Undergrad: Hamilton College, Chemistry.

Graduate: Boston University School of Medicine, MD PhD program

Residency: Massachusetts General Hospital, Anatomic Pathology

Fellowship: Massachusetts General Hospital, Cytopathology

Employment:

• Assistant Professor of Pathology and Lab Medicine
• Medical Director of Pathology Informatics
• NCCC Cancer Faculty Fellow
• NCCC Translational Engineering in Cancer Member

Research Philosophy: As scientists we have an obligation to perform high quality and moreover, useful research. Resources are limited so it is up to each of us to perform at our peak potential. As such, we should avoid obvious projects. If something is obvious, someone else will do it. We should instead pick research topics that stretch our limits and make us uncomfortable. When I started my research career I was struck by the sheer amount of mindless, repetitive data analysis tasks we were required to carry out. I knew there had to be a better use of our (my) time and set about to teach myself to program in VBA (laughter is appropriate). The learning curve was incredibly steep and I almost gave up on multiple occasions, but I can confidently say that stretching my limits to learn even rudimentary programming was the most impactful decision of my life. This experience taught me that hard problems are useful problems. Most people avoid things that make them uncomfortable, if we do the opposite, we become experts in a rarefied niche filled with unexplored avenues. I ask anyone I work with to justify the projects they choose and ask themselves will this make a difference? Will I learn something new? Do I feel like an imposter even talking about this topic? Does it give me goosebumps thinking about the possible outcomes? If the answer to the first 3 questions is yes, you have yourself a good research question, if the answer to all four questions is yes, you have a great research question, the kind of question that will get other people excited, the kind of question that will get you platform presentations at conferences and publications in high impact journals. Even if it takes you a week to figure out what that question is, it will be worth it.

Research Interests: My principal research interest is the development of quantitative tools for use in the practice of digital pathology. Pathology is a largely Victorian practice that has yet to embrace digital practice to the same extent as, for instance, radiology. The reasons for this, are of course, more complicated than they might seem. The average size of a layer in a radiology image (e.g. a CT scan) is 1024 x 1024 pixels. Contrast this with a whole slide image (WSI) scanned at 400X (the practical maximum magnification through air) which comes in at 80,000 x 80,000 pixels per color channel. It was not until recently that scanners were developed which could create such enormous images and most computers still have insufficient RAM to open them. The other reason for the delay in adoption in digital pathology is the fact that glass slides and microscopes are 125 year old technology. They are very fast and very efficient to screen. The pathologist typically pushes the slide around with their hand and adjusts focus and magnification with the other. For simple diagnoses, pathologists may dwell on a slide for as little as 10 seconds. Digital image viewers, by contrast require the operator to drag the image around with a mouse. The process is slow compared to glass and quickly results in wrist strain for the operator. The current digital interface method is thus slower and more laborious than glass slide screening and does nothing to take advantage of the digital format. It’s like putting radio on the internet, a gross underutilization of the technology. Therefore, we focus on killer applications for digital pathology, algorithms that automate tedious tasks, detect rare events and minimize the risks of future litigation due to a momentary lapse of judgement by a fatigued practitioner. These algorithms serve as our “foot in the door” so we can pursue truly revolutionary techniques. Our overarching goal is not to automate tasks humans can perform but to perform analyses that no human ever could do to insufficient attention or features that are too subtle or too high dimensional. It is also extremely important that medical practitioners be involved in the development of these algorithms. There is a huge interest in industry in developing health care applications. This research and product development is often performed solely by engineers who have no real conception of what the real problems in medicine are and who have never interacted with patients, nurses and hospital technical staff. These are also the same companies that sell our personal information to nefarious corporate entities, disinformation mills and hostile foreign governments and seem incapable of acting in an ethical manner. Jaron Lanier, one of my personal heroes, likens these companies to psychopaths who we willingly cede our personal information to in return for convenience. It is therefore imperative that we perform high quality and ethical research and educate our colleagues as to the danger these companies pose.

More about me: I live my life by a single philosophy; that humans develop technology so they don’t have to do any work. Our end goal as a species is to do nothing we don’t want to. To this end, I make sure to balance my actual life with my work life. I’m attracted to computational research because I can write a script to do 1000 man hours of work, while I go to Whaleback and ski all afternoon. I love the outdoors. I love to ski, mountain bike, hike, sail and garden. I love video games, RTS, FPS, Platformers, Space Sims (Freelancer), etc. I love pickup trucks, although I don’t have one. I like country music, 90’s pop and instrumental progressive / djent. I also like to build horrible furniture in my woodshop. I have a wonderful family composed of my wife Molly, my son Walter and my daughter Nancy. I hate woodchucks, black flies, ticks and yellow jackets, I’m ok with most other lifeforms.

Even more about me: Louis Vaickus and Joshua Levy co-head the EDIT Machine Learning Laboratory.