Author: ChadVanderbilt

Dear Reader,

The purpose of this blog is to create a useful corner of the internet where people who have interest in applications of machine learning in Anatomic and Clinical Pathology can go to learn more about the latest developments in this quickly developing field.

I am currently a 4th-year Pathologist-in-Training (Resident) going into molecular pathology pathologist in molecular pathology fellowship. For more about me, here is my CV.  My interest in machine learning and artificial intelligence has developed as a side effect of many observations from clinical medicine as well as some data related research projects I have completed over the last several years.

In additions to my advanced studies in clinical medicine and pathology, I am heavily influenced by the works of Nassim Teleb, Daniel Kahneman, Pedro Dominguez, Phil Tetlock, Brian Christian, Ray Kurzweil, and many others.  Other important resources that I draw from for my ideas are the Talking Machines Podcast and lectures of the Long Now Foundation.

I find that these influences from outside my area of professional experience are at least as important as my professional training.  As NN Taleb discusses at length in his books and Daniel Kahneman has shown with his research, many people are bound to their domain of expertise and fail to apply knowledge from one domain to another.  I have found that by gaining the knowledge of people who study other disciplines, one can apply these ideas to her/his own field and thrive.  This is what I am hoping to achieve with this blog/website.

Machine learning is by and large a field within the domain of computer science, most often applied to applications within the technology industry. The most famous examples of Machine Learning are product recommendations for websites, computer voice and text recognition, Chess playing, Jeopardy winning, and self-driving cars to name a few.

The power of these algorithms is just beginning to trickle into the realm of clinical medicine.  At the moment, we use some forms of machine learning for creating surgery schedules and certain applications of calling mutational variants in DNA sequencing.  However, I have observed that more and more research articles are being published in research journals on how to apply these methods in ways that the algorithms can be used to directly manage patient care.

The reason that I have decided to write from the perspective of pathology is not only because that is where I have the most advanced training but also because pathology is where I see the most practical applications of machine learning.  In order for machine learning to be useful, large data sets must be available.  As it happens, the largest data sets are in the scope of influence of anatomic and clinical pathologists: (1) clinical laboratory data, (2) molecular and genomic data from tumors and constitutional, and (3) the data contained on histology slides.

The major sources of data in clinical medicine outside of pathology are electronic medical record data (i.e. clinical notes and vital signs) and radiology.  I will likely incorporate occasional research as related to these data sets when the work is compelling but will try to limit this website mainly to applications within pathology.

Comments related to posts are encouraged as they lead to productive conversations.  If you know of research work that fits into the scope of this blog, please feel free to email me: last name dot first name at Gmail

-Chad Vanderbilt, MD