אין עדיין תגים.
So in the last blog post I addressed what generally texture analysis might do for us. The topic is too broad to summarize in one blog post, so I just threw up one general article and some nice video where you could see one use. And then I spoke to another colleague who far outranks me. You would think I'm speaking in my own language. Digital biopsy- what the heck and I talking about?
Let me start out by saying that radiomics and texture analysis can not replace pathological biopsy, at least at this point. But it can tell you things you aren't going to get from biopsies. Like- how well is this tumor responding to treatment? Do I need to do another biopsy? And so on....this deserves serious articles, not a blog post but anyways...
Let's get on to part 2...why on earth would you use the program I developed and not something more well known? So without making a bunch of graphs no one would understand or agree on I'll break it down:
Totally Commercial: Oncoradiomics, TexRad etc.
Benefit: nice shiny interfaces, off the shelf...
Problems: You can't look at the code and tell what the heck the program is actually DOING. For research purposes this is a non-starter...
Semi-commercial: Matlab.
Benefit: You CAN control come of the code, and there are lots of people around who have been programming inside MATlab for decades
Problems: You need to PAY for this mess...I hear it's like a secret love child of scientists who don't know how to program and programmers who secretly hate scientists...but I wouldn't know can't afford buying not only the basic program , but all the add on modules I would need, then having to write a program to look at what I want to look at anyways....I mean if you have to write the program anyways- why pay?
Open Source: Debora + Voxelmath (my programs), various plug-in modules for Slicer, MaZda
Benefit: You CAN control the code, completely, and they are FREE
Problems: Not as flashy as commercial projects.
Now inside the open source products, mine has at least one interesting unique quality- you can set the bin size on histograms. Which means you can do some things with entropy even those guys with the funded lab and prestige can not....if you know what you are doing....
This brings up an interesting issue- do radiologists know what they are doing with computers? Well, I would describe the situation this way: analagous to the car situation.
Almost all radiologists drive to work in a car, or get to work on a bus. Some percentage of radiologists have no idea what is going on in their car, they just get in their car and drive. They probably picked which car they bought by the color, or because their best friend said it was good or...well, you get the point. Other radiologists know the basics about cars, and they understand when they read about them. Those radiologists could, if they had to, look under the hood and change the oil. Now here is the interesting point- almost ZERO radiologists are also mechanical engineers who work in the automotive industry and could make that car fly if they wanted. So to bring this analogy home, I'm sure there are some radiologists who are also PhDs in computer vision, and also work for industry....but for mere mortals like myself, computers are just a tool we try to deal with and not crash. Yes, I know, they have fellowships in informatics...but if you look closely at the requirements and curriculum for these, it's obvious that they are not going to turn a professional radiologist into a professional computer scientist as well. Maybe in the future just as there is a physicist or two in each radiology department, there will also be an informatics person. Maybe in the future Grainger's will have not only Physics sections, but also computer informatics sections...Who knows.