There are two approaches to inhance the visibility: on one hand, using a much high x-ray sensitive detector; on the other hand, using new imaging mechod, e.g. X-ray phase contrast imaging.

There are at least three approaches: besides different physics (detection, phase contrast, etc), use more intelligent software in post-processing.

Apart from excessive X-ray doses, we should aim to reduce amounts of contrast agent -- nasty stuff.

I agree with Jinchuan Gua. However, an increased sensitivity of the detector will not help if there are no -or hardly any- events to detect (too much bony tissue between X-ray source and detector).

The staff can be shielded; the parts of the patient that need to be imaged obviously cannot.

Another (costly) solution is an MRI in the operating room.

I cannot agree more about development of good post-processing.

My experience with CdTe (high sensitivity to high energy X-rays), is that Si can still provide better imaging results on raw data.

And again with Tim, contrast reduction is a must.

Single-photon counting ASIC's can help, as they can select the X-rays energy window, thus reducing the impact of scattered photons and unwanted energies.

Also working on the X-rays source should be investigated, so to further help the detectors doing their job properly.

Detectors with single photon counting with energy discrimination capability (e.g. medipix detectors) should in principle offer the best image-quality/absorbed-dose ratio. I also agree that software post-processing is also very important and it is still an interest research area with a lot of room for improvement.

Optical diffusion might be an interesting method to test when just soft tissues are involved, which act as a translucent media for photons in the visible range. With the appropriate processing it is possible to extract valuable information of the internal density fields. Diffusion Optical Tomography (DOT) is an example of this technique. Nowadays it is also possible to measure the temporal evolution of the light pulse, which provides additional information.

Thank you for all the valuable answers. I believe that an intelligent post-processing algorithm can solve a big part of the problem because such algorithm can improve the quality of the images whether soft tissues or bones are involved. In addition to that, enhancing the quality of images will lead to enhance the diagnostic performance which mostly concerns the medical staffs.

Excellent topic. Thanks for asking! I'm very interested in the answers.

In Bragg XRD Microscopy we are using detector "gain" to:

1. Reduce image acquisition time (dosage as in your case)

2. Enhance weak image contrast.

Ravi, are you sure that using "gain" will actually enhance image contrast?

As per contrast definition, you should get no improvement at all on that side.

C = abs((a-b)/(a+b)), so that if you change your gain factor you get C = abs((ga-gb)/(ga+gb)) = abs(g(a-b)/g(a+b)), whence C = g/g abs((a-b)/(a+b)), and g/g is just 1!

Moreover, I believe you are using a charge-integrating detector, or your comment about reducing image acquisition time would make no sense.

In this case, your improvement in contrast is due only to the signal standing taller vs. noise.

I am not very experienced at that, so I invite the others to also contribute to the discussion.

Using a bow-tie filter reduces the dose, or -if you want- with equal dose, increases the image-quality of cone-beam CT.

Clever software can do a lot, but NOT if there is no signal (e.g. lateral images through the shoulders)!

For several years I've been involved in cone-beam CT reconstruction software. Because I had access to the raw images, some processing could be done before reconstruction. I was able to greatly reduce streak artifacts caused by dental fillings. The reconstructed cone-beam volume 'looked' better than a diagnostic CT-scan.

Unfortunately the algorithm never made it in the clinic (liability issues!@#$%).

Marino! You are most likely correct. However, using detector gain it is possible to acquire weak images quicker, hence minimizing the "dosage". My experience with radiographic imaging is limited. I will learn. Now, enhancing contrast is a different story. For this purpose there are plenty of advanced algorithms available through OPEN CV potentially. As Lambert points out "clever software can do a lot". I'd suggest exploring the following LinkedIn group:

http://www.linkedin.com/groupItem?view=&gid=2642596&item=269436684&type=member&commentID=160463363&trk=hb_ntf_COMMENTED_ON_GROUP_DISCUSSION_YOU_LIKED#commentID_160463363

BTW we are using a scintillator with an Image Intensifier for our work. We are able to detect very weak X-ray diffraction images in real time (30-300fps) depending on the hardware we use.

What is the best spatial resolution and frame rate possible for digital medical radiographic imaging these days?

Within clever software, there are two (equally valuable) kinds of process -- both of which, of course, require a reasonable signal to process.

There is statistical processing like noise reduction, contrast enhancement, etc., 'bottom up' and driven by the data. There is also the 'top down' approach, structured like the human visual system to detect features on partial evidence: in 2D images, for instance, 'snakes' (active contours) do a better job of finding edge curves and boundaries than any bottom-up edge detector can do on its own -- though they need edge intensities in their own input. They can't stand alone.

Like human vision, this inevitably has artefacts. What you see is what you guess, and you sometimes guess wrong. But intelligent guessing (like using knowledge of bone anatomy to remove bones from head CT, improving blood vessel visibility, as pioneered by GE) is an inevitable part of medicine.

The answer depends very much on what limits the visibility of the diagnostic feature: is it quantum noise, system noise, or overlapping structures (aka anatomical noise). Dose level has an impact mainly on the first one whereas for the other two dose is of minor importance.

There are various phase-contrast methods including inline (the simplest), grating methods and finally analyser based imaging although this last is mainly a synchrotron only technique.

Some companies are already working on incorporating phase-contrast into x-ray imaging equipment for commercial use (both inline and grating methods I believe).

Other methods that can increase the information content for a given dose are energy sensitive detectors where dividing the image into different energy bands can provide DEXA type information form a single acquisition (eg see MediPix which is being commercialised here: http://www.marsbioimaging.com/mars/). Also improved detector sensitivity and improved contrast agents may also help but the gains here are probably smaller.

I think it will required to work on the x-ray detetcors so that they can be highly sensitive to very low power/intensity X-rays. Hence it will reduce the high power x-ray requiredment thus reducing X-ray doses. Truely this will be the major component which we can not control through just image processing. After that we can work on the image processing to enhance the raw images.