Understanding Artificial Intelligence from a Cytopathologist’s perspective
“Artificial Intelligence (AI) in Pathology can be quite an intimidating subject”, says Dr. Joerg Schwock, “There are a lot of terms we pathologists don’t routinely deal with.” Trying to understand scientific papers on the subject is not always easy and reading them in detail can be an exhausting exercise without the proper education.
An Assistant Professor in the Department of Laboratory Medicine and Pathobiology in the Temerty Faculty of Medicine, he is a Cytopathologist at Toronto General Hospital in the University Health Network (UHN). He is also an education section editor of Acta Cytologica – the official journal of the International Academy of Cytology.
He was approached by the journal to lead the development of free, online educational content to accompany issues of the journal, including a recent edition on Digital and Computational Cytology. At first having limited knowledge of AI himself, it was an opportunity for personal development, and he now admits AI “is not as frightening anymore”.
Cytopathology is a discipline within Anatomical Pathology which attempts to diagnosis disease by evaluating individual cells from the lesion. Due to a history of automated screening of Pap tests, those practicing Cytopathology are poised to embrace automation in the laboratory. However, the introduction of AI offers an entirely new level of possibilities, once current technological barriers have been overcome.
“AI is definitely something I see on the horizon”, says Dr. Schwock, “We have to interact with the change that's coming and be prepared for it rather than pushing the thought of it away and fearing for our jobs. This is our opportunity to shape AI to our advantage because we are the content experts”.
The first of its kind for Acta Cytologica, the journal decided to pair in-depth journal coverage with complimentary educational modules so readers can delve into, and understand, topics with which they may not be so familiar. When Dr. Schwock was approached by the Editor-in-Chief about the idea, the special edition on AI was already in preparation and it “all converged rather nicely”.
AI experts Dr. Ewen McAlpine and Dr. Pamela Michelow from the University of the Witwatersrand, South Africa, the journal publication manager based in Switzerland, and Dr. Schwock came together through a series of virtual meetings to develop the e-learning module.
“This international collaboration was very exciting,” says Schwock, “It was an opportunity to bring together like-minded people to make a somewhat difficult topic much more digestible to average readers”. Approaching the topic from the perspective of a practicing Cytopathologist, and not a computer scientist, was an aspect Dr. Schwock felt was important and he provided that perspective.
The course explains key concepts, something Dr. Schwock hopes will make AI more accessible. “I was initially put off by the vocabulary involved so with the help of our expert team we stripped it to its bare bones for someone who is not familiar with the subject”.
Aside from posing and answering questions, the module walks participants through the creation of an AI model for cytopathology to make the process more understandable. Participants also get an idea of what is involved in image annotation, whilst limitations and challenges of AI in cytopathology are acknowledged and documented to stimulate future research.
Although created for a Cytopathology journal, the course is free and relevant for anyone involved with pathology.
“There is great potential for digital and computational pathology and much progress has been made in the field,” adds Schwock, “Formidable barriers still to be overcome but I am confident solutions will be found soon. We, as Pathologists will not be replaced by computers, but substantially supported by applications that could, for example, pick out abnormal cells and pass the data to a human for final adjudication.” It will enhance our diagnostics and research activities.