Convert pixel/grayscale values ​​to decibels

I would like to know how to convert uniformity in ultrasound images into decibels.

I found that uniformity testing procedures involve several steps:

1. Apply coupling gel to the scanning surface.
2. Position the transducer of the standard ultrasound system above a phantom region without targets, only containing a speckle pattern.
3. Adjust instrument settings individually in the ultraosund equipament
4. Freeze one image for each setting combination and obtain a hard copy.
5. Perform computer analysis by placing a series of 4 small ROI (Region of Interest) along the axial direction. Compute the average pixel value (PV̅̅̅̅i, i=1,...,4) in each ROI. Calculate the difference between the average pixel values of the two quadrants with the highest (MAX(PV̅̅̅̅i)) and lowest (min(PV̅̅̅̅i)) average pixel values, recording this result as the uniformity parameter:U = MAX(PV̅̅̅̅i) - min(PV̅̅̅̅i)This parameter represents the maximum variation of mean grey levels among discrete quadrants. To compare results with tolerance limits suggested by international standards, uniformity is converted to the decibel scale (how?).

In the following image, we have an example of automatic selection of ROIs on MATLAB for image uniformity assessment.

Any help is welcome!

Answer 1

[William Rose]

@ESTER LIMA, given two levels, PVmin and PVmax, whose values are proportional to the intensity of the ultrasound signal, the decibel difference is

10*log(PVmax/PVmin)

I used a factor of 10 above, not 20, because I assumed the gray level is proportional to intensity rather than amplitude. If the gray level is proportional to amplitude, then the dB value is

20*log(PVmax/PVmin)

There is a reasonable chance that the actual grayscale value is not proportional to either amplitude or intensity, because the user can adjust image gain and contrast. This creates a nonlinear mapping from amplitude (or intensity) to gray level. Ultrasound devices may use proprietary or public domain algorithms to map from intensity to gray scale. Matlab's histeq() is an example of a public algorithm for this purpose. In such cases, I would use the actual ultrasound intensity ratio, and the fomula above with "10", for the dB difference. But I can't tell you how to extract the intensity from the grayscale level. You would have to know more about the machine's grayscale mapping.

[ESTER LIMA]

@William Rose, thanks for the reply!

After doing the 4 rois in imageJ and getting the highest (MAX(PVi)) and lowest (min(PVi)) average pixel values, I put them in the formula 10*log(PVmax/PVmin).

I tried using a factor of 10 and 20, applying this methodology to a uniform and a non-uniform image. However, following the regulations I sent in the original question, images with more than 4 decibels would be non-uniform, which was not in agreement with the result after applying this formula. I found values ​​lower than 4dB for both images.

The images I used to test this are in the following link: https://drive.google.com/drive/folders/1tlz8eVrd79o6kQwMeo0d3G3rGkT5_33f?usp=sharing

[William Rose]

@ESTER LIMA, I could not open those image files. They have no suffix. I tried adding .jpg and .tif and .bmp, but could not open the image file.

[ESTER LIMA]

@William Rose To open these images in DICOM format, it is necessary to download the folder and open it in a program compatible with images in this format. The simplest one is ImageJ, which I believe you are familiar with. I also managed to open them using Python with the pydicom library.

[ESTER LIMA]

@William Rose

DICOM format is essential in medicine for securely storing and sharing standardized medical imaging data, allowing efficient analysis and diagnosis across different imaging modalities.

Anyway, I just uploaded to the folder using the same link, both images in PNG format.

https://drive.google.com/drive/folders/1tlz8eVrd79o6kQwMeo0d3G3rGkT5_33f?usp=sharing

[William Rose]

@ESTER LIMA, I'm afraid I do not have more time to work on this problem. Good luck with your analysis.