How does technological innovation in the medical field give rise to new and expert ‘ways of seeing’? Researchers Jojanneke Drogt and Megan Milota, working in collaboration with pathologist Shoko Vos, trace the journey of a dermatological mole ‘in’ and ‘out’ of visibility to reveal different modes of ‘perceiving’ in relation to medical images.
This paper was presented at the Northern Network for Medical Humanities Research Congress 2021.
Jojanneke Drogt and Megan Milota write:
In order to diagnose an illness, physicians need to gather and analyze relevant information. This information can come from a patient’s description of their symptoms, from blood or urine tests, or from medical images like scans, x-rays, and imaged tissue samples. As members of the interdisciplinary research group RAIDIO (Responsible Artificial Intelligence in clinical DecisIOn making), we are studying two fields of image-based medicine: radiology and pathogy. We are particularly interested in how digital technologies—like scanners, digital analysis interfaces and artificial intelligence (AI)— shape physicians’ ‘ways of seeing’. How do medical visibility and invisibility relate to one another?And more specifically, what are the differences between directly perceiving a part of your body versus looking at a microscopic medical image of it?
The complexity of image analysis
Answering these questions requires a grasp of the complex relations between the medical, social, historical and philosophical dimensions of medical practice. Although the scope of this article is necessarily limited, we will highlight some important steps and concepts in medical image analysis. In many ways, this article is the culmination of a collaborative journey. We, Jojanneke and Megan, are not medical professionals, and when we started our interviews with pathologists, we knew very little about what their daily workflow looked like. As outsiders, we tried to piece together and make sense of the various steps in the diagnostic process, which we’ve distilled below. Our co-author Shoko, a pathologist and member of our research team, willingly answered our questions and explained the basic contours of her work by means of the digitalized tissue samples included below. We’ve also drawn upon our own observations and experiences as dermatology patients when considering the journey of our tissue samples—in this case moles—from our bodies to a pathologist’s digital workspace.
The complexity of medical imaging can best be understood by looking at a specialized image, such as Figure 1. Without any previous experience of looking at an image like this, it is hard to identify what it represents. It could be mistaken for a work of art, an aesthetically pleasing pattern of pretty pink and purple blending in one another. From afar, it might be a picture of another planet, a foreign world perceived from above, riddled with craters. Even if you know this is a picture of cells, an image of some sort of tissue, it remains to be seen what kind of tissue it is (let alone what disease it might harbour). By taking a new look at such an image, it becomes apparent how foreign it actually is: it is utterly unlike the physical body we perceive with our eyes, feel with our hands, and imagine with our minds.
More familiar to the ‘non-medical eye’ are the normally visible parts of our bodies. This is shown when we compare the microscopic image above with the original source of the picture: a mole. Moles are found on most skin types and are usually no more than harmless pigmentations or cell groupings. Normal parts of our bodies become clinically relevant, or abnormal, when they develop signs that might indicate the presence of disease. Harmless moles can potentially become melanomas. By forming the medical hypothesis that a mole might be developing into cancer, a normal and healthy part of a body becomes something full of medical meaning. Moreover, invisible ideas such as causality arguments, disease histories and narratives concerning moles and melanoma may provide initial clues as to a particular mole’s (ab)normal status (Grmek, 2018). Once a mole has been marked as ‘abnormal’, it becomes the the object of medical scrutiny.
Creating medical visibility
Medical imaging is used when a healthcare provider forms the hypothesis that a disease might exist in the body. In order to confirm or deny malignancy, a mole must first be removed from the body. In this way, the mole begins it’s own journey toward clinical visibility as a microscopic image. It is important to realize how labor-intensive the processes of creating ‘new’ forms of medical visibility can be. In the case of a mole, it is sent to a pathology lab, where it is unpacked by lab technicians. They must describe the mole, determine the resection margins (is the mole completely excised?) slice the tissue, dye and mount on slides.
In digital pathology departments, these slides are loaded into a scanner and the resulting images are sent to an analysis interface on the assigned pathologist’s computer. Only once this process is completed can the diagnosing pathologist begin their work of anwering the question posed by the treating physician who sent in the tissue sample in the first place. In this case: is this mole a form of cancer or developing into cancer? To answer this question, they annotate the tissue sample identifying and marking portions of the image that look like potential malignancy. By means of these markers, the likelihood of cancer is established. This level of certainty is variable and depends on the complexity of the tissue itself.
Translating medical images to the patient
Using the microscopic image, a medical expert confirms or denies the possibility of cancer within the mole. But what kind of meaning is actually being attributed to the picture? As laypersons, we need the pathologist’s translation before we can ‘see’ what this picture actually depicts. Normally the pathologist gives this translation to the treating physician, who again translates it to the patient. For example, the pathologist’s report might say: ‘’compound navus without atypical features. Some superficial dermal activity, yet not enough to confirm malignancy. Resection margins are not reached.’’ And the treating physician may then say to the patient: ‘‘there is nothing wrong’’. Each time a translation occurs in this process, the pathologist’s diagnosis could lose part of its medical meaning. By the time the translation reaches the patient it may been pared down and is usually stripped of its medical terminology. The assurance that the mole is benign may assuage the fear of cancer, yet the uncertainties surrounding the disease itself might remain.
A cycle of visibility and invisibility
Through this medical process, a mole has made a whole ‘journey’ or ‘cycle’ of visibility and invisibility, which we’ve tried to illustrate here. It begins with cells which divide and become visible to the human eye as a mole, which further develops until it becomes a cause of concern and is removed. At this point the mole becomes invisible again, an incision scar the only trace of it’s existence. What once was an integrated part of a body kept alive by blood vessels is now preserved in wax and sampled on a slide. As a medical image the mole becomes visible again, albeit only for a few highly trained experts who can interpret the specific details of the image. As one pathologist remarked to us in an interview: ‘‘Even many of our clinical colleagues have no idea what we do here, so that’s (…) even harder for the patient to get something meaningful out of this.” This is how we’ve ascribed meaning to this process: what was once a visible mark on a human body becomes, through a process of selection, reduction, and renaming, a colored smudge on a slide or on a screen. Something illegible to an untrained eye but full of significance for a pathologist. On a slide, a disease cannot hide. The white noise of confounding factors, mundane distractions, and the fears and worries of the patient are turned down to a murmer. In this silent encounter a pathologist hunts out a disease at cellular level.
Visible to whom?
Many medical images, such as the image below, are therefore only accessible to the ‘trained eye’. In order to ‘see’ what is happening on the slide, this trained eye must integrate a range of relevant medical knowledge with the image itself. As another clinical pathologist remarked in our interviews: ’’What we do is gather things […] from a lot of information sources. We consider clinical information along with the dermoscopic image. And we add to that the integrated image and perhaps the molecular data as well. And we integrate all of these things in order to ultimately arrive at a diagnosis.’’ Taking this into account, deciding what it is we are looking at is actually a decision making process, an integration of various parts of the medical world, knowledge, experience and intuition.
Invisible frames to the medically visible
Tracing the life cycle of a mole requires a shift in focus from the external reality of disease to the challenging practice of diagnosing internal disease mechanisms. As we have tried to show, the specialized practice of looking at medical images belongs to those who have developed a ‘trained eye’. But other fields can contribute to the ways we look at and understand medical imaging as well. For instance, in philosophical discourse concepts like explainability and responsibility are frequently mentioned in regards to medical imaging (e.g. Lalumera and Fanti, 2021). Scholars in the history of medicine focus, in turn, on the importance of pathology’s place in the medical system (e.g. Van den Tweel and Taylor, 2010). And from an anthropological perspective, attention can be paid to the ways in which image-based medicine develops its own unique conventions, practices and communication strategies around a technology such as the microscope or the digitalized tissue sample (e.g. Van Dijck, 2011).
The future of image analysis
The field of image-based medicine continutes to develop; more and more departments are currently making the switch from analog to digital workspaces. In these digital spaces, artificial intelligence (AI) applications are now in the process of being developed and validated (e.g. Aeffner et al., 2019). With each new development, new interactions take place between visibility and invisibility. In the cycle we described, humans interacted with technologies at each step in the process. In the foreseeable future, it’s likely that AI applications will be added to the list of diagnostic tools at a pathologist’s disposal. The possible collaborations between medical professionals and algorithms raises a whole new set of questions. For instance, to what extent will AI change a pathologist’s ways of seeing and knowing? Will this new technology find connections that are invisible to the human eye? At the moment there are no answers, just many paths forward, not all of them clear or even yet visible.
Jojanneke Drogt is a PhD Candidate in the philosophy of technology at the University Medical Centre Utrecht (UMCU) and Utrecht University. She is particularly interested in applying theoretical and ethical frameworks to specific (medical-)technological contexts.
Dr Shoko Vos is a pathologist and postdoc researcher at the University Medical Centre Nijmegen (Radboudumc). She works with digital pathology on a daily basis and likes to critically reflect on new developments in her field. She has published several biomedical as well as ethical articles.
Dr Megan Milota is an assistant professor in narrative medicine at the University Medical Centre Utrecht (UMCU) and Utrecht University. She has published on a wide range of topics related to medical education, narrative ethics, and qualitative research methods.
All are part of the RAIDIO project (Responsible Artificial Intelligence in clinical DecisIOn making), which is funded by a Dutch research grant (406.DI.19.089). Members of the team are: Prof dr Annelien Bredenoord, Prof dr Sally Wyatt, Dr Karin Jongsma, Dr Megan Milota, Dr Shoko Vos, Dr Flora Lysen en Jojanneke Drogt.
Aeffner, Famke, et al. “Introduction to digital image analysis in whole-slide imaging: a white paper from the digital pathology association.” Journal of pathology informatics 10 (2019).
Grmek, Mirko. Pathological realities: essays on disease, experiments, and history. Fordham University Press, 2018.
Lalumera, Elisabetta, and Stefano Fanti. Philosophy of Advanced Medical Imaging. Springer Nature, 2021.
Lindley, Sarah W., Elizabeth M. Gillies, and Lewis A. Hassell. “Communicating diagnostic uncertainty in surgical pathology reports: disparities between sender and receiver.” Pathology-Research and Practice 210.10 (2014): 628-633.
Van den Tweel, Jan G., and Clive R. Taylor. “A brief history of pathology.” Virchows Archiv 457.1 (2010): 3-10.
Van Dijck, José. The transparent body: A cultural analysis of medical imaging. University of Washington Press, 2011.