How to Filtering DICOM Images without any predefined set value ?

How to Filtering DICOM Images without any predefined set value ?

• Image was Dynamic
• Without Set any value
• DICOM Filtering performed based on Image Pixels

[Rik]

Your question is not really clear. What are your inputs? What are your desired outputs? What have you tried so far? The question you're asking is almost impossible to answer in a general way. I would suggest reading the tutorial and editing this question to make it more specific.

[voxey]

Hello Rik:

My Question: 1. DICOM Image Noise removing automatically ?

Answer 1

[Walter Roberson]

You cannot do that.

Cosmic Microwave Background Radiation (CMBR) was thought to be instrument error when it was first noticed while observing stars -- something to be filtered out as noise. When they went to track down the source of the instrument error, they discovered that it is instead a low level signal, and that as they studied more it turned out to have lots of texture. These days they are busy building some of the most sensitive observing devices ever in order to measure it more accurately -- observing devices that have to deliberately block out stars because the stars are too bright.

Thus, considering any one photograph, the CMBR might be noise to be removed if the intention is to observe the stars, but given the same photograph, the stars might be the noise to be removed if the intention is to observe the CMBR.

Therefore there cannot possibly be a fully automated algorithm that removes noise, because "noise" is situational according to intention rather than according to physical characteristics of what is being monitored.

[voxey]

Thank you Walter Roberson, thank you so much,thanks for , given lot of information.

any possiblity to extract high level and low level object[components]. ?

[Walter Roberson]

What do you mean by high level and low level in this context?

[voxey]

Hi @ Walter Roberson

High level --> High intensity components

Low level--->Low intensity components .

[Walter Roberson]

Wavelets perhaps

[voxey]

Hi Walter Roberson,

Can you please give a sample code?

[Walter Roberson]

https://www.mathworks.com/help/wavelet/getting-started-with-wavelet-toolbox.html

Wavelets are in a way similar to fourier transform, in that they operate on multiple frequency scales and break the signal into low frequency components and higher frequency detail. You can do that multiple times, each time breaking up the high frequency component into subsignals. Eventually you get to the point where you decide that the leftover unmodelled part does not contribute significantly to the signal.

This is suitable for some definitions of what "high intensity" and "low intensity" can mean.

Wavelets have been used for texture analysis, which can be import for medical image analysis https://www.researchgate.net/publication/3707527_Wavelets_for_texture_analysis_an_overview

As someone who has been involved in analyzing dicom images from MRI and MRS, I think that your project is doomed to failure. We had direct access to brain surgeons and patientddatasets. We found that relatively crude techniques of analyzing images could classify at about 75% accuracy, and that with a lot of complicated analysis we could push up to 80 to 81 percent, but we couldn't get much more than that.

Until, that is, we switched from MRI to MRS, im which case we could do significantly better, including in some cases better than the top pathologists we were working with. The features that turned out to be important were seldom immediately obvious from the spectral plot, and we broke new medical grounds in discovering that those chemicals were involved in those kinds of tumors.

You are asking for the ability to to have a program take any arbitrary dicom image of any body part and automatically figure out what parts are important and what parts are not, without any guidance as to what is being investigated. That is not going to work.

I have some friends who are technically obese. When they go to the doctor, it is not uncommon that the doctor starts and ends with lecturing them about losing weight, and they often have difficulty getting the doctor to even pay attention to what they went to the doctor about. One of them hurt their arm and the doctor refused to even look at their arm until they got angry at the doctor.

Another of them spent over a decade attempting to convince the doctors to check for a hernia, each time being refused and told that there was nothing wrong with them except that they needed to lose over 100 pounds. When they finally convinced a doctor to order hernia tests, the tech said they had never before seen such a large hernia, and the first doctor to look at the results sent the tests back to the lab complaining that the test had been administered improperly, that no one could have such a large hernia.

Your proposed program, if implemented, would inevitably do the equivalent of just telling patients that they needed to lose weight, as that would be by far the largest most intense signal: without guidance about what was being investigated your program would ignore the fractured arm bone or the hernia.

[voxey]

Walter Roberson

Thanks for spending your golden time, thanks for guidance . i got lot off knowledge from you. thank you so much.