[Aeskulap-devel] More opinions

[Anthony Butler]

Hi All,
  I'm pleased to see this list is starting to get some traffic!

First off, I'll introduce myself to give a background to my opinions. I'm a 
radiologist in NZ. I'm currently taking time of clinical work to do a PhD in 
computer aided diagnosis systems (eigenimaging) through my local Uni's 
computer engineering dept. As a clinician I've regularly used efilm, 
intelliviewer, and osirix.

Please don't take these opinions as anything other than constructive feedback. 
They are really just a wish list from a single person. I think aeskulap has 
the most potential of any cross-platform open source viewer I've tried. 

I'm interested in aeskulap because as I'm approaching the end of my thesis I 
want to be able to put the image processing routines I've worked in into a 
clinical viewers. Currently all my image processing work has been developing 
routines matlab. If I want people to test/trial them in clinical practice I 
think the best thing is to put them is in an open source viewer. Apart from 
osirix which is tied OSX libraries, aeskulap seems the most promising and 
stable target.

Mitchell L. has been talking about user interface. I have to agree with him 
about what functions are important. Modelling the interface around efilm is a 
good idea. Its what most people know. Intelliviewer is similar, but allows 
for customisation and I personally prefer it and have used it for the last 18 
months. 
http://rsna2005.rsna.org/rsna2005/V2005/exhibitor_list/displayEx.cvn?exbID=842

General interface
============

Most people work with the right hand on the mouse, left hand on the keyboard- 
similar to playing a 1st person shoot-em-up game. This means all important 
keyboard functions need to be a single key, not key combinations such as 
"crtl-a"

Top level functions on the mouse should be
1) Scroll
2) Zoom
3) Pan 
4) window/level
These are the 4 most used functions while reporting. I personally like Pan on 
left buttom, zoom on right buttom, scroll on the wheel, wheel-button window. 
However since going to 64 slice CT we now have to scroll a lot more and some 
people are prefer L-pan, R-scroll, wheel-zoom, wheel-button window. Hence it 
needs to be user configurable. 
F1- F8 (or keys 1-8) should be a set of preset window/levels values.

I like the keys "home" and "end" to take me to the first and last images 
within a stack.

I also agree with Mitchell that pan is currently too limited. You should be 
able to pan any part of the image to any part of the screen. Arbitrary 
rotations and image mirroring (flip L-R) will be need to be implemented at 
some point. Maybe not till later in development, but you should plan for 
them. 

Measurement tools
==============

>What I'm currently missing is a list of needed features on "pixel value"
There are the two measurement tools I consider essential. They constitute 
around 95% of what I use.

1) As Mitchell said, a region on interest (ROI). eg circle/oval which report 
the mean HU within that circle. This is essential for enhancement of tumours 
etc. (eg adrenals). When commenting on density+enhancement you need the mean 
of a small sample of pixels, hence the ROI. Individual pixel values are much 
less relevant due to image noise.

2) Distance measurements. These are required for measuring tumours, midline 
shift, size of stones, depth of fluid etc. No viewer could be used without a 
distance measuring tool. Later in development things like, angles, area of 
ellipses, COBB angles can be added, but they are not essential for most 
things other than orthopaedic reporting.

I've seen these features implemented using several different interfaces. 
Personally I prefer a single keystroke to activate them, have the mouse 
pointer change to show the change in mode, use the left mouse buttons to take 
1 measurement, then return to normal mode.
ie. to measure a line
1) push "L" on the keyboard
   mouse pointer changes (?have  the word "line" above it)
2) left click to mark start of line
3) left click to mark end of line
   mouse pointer changes back to normal
   line is displayed with measurement number on beside it. eg "1"
  distance is displayed in bottom left corner of screen
  "1 Line:  10mm"

A similar interface for all measurement tools is required. ie ROI you click 
"r" to active the mode. Bottom left of screen is displays "2 ROI: 55 HU. Good 
viewers let you customise the colour, location, units, and font size.

If you enter the mode by accident, or decide you didn't need it. Pushing the 
"esc" key should stop the measurement and return you to normal mode. Pushing 
the "delete" key and selecting a measurement (eg ROI or line) should remove 
that measurement.

Image processing plugins
==================
What I really like to see is a plug in feature where people (including me) can 
write image processing routines without having to delve to deeply into the 
code. Something similar to the plug in interface osirix would be ideal. This 
will allow for image enhancement and computer aided diagnosis routines to be 
developed outside the core functionality of the viewer. Now is probably the 
point in the development where you need to think about how this might be 
implemented. 

http://homepage.mac.com/rossetantoine/osirix/Index2.html

There are heaps of routines for registration, segmentation, enhancement, 
feature extraction etc. that should be implemented outside the core 
functionality of the viewer. The interface should allow passing of image 
data, patient data such as age, and user selected control points to a  plug 
in. The plug in can then pass back a processed image +/- some text to be 
displayed.  



Hope this helps and keep up the good work.

Anthony





-- 

Anthony Butler
address@hidden
http://www.butler.co.nz/
skype:elivs.nz

Today is the first day of the rest of your life.