It amazes me, the way government funded research projects on “the bleeding edge” of science can pour money into projects and all that comes out is a set of a few bar graphs. Of course, the science and work behind them are filled with the glow of human enlightenment (I’m kinda a fan of science…) but why not make it look nice. It just seems so efficient and an easy way to get more bang for the buck. This came to mind while working on my main work project. We’re involved in a multi-site experiment that contains a wealth of genetic, demographic, MRI and clinical data. The multi-site aspect makes it difficult to collect and really get a grasp on. Part of my role, is standardizing and monitoring the way data is collected, stored and analyzed. The only way I saw to really get my claws into it was a series of web based, automatically analyzed and updated web pages. The pages are designed to show different modalities on different pages, drilling down goes from a cursory level to site level to subject level. Before this idea came along, this data was just numerical and inelegant to visualize to say the least. We’ve been able to identify a number of early trends, including errors this way, that otherwise would have rendered millions of dollars worth of collected data worthless. In images, since the permissions to view these are changed regularly. Starting with the home page:

This level summarizes each of the modalities, drilling down into each one:

Flowing down to the subject level yields a large, ugly table. Least you judge, these tables are .csv files that ANYONE can download and quickly work with themselves in R, SPSS or even (ick) Excel. Maybe not visually appealing, but massively efficient.

Never the less, clicking on a subject brings you to visualized single subject data.

Imagine, for each of 100+ subjects, the power of being able to actually SEE your data. With so many subjects and sites collecting data at once it is remarkably unwieldy to manage. These pages generate autonomously and have made a world of difference in the way research is done, and I hope more labs and groups take the time to adopt systems like these. A little richer than a couple of Excel bar charts to say the least!

Advertisements