Wednesday, August 02, 2006

Edge Detection Rule(r)s!!


I haven’t posted much in the way of pure PACS lately, in my quest to be a politician and policy maker (or breaker as the case may be). My good ideas seem to be coming less and less frequently; must be old age, or stress, or some combination thereof.

Finally, my rusty steel trap of a mind has come up with something new. Well, it really isn't new, but it is a simplified and useful twist on an old concept.

Edge-detection is a well-known mathematical procedure, useful in the detection of abnormalities on mammograms, and nodules on CT's. There are a number of Artificial Intelligence (AI) programs out there that sniff out badness on these exams and do all sorts of measurements allowing for followup.

My idea is much simpler, and applicable to the everyday PACS world. In fact, it occured to me whilst measuring innumerable lung nodules on a cancer patient. It is a true pain to draw in short and long axes, especially when the nodule is in the subcentimeter range (and even more so when your non-averaGE PACS won't let you show more than one pair of measurements on the screen.) So, let's create an edge-detection-based ruler! As in the rather primitive mock-up above, the user draws a circle or ellipse around the nodule in question, and tightens up its radius to a reasonable level. The computer then takes over, detecting the edge of the nodule, finding long and short axes, and even giving average density, total area, etc. Yes, this could be expanded into 3D, seeking the nodule in the slices above and below, but that adds so much complexity, it probably isn't worth it. All I want is my measurements done for me without tedious submillimeter movements of the mouse. The Dalai Edge Detection Ruler would do the trick.

OK, which PACS company is going to make this happen? I'll take 50% royalties, thank you. Uh, how about 30%? 20%? Oh well, just put this in action and I'll be happy!

3 comments:

Anonymous said...

Hi Doc, what you write about measurement in GE PACS is a little outdated. It used to be like that in the past, but the newer versions can display 4 measurements + 1 COBB's angle, out of all the created annotations. Just for the sake of truth. ;-)

Dalai said...

Thanks, GE of UK for that. Glad to see you support your Yankee cousins! Version 2.x, the one I still have to use, has the two measurement limit, as did Version 3.0 as tested at RSNA. You are telling us that Version 3.x, or maybe 4.x, allows double the number of measurements! WOW! Now, tell me why GE presumes to make the medical judgement that I only need to see two, or four, or eight, or 20 measurements on screen at any one time? You are interfering with patient care on that one, and you should really reconsider that approach. When do we get rid of the all-caps and space-after-comma restriction for searches, hmmmm?

Anonymous said...

How about clicking in the middle of the module and the computer does the rest? Can you send me some images with nodules? I would like to do a try...
rafael[at]charrua.com