Saturday, August 4, 2012

Background creates context

A crop germinates in a context where the conditions allow. The ideas herein have been ruminating in my mind for some time. With this blog, the intention is sketch out my own background to enable the reader to understand where I am coming from. I hope to use this writing to further refine the ideas which have stayed locked in the silo of my own mind for too long.

The purpose of this blog is to lay out a vision. In the book Where Great Ideas Come From, Steven Johnson describes something he calls a slow hunch. This is the process by which ideas take a long time to go from acorn to oak. I have always had an interest in writing, it is a way in which I find that I am able to refine a thought. This public diary may prove more helpful on a personal level, forcing me to clearly articulate a vision, than it will be for the reader. I am putting it in the public space because I believe these personal interests may have broader appeal.

My Background
I was born in Mississippi, son of a single mother. For the first 11 years of my life I lived with my mom in the southwest corner of Arkansas, USA. When I was 11, my mother took a job as teacher with the Department of Defense Dependents Schools. This is the organization that hires teachers for schools on U.S. military bases overseas. These schools where children of U.S. military members attend school. She left a job teaching kindergarten in the United States and we moved to Japan first. After 2 years we moved to Seoul, South Korea. I spent 8th through 10th grade living in Seoul. We moved again, and for junior and senior year of high school we lived in Mons, Belgium. After high school, I returned to Arkansas to attend university. I went to a small liberal arts college in central Arkansas which was geographically close to my father. I completed medical school in Washington, D.C. Today, I am in my 3rd year of postgraduate clinical medical training (termed residency in the U.S., house officer in the British system) in family medicine in North Carolina.

Concept: The value of the crowd.
Project 1, a foundation in a mental pyramid
I have known Zach since high school. He and I have grown in friendship since that time. I had breakfast with Zach at a Panera several years ago and described an idea for using cell phones to identify dermatologic skin abnormalities. Smart phones have reached sophistication with a combination of camera, processing power, and internet connectivity that enable this to move out of the realm of science fiction. The idea is for a cell phone to take a picture of a skin lesion and use artificial intelligence algorithms which would correctly identify the diagnosis. To imagine how this would work, picture an episode of CSI (Crime Scene Investigators). During the crime solving music montage there is screen shot of a computer with fingerprints cycling rapidly through the screen until the correct match is made. The criminal is caught! I have also seen this on the silver screen with scenes of facial recognition technology with the computer successfully picking out the right face. This is not science fiction, airports are currently using the technology. This article about facial recognition for airport security is almost a decade old. The cognitive leap to replacing faces with skin pictures is a short distance. A dermatologist in training spends time gaining exposure to thousands of mental images of abnormal skin (along with associated data points of gender, age, other symptoms, etc) and cataloguing that information attached to a “correct” diagnosis. An algorithm will search the submitted image for characteristic qualities (like the ID points on fingerprints) and produce a diagnosis, or list of probable diagnosis - in the same way a human physician operates. In fact, a version of this concept is operational as a medical device. A machine, Melafind, can detect a cancer in a specific type of small pigmented skin lesion better than a trained Dermatologist. The idea was to expand this to any type of skin lesion, not just something that was a potential melanoma.

Zach became interested in this idea. Our Panera coffee shop breakfasts continued sporadically. Zach emailed me and said he was having trouble locating a skin image database that would be  sufficient in size to attempt such a project. The problem is that there exists no publically available skin image database. In order to “teach” an algorithm what the important characteristics of skin lesions are, you need a large set of photographs that are tagged with an answer, a diagnosis. Those types of image collections are all private. They exist either as proprietary collections, or they are in the possession of the academy for research or teaching purposes. Without the database, it is not possible to begin the development of this algorithm.

This was a daunting set back. The energy had stalled forward progress. The next link in the development of the concept was exposure to the ideas of Clay Shirky. The ideas, freely available on the internet, have shaped my appreciation for the potential of collaborative solutions. One of the things I hear in his work is this idea that the answers to problems which emerge from the crowd are more robust and have more potential. Organizing a truly crowd-sourced answer in a pre-internet world was virtually impossible. The internet, which is estimated to have penetration to a third of the global population, enables the the concept. One  widely used example of this methodology is Wikipedia. The challenge which Wikipeda addresses via the crowd is how to organize/update/and disseminate current and historical information on broad range of current and historical topics. The efficacy of the new method is punctuated by the passing of an old method which was previously used to address this challenge. Encyclopedia Britannica announced that it will stop print publication this year. Similarly, Twitter and social media are challenging the professional journalistic class in answering the question of how society keeps itself informed. For exposure to these ideas, I owe Professor Shirky an enormous debt of gratitude.

How Society Keeps itself healthy - A Professional class of Doctors - Can the crowd compete?

Zach and I met at Panera yet again. The barrier to forward progress was a current image database system characterized by proprietary images, locked away, unavailable for public use. This was an impediment that was scuttling a grand scheme to bring the power of technology to bear for improved health. The next conversation was full of questions. Since we don’t have access to a skin database how can we get the volume of necessary images to proceed? Could we get people to actually submit pictures of their own skin? What if we tried to crowd-source a database of skin images with diagnoses? On a technical level -how could we guarantee the quality of images submitted? Is this even legal? How can we protect the identity and personal health information of willing participants? Will this work?

We don’t yet know the answers to those questions yet. But with a low barrier to entry and a strong conviction in the belief that others will understand the potential of what we are trying to do, we are going to try.

In the posts that follow I will elucidate answers to those questions as well as architect the pyramid, for which Project 1 is a first building block, a proof of concept.