It seems like artificial intelligence (AI) is everywhere. Have you tried the Google Arts and Culture app? Take a selfie, and the app will use image matching AI to match your face to famous works of art displayed in museums around the world. My face was matched to this portrait of a lovely young woman from Argentina's national museum of art. This app may be artificial, but it must be pretty darn intelligent if it knew enough to find a pretty face to flatter and please this vain and aging user!

I think that AI is a big part of the future of health care — this was one of the messages in my Papanicolaou Address at the American Society of Cytopathology's annual meeting last November. I was involved in the early days of neural networks (a form of AI) for cytology screening in the late 1990s, as was Vice Chair of Clinical Services John Bishop (1,2). But 20 years ago, AI in cytology was an idea ahead of its time. Neural networks are now more advanced and there is new interest and promise. A study by Hou et al. demonstrated that the new neural networks could classify lung carcinoma into histologic subtypes as well or better than expert pulmonary pathologists (3). Amazingly, AI systems can distinguish short-term vs. long-term survivors in patients with Stage 1 lung cancer just by evaluating hematoxylin and eosin stained slides (4). AI is also being used to improve readings of retinal images in ophthalmology (5). But AI is not just about images. Baylor University is employing AI to gather and analyze patient data from the electronic medical record – including lab data — to improve length of stay, minimize readmissions, and identify patients at risk for sepsis (6). Stephen Phinney, a UC Davis professor emeritus in Internal Medicine, co-founded an AI-based company called Virta that uses patient collected data to support nutrition and dietary management in patients with diabetes (7). Notably, a venture capital group funded Virta with a $37 million investment (8). AI is clearly on its way to becoming a big business in medicine.

For all these reasons, the Council of the Association of Pathology Chairs (APC), identified AI as the next “big thing” that academic departments of pathology and laboratory medicine need to incorporate into their curricula, research and clinical programs so that we can shape our own future. A special session on AI is therefore planned for the APC's annual meeting in July.

So what can we do to prepare for our future with AI right here at home? An editorial last year by Sharma and Carter in Archives of Pathology and Laboratory Medicine encourages pathologists to make friends, not enemies, with AI (9). An upcoming editorial in that same journal by our very own chair emeritus, Distinguished Professor Ralph Green, describes the potential disruption that AI can bring, but emphasizes the growing need to include AI in our residency curricula since AI provides opportunity for new efficiencies, for addressing the predicted workforce shortages, and for increasing professional satisfaction by removing drudgery.

Their call to action has already been heard — members of our department and others are moving us forward into in the AI world. Pathology faculty member John Paul Graff has co-founded an interest group in AI for faculty, housestaff and others, along with Radiology's Tom Loehfelm and members of the faculty from Computer Science and other departments. We are starting to invest in AI, too – we funded two grants from our department's Collaborative for Diagnostic Innovation seed grant program: Jason Adams' (Internal Medicine) and Michael Johnson's (Emergency Medicine) project “Validation of a Machine Learning-Based Automated ARDS Diagnostic Test” and a project from two faculty in the School of Veterinary Medicine Hao Cheng (Animal Sciences) and Erik Wisner (Surgical and Radiological Sciences) on “Application of Artificial Intelligence in Detecting Canine Left Atrial Enlargement on Thoracic Radiographs.” I'm also really excited that a UC Davis undergraduate contacted me just this week about an AI project utilizing image recognition neural networks that he developed for automatic classification of cytologic smears. He wants to work with our department to take this beyond the proof of concept stage.

But what about the really big question on everyone's mind – should we be worried that AI will replace us?? I don't think so – I think that our roles as pathologists, cytotechnologists, and laboratorians will evolve in new directions as we adopt new tools, just as we have always done. In a recent editorial in the Journal of the American Medical Association, Jha and Topol foresees the pathologist of the future as an information specialist who “…would not spend time…examining slides for “Orphan-Annie” nuclei…” and instead “…interprets important data, advises on the value of another diagnostic test and integrates information to guide clinicians (10).”

Serving as information specialists will be an important and foundational component of our future roles. But as I pointed out in one of my previous blogs (11), the real job behind our job as is to reduce anxiety among our patients and our provider-clients. I believe that this will require more than timely, accurate, or integrated information. As health care practitioners, we will need to provide the important human touch that machines can never have – we will need to create and maintain relationships with clinicians and patients to ensure a positive experience for those accessing laboratory services and to ensure the best patient outcome possible. AI can also never be creative or strategic in the ways that human are (12). New innovations and applications will inevitably come from the human users.

I'm excited that we are moving into the AI world, and optimistic that we will become a leader in this area, thanks to our faculty and staff's interest, talent, and most importantly their wonderful human qualities — curiosity, creativity, compassion and caring — that will take this incredible technology to the highest level. I look forward to sharing more as our activities with AI unfold.


  1. Howell LP, Belk T, Agdigos R, Davis R, Lowe J. The AutoCyte Interactive Screening System: experiences at a university hospital cytology laboratory. Acta Cytol 1999; 43:58-64.
  2. Bishop J. Kaufamn, Taylor. Multicenter comparison of manual and automated screening of AutoCyte gynecologic preparations. Acta Cytol 1999; 43: 34-38.
  3. Hou L, Samaras D, Kurc TM, Gao Y, Davis JE, Saltz JH. Patch-based convolutional neural network for whole slide tissue image classification. Proc IEEE Comput Soc Conf Vis Patt Recog 2016; 2016:2424-2433.
  4. Yu KH, Zhang C, Berry JG, Altman RB, Re C, Rubin DL, Snyder M. Predicting non-small cell lung cancer prognosis by fully automated microscopic pathology image features. Nat Commun 2016; 7: 12474.
  5. IBM research is training Watson to identify eye retina abnormalities. IBM Newsroom, 2/21/2017: (Accessed 1/30/2018).
  6. How AI can help hospitals manage length of stay. HIMSS Media, 12/20/2017. (Accessed 1/30/2018)
  7. Maney K. How artificial intelligence will cure America's sick health care system. Newsweek, 5/24/2017. (Accessed 1/30/2018)
  8. Buhr S. With $37 million in funding, health start up virta aims to cure type 2 diabetes by watching what you eat. Tech Crunch, 3/8/2017.
  9. Sharma G, Carter A. Artificial intelligence and the pathologist: future frenemies? Arch Pathol Lab Med 2017; 141:622-623.
  10. Jha S, Topol EJ. Adapting to artificial intelligence: radiologists and pathologists as information specialists. JAMA 2016; 316:2353-2354.
  11. Howell LP. Do you know the real job behind your job? Pathology chair blog. August 31, 2017. Accessed 1/30/2018.
  12. Pistrui J. The future of human work is imagination, creativity and strategy. Harv Bus Review, January 18, 2018. Accessed 1/30/2018.
  13. Howell LP. Do you know the real “job to be done” in your job? Pathology Chair's Blog, UC Davis Health.