A way to segment bone in the Ultrasound Image

6 ビュー (過去 30 日間)
Stelios Fanourakis
Stelios Fanourakis 2019 年 6 月 30 日
As you can see I attach two images. The original us image and the other is after applying a harry's edge detector filter.
As it is obvious, there is a small pattern of white pixels outlining the bone area in the processed image.
Two questions:
1) Is there a better method to preprocess the image and eliminate the randomly white pixels (dots) on the processed image and keep only the major two white lines (curves). The one at the top that looks like a downward curve (hemisphere) and the bone outline. All the rest white speckle dots should be removed.
2) I was thinking of an algorithm to create a pattern of the white pixels in the bone area. One that will estimate certain distance from the white pixels area of the bone (neightbor pixels) and register them, creating a specific pattern of white pixels that can be segmented.
I hope that I can be clear in my thoughts.
Waiting to your feedback.


Image Analyst
Image Analyst 2019 年 6 月 30 日
You could try a median filter or something better. See Vision Bib for some ultrasound processing algorithms.
Or you could just threshold and clean up afterwards, like throw out blobs smaller than a certain area or with an aspect ratio that is not right.
  5 件のコメント
Stelios Fanourakis
Stelios Fanourakis 2019 年 7 月 1 日
Thank you very much Dr Image Analyst. Your contribution is more than well appreciated and works marvellous. However, in my case I want it to be automated. It is quite easy to be done using some edge detector filters like the Sobel one or the Noble one, that I have tried so far. However, they come up with lots of noise.
At the moment, I am searching for all possible methods (literature) to do it before I decide to which one of those I will finally employ to my project. If it would be automated and a contour was also generated at the same time that would be magnificent. Since if there would be something like a convox hull (closed contour area of the bone surface and magnitude) that would be easy enough to recontruct it into a 3d object.
Anyhow, thanks very much for your contribution.


その他の回答 (0 件)


Find more on Biomedical Imaging in Help Center and File Exchange

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by