Ok, I'll admit it. I'm not in New York City. Earlier this week I returned to Ithaca to catch up on the latest cow tipping action, and to move out of my apartment. Three days later, my entire life is in a series of carefully sealed and labeled boxes, my movers still haven't shown up ("Today! We promise. We promised yesterday? Well we mean it this time!"), and I'm sitting on a living room box typing a BME blog entry on my cell phone. That has to count for something, right?
Any ways, back to my project. We haven't fully worked out the automatic segmentation yet, as Noel is still tweaking a few things, but from the look of it, the comparisons between the Phase Contrast derived measurements and the manually segmented magnitude images are intriguing. It appears that while they have a somewhat high correlation (0.95 or greater), one always seems higher than another.
From these early results, it is easy to see that the two forms of "ground truth" may not even agree! How, then, can we ever judge the accuracy of te automatic segmention algorithm? Perhaps it will be similar to the PC flow analysis, but not the magnitude based measurement. And if so, is this more accurate than i it was the other way around?
An interesting question that, for me in any case, remains unsolved. Looks like this phone doesn't seem to get along with PubMed. Alas...