2
$\begingroup$

How to objectively assess the quality of a medical image(CT,PET,MRI)? What factors should be considered?

$\endgroup$
  • $\begingroup$ Basically I am using SSIM $\endgroup$ – gmotree Jul 9 '15 at 0:58
2
$\begingroup$

It is a million-dollar question. It depends on the modality and the task the radiologist has to accomplish.

The gold standard is to give radiologists your images, assign them a task and measure how effective they are. However, this requires time, money and clinical images. Typically, this is used as a last resort to convince the regulatory agencies (the FDA for instance) that your medical imaging system is effective.

There are other ways of doing this without radiologists, with "model observers". Basically, you define an algorithm (often a variation on linear template-matching) with is supposed to mimick the performance of a human observer, and you measure the performance of the algorithm. However, not all tasks and modalities are suitable to such an approach.

You can also use metrics measured on the image such as the Signal Difference to Noise Ratio, or metrics measure on the whole imaging system, such as the Detective Quantum Efficiency for X-ray systems.

For a good overview, read this: http://www.ncbi.nlm.nih.gov/pubmed/9419600 Image quality, the ideal observer, and human performance of radiologic decision tasks. Burgess A.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.