor boards. Representative sections of tissue are retrieved from the specimen for subsequent microscopic examination, to confirm the primary diagnosis and to assess any additional suspicious or unusual pathological changes, as well as the surgical resection margins for possible involvement. These tissue samples are then placed in a fixative for several hours and subsequently subjected to dehydration, through graded alcohols, immersion in an organic solvent (xylene), and infiltration by hot, liquid paraffin, typically an automated, overnight process. The tissue sections are oriented and permanently embedded in paraffin blocks (FFPE) the following morning. The paraffin tissue blocks are then forwarded to microtomy workstations, where 4 micron thick tissue sections are cut and affixed to glass slides. Finally, the resultant slides are subjected to histochemical and immunohistochemical staining followed by cover-slipping, resulting in the diagnostic form universal to Pathology: the glass microscope slide. These slides serve as a miniscule fractional view of the totality of tissue actually contained in a typical block. Thus, at this point in the workflow, before any slide is reviewed microscopically, sampling error is possible.The slides are delivered to the pathologist for interpretation by the next morning or afternoon (at this point, the slides could be scanned into digital slides).

A pathologist performs an initial review of these slides, and either makes a diagnosis or orders additional slides to be sectioned and stained, utilizing some or all of the remaining tissue left in the case's paraffin blocks. These slides are often cut and stained for specific organic elements, biological markers or molecular epitopes (e.g., protein or DNA) via a plurality of marker chemistries [histochemistry, immunohistochemistry, in situ hybridization, in situ PCR (polymerase chain reaction), etc.]. Alternatively, the pathologist may order a series of deeper hematoxylin and eosin (H & E) sections from the original blocks to better assess the presence or absence of a particular diagnostic entity in the totality of the tissue that has been sampled (a common scenario being where the pathologist is intent on elucidating the presence of malignancy when initial sections are equivocal). (At this point, the slides could be scanned into digital slides). All the slides (which can range from 3 - 30 or more in number, depending on the case at hand) are thoroughly examined under the microscope, and these findings are then correlated with the gross examination, the clinical history, and the radiological findings. Finally, a diagnosis is made. To be clear, the pathologist is legally responsible for everything on all the glass slides 1 just as the radiologist is legally responsible for all the images rendered from the radiographic studies. After the surgical pathology report is issued, the slides and the tissue blocks (and digital slides) are stored.

   A Diagnosis Versus an Interpretation Top


A radiological image is performed at the request of the clinical team as a tool to guide treatment decisions or to bracket the overall anatomical frame of reference. Often, radiology images (such as a CT scan, MRI, or X-ray) are reviewed by both the clinical team and the radiologist. The radiologist then issues a report containing an interpretation of the findings, and a list the types of diseases that may have these particular imaging findings. Most frequently, the radiology report constitutes a list of impressions, rather than a single definitive diagnosis. 2 This is related to the limited spatial resolution of standard radiological modalities, causing many disease processes to appear similar or even indistinguishable. This limitation can be very challenging for the clinical team, who are often seeking a definitive diagnosis in order to initiate definitive treatment. In this context, typical radiology reports might read as follows:

- "This lesion likely represents a benign reactive process owing to its small size, current location, and the presence of adjacent edema, with further confidence in this impression conferred due to the patient's age and the reported history of slow clinical onset. However, slight hyperechosity in the central region of the lesion may be associated with a malignant process. Close follow-up, an MRI study or biopsy is recommended. Clinical correlation is recommended."

-"The upper outer quadrant exhibits a hyperdense, irregular nodularity measuring 2.1 × 1.4 × 1.2 cm in greatest extent. Biopsy is recommended for definitive diagnosis."

Such reports often result in the clinical team's needing to defer to histopathological examination of permanent sections, as a vehicle for arriving upon definitive and clinically-actionable diagnosis.


A pathologist renders a diagnosis based upon the tissue obtained from the patient. Sometimes, diagnoses can be rendered at the time of gross examination (e.g., epidermal inclusion cyst). However, a pathologist is not only responsible for making a diagnosis on the lesion of concern to the clinical team, but also for commenting on any other disease process that may be present, whether observed grossly or microscopically. Therefore, it is not unusual to have one or more diagnoses on a specimen. Often, especially in the case of large specimens, only representative areas are sampled and rendered as slides. Sometimes microscopic lesions go unnoticed and are only brought to the attention of the pathologist after the fact by the clinical team, who were themselves aware of an outside institution's pathology findings or of preexisting molecular imaging results, but failed to communicate this information to the pathologist along with the case itself. Therefore, a pathology image alone is only a partial 'window' into the disease process of the specimen / tissue.

When an organ is resected for cancer, as may be done for prostatic adenocarcinoma or ductal carcinoma of the breast, numerous slides are rendered to answer many questions that will impact patient management, including: Is there cancer? How much cancer is present? What grade is the cancer? Is there only one type of cancer present? How close to the surgical resection margin is the cancer? Is the cancer micro-invasion only present on one slide? Which molecular markers and their respective intensities are expressed by such tumors (i.e., breast cancer). Is the ER, PR or Her2 / neu IHC stain positive?

The associated pathological findings have an immediate impact on therapeutic selection. The most straightforward example is that of treatment of breast cancer patients with trastuzumab (Herceptin® ), which requires pathological documentation of the presence of overexpression of HER2, as detected by immunohistochemical and / or in situ hybridization studies performed on the breast cancer specimen. Other patient management decisions relate to the identification of other disease processes that might be present in the same specimen (e.g., a reactivation of the patient's lymphoma). Therefore, the pathologist must be confident and certain of his / her diagnoses, because immediate action is often taken after a pathology report is issued per a treatment protocol, such as: a BRCA-positive breast cancer patient undergoing contralateral mastectomy for the other breast and receiving additional radiation on the ipsilateral side, along with an extended chemotherapy protocol.

   Pathology Slides as Molecular Repositories Top

When a digital slide is created, it is a permanent virtual record of the physical glass slide, a physical entity that can be lost or broken. The digital slide can be further analyzed and zoomed in for manual inspection by the pathologist, or additional image-enhancements / post-image processing techniques can be applied to the digital data representation in the digital slide. Similarly, the radiology images can be further analyzed and processed to extract more information for the radiologist's use. However, the radiology image itself is the diagnostic entity for the radiologist to act on; there is no additional recourse for simplified access to additional radiology data unless the patient is reimaged in subsequent studies.

However, in pathology, the physical paraffin block also represents a molecular repository for immunohistochemistry (IHC), in situ hybridization, PCR, and a growing plurality of high-throughput interrogative technologies. For example, these molecules can be probed with IHC to identify biomarkers that will indicate which type of chemotherapy or hormonal therapy the patient will receive, as per the HER2 example. Furthermore, when tissue in the block is exhausted, it is possible to un-stain an H & E slide and re-stain it with the antibody probe of choice. An image along does not possess this potential. Conversely, every paraffin block and/or slide derived from that block is a complete molecular repository of the patient and the disease state in that organ, at the time the tissue was harvested, and serves as a resource for the pathologist to ask additional questions and delve deeper into the disease analysis. This, in turn, will impact the patient's final diagnostic assessment. In terms of the pathology workflow, the digital slide constitutes an additional (and potentially redundant) diagnostic entity in the aggregated collection of 'raw material' resources, of which the pathologist has access when formulating the diagnosis. Clearly, some measure of process evolution is still required to fully integrate the use of digital slides into existing workflow in a manner that allows the full appreciation of any remaining 'analog' material. Without this consolidation, pathology is faced with the apparent paradox of actually creating more work for itself, if use of whole slide imaging work flow is simply layered upon the existing microscopy-based sign-out.

   Discussion Top

Whereas Radiology is a specialty that typically renders impressions, Pathology typically offers the definitive discriminant power of a true diagnosis, along with any associated prognostic / theranostic value-added data. Although a radiologist is asked to render impressions on received images alone, a pathologist is tasked with the generation of diagnoses from received tissue (all of it - and not just that which ultimately makes its way to the slide or to the pathology image data set). The diagnostic evaluation actually begins with the pathologist examining the specimen. Representative sections of the specimen are then taken for microscopic evaluation and for ancillary studies, which can aid in deriving a diagnosis. Thus, the fundamental differences between Radiology and Pathology knowledge generation models lie in the observation that the image itself is the diagnostic driver in the former, whereas, it is the specimen that serves as the driver in the latter.

Although an image can be made of a specimen, both grossly and microscopically, and a diagnosis can be made from such resulting pathology images, it is actually the specimen, collectively, within the one or more tissue blocks associated with the case, that is of greatest value in reaching a correct diagnosis. And while a pathology image is a fractional representation of the specimen block, it should be emphasized that the pathologist has the ability to derive additional data from the specimen block in a continuously evolving and expanding manner, throughout the diagnostic workup (as illustrated by the work flow as already described earlier), whereas, radiology datasets generally represent static constructs. In many instances, a pathology image can supercede the capacity of a radiology image to provide the data required to render an immediate diagnosis. Consequently, pathological interpretation remains the gold standard for definitive diagnosis, which, in turn, determines both prognosis and associated treatment. Thus, it is unambiguously the specimen itself that contains the greatest amount of disease- and patient-specific information (DNA / RNA / protein data), and the pathology image is merely a window into a subset of the totality of such information. It is the ability to carry out a retrospective query upon archival tissues that sets apart the pathology diagnostic modality domain from that of the radiologic modalities, which have no archival equivalent. This can be best illustrated by the example of pulmonary cytopathology, which exceeds the diagnostic power of radiological imaging (used to locate, quantify, and provisionally identify lesions) by also providing definitive identification and more importantly, malignant potential. Even a small quantity of cytologically identified diagnostic cells will allow the physician to commit to performing a definitive therapeutic management step, such as an excisional biopsy, complete resection, neoadjuvant therapy or non-surgical therapy, as appropriate.

In the cytopathology workflow, fine needle aspirate (FNA)-acquired tissue is placed onto a slide. If one applied the model intrinsic to contemporary radiology workflow at this stage, an image would be made and it would be simply interpreted, with the process stopping here. However, with the added repertoire of pathology workflow, cells identified on first review of FNA material, which might represent either a malignant or a reactive processes, can be further adjudicated by special stains (e.g., CK7, CK20, TTF1, etc.) that provide exquisite sensitivity in defining both lineage and biological potential; attributes not so immediately available in the radiology-only setting. If a critical fraction of FNA material is positive for putative markers of cancer or any other illness, it is a relatively easy process to definitively render one or more clinically-actionable diagnoses.

In the setting where initial microscopic sections fail to provide diagnostically compelling evidence, subsequent deeper tissue sections from the paraffin block can be obtained. These may provide additional opportunities to generate a definitive diagnosis. This example again demonstrates how it is the tissue itself, and not the derived images that is the actual diagnostic driver.

   Pathology Images in Research Applications Top

With regard to basic and clinical research applications, pathology images, gross images, and WSI datasets also differ from radiology images, in that they represent a window back to the original specimen in which rests the referential the DNA, RNA, and protein material. Thus, digital pathology imaging serves as a bridge between the original tissue physical state and its molecular features. Similarly, this allows for bridging of the specimen's physical state with that of the clinical phenotype.

In summary, pathology imaging is fundamentally different from that of radiology. Additionally, there are basic differences in pathology workflow that have implications for clinical management and therapeutic decision-making. As already emphasized, the tissue itself is the diagnostic driver in pathology, with its images merely serving as portals into a vast continuum of molecular and morphological data that represents the totality of information intrinsic to the biome.

There are many lessons and challenges that radiology has faced and overcome in moving to digital imaging that can and should be similarly leveraged as pathology moves to more fully adopting digital pathology uses. These include lessons learned in areas such as data management, digital image annotation, and application to telemedicine, education, and research. However, pathology images also have unique challenges associated with larger number of images associated with each case, larger individual file sizes, and possibly, the greater number of geometric annotations and observations that are generated by pathologists. As clinical practitioners and investigators in pathology consider the use of digital imaging and the advantages and limitations of the various digital solutions offered, these key differences from radiological images should be kept in mind.

   Footnotes Top

  1. Radiology is beginning to see bits of this quandary of having to review a significant amount (more than 10) of images per case study. Consider the spiral CT scanner that generates 1300 slices.The radiologist is responsible for reviewing every image.
  2. In general, for radiology images of chronic disease, impressions rather than diagnoses are rendered while for acute diseases such as a ruptured aorta or pneumothroax, diagnosesare rendered that result in immediate therapeutic intervention.

This article has been cited by
1 Breast Cancer Classification Using Deep Learning Approaches and Histopathology Image: A Comparison Study
Faezehsadat Shahidi, Salwani Mohd Daud, Hafiza Abas, Noor Azurati Ahmad, Nurazean Maarop
IEEE Access. 2020; 8: 187531
[Pubmed] | [DOI]
2 Review of the current state of digital image analysis in breast pathology
Martin C. Chang, Miralem Mrkonjic
The Breast Journal. 2020; 26(6): 1208
[Pubmed] | [DOI]
3 Breast Cancer Histopathology Image Analysis: A Review
Mitko Veta,Josien P. W. Pluim,Paul J. van Diest,Max A. Viergever
IEEE Transactions on Biomedical Engineering. 2014; 61(5): 1400
[Pubmed] | [DOI]
4 Simultaneous Sparsity Model for Histopathological Image Representation and Classification
Umamahesh Srinivas,Hojjat Seyed Mousavi,Vishal Monga,Arthur Hattel,Bhushan Jayarao
IEEE Transactions on Medical Imaging. 2014; 33(5): 1163
[Pubmed] | [DOI]
5 Surgical pathology reporting at the crossroads
David W. Ellis
Pathology. 2011; 43(5): 404
[Pubmed] | [DOI]




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

  In this article
    A Pathology Imag...
    A Diagnosis Vers...
    Pathology Slides...
    Pathology Images...

 Article Access Statistics
    PDF Downloaded1405    
    Comments [Add]    
    Cited by others 5    

Recommend this journal