4 ms·

X-ray ‘ghost images’ could cut radiation doses

25 points by yskchu 9y ago

gus_massa 9y ago
Is this "compressed sensing"? Is "ghost images" an official name or it was made up just for this article?
- phyzome 9y ago
  I think I've seen this term used before for single-pixel cameras.
m3kw9 9y ago
I don’t get if this light source is the x-ray source and how shooting it through a filter thousands of times lowers radiation?
- VikingCoder 9y ago
  Each detector pixel receives a tiny fraction of the emitted radiation, right? The problem is that xrays scatter. So it's not entirely obvious what path the ray took as it passed through the body. That's noise. There are a lot of techniques to minimize that noise, but it's always there. Okay, now switch to a single pixel detector, with no filter, and picture if you left the amount of radiation exposure the same... Well, there's just not enough information in the signal. But if you add a spacial filter, and you know its exact structure, then that gives you a lot of information. And it turns out that the math is such that it reduces noise A LOT. So, then you have a choice - keep the radiation the same, and increase the resolution... Or reduce the radiation, and accept the same resolution you had before. Well, you may not be able to actually increase the resolution, for lots of reasons. So you might not be able to make that trade-off. I know that's hand-wavy, but I hope it helps.
  - petermcneeley 9y ago
    Ya I dont know how much sense this makes. Knowing the final detected position is information that you are throwing away when you have a single detector. The ultimate would be to know the exact ray as it entered and then the exact position (and angle) as its detected.
    - VikingCoder 9y ago
      "The ultimate," sure, but in practice that's not going to happen. Also, don't forget, there are gaps between the pixel detectors. Maybe there's just not enough information in multiple detectors to make the aggregate signal loss of those gaps worthwhile.

25 points by yskchu 9y ago

gus_massa 9y ago

Is this "compressed sensing"? Is "ghost images" an official name or it was made up just for this article?

phyzome 9y ago
I think I've seen this term used before for single-pixel cameras.

m3kw9 9y ago

I don’t get if this light source is the x-ray source and how shooting it through a filter thousands of times lowers radiation?

VikingCoder 9y ago
Each detector pixel receives a tiny fraction of the emitted radiation, right? The problem is that xrays scatter. So it's not entirely obvious what path the ray took as it passed through the body. That's noise. There are a lot of techniques to minimize that noise, but it's always there. Okay, now switch to a single pixel detector, with no filter, and picture if you left the amount of radiation exposure the same... Well, there's just not enough information in the signal. But if you add a spacial filter, and you know its exact structure, then that gives you a lot of information. And it turns out that the math is such that it reduces noise A LOT. So, then you have a choice - keep the radiation the same, and increase the resolution... Or reduce the radiation, and accept the same resolution you had before. Well, you may not be able to actually increase the resolution, for lots of reasons. So you might not be able to make that trade-off. I know that's hand-wavy, but I hope it helps.
- petermcneeley 9y ago
  Ya I dont know how much sense this makes. Knowing the final detected position is information that you are throwing away when you have a single detector. The ultimate would be to know the exact ray as it entered and then the exact position (and angle) as its detected.
  - VikingCoder 9y ago
    "The ultimate," sure, but in practice that's not going to happen. Also, don't forget, there are gaps between the pixel detectors. Maybe there's just not enough information in multiple detectors to make the aggregate signal loss of those gaps worthwhile.