Seems my previous post did not make it through. You might need to develop a customized segmentation algorithm/software for this task, and due to the complexity of the segmentation in this context, you could benefit from the power of deep/machine learning if you have labeled data.

Hello,

According to ImageJ's wikipedia page https://en.wikipedia.org/wiki/ImageJ

"ImageJ can calculate area and pixel value statistics of user-defined selections and intensity-thresholded objects." So the short answer to your question is "no."

The longer answer is that you probably need a software implementation of a semantic segmentation algorithm. The ultrasound data you present is naturally very noisy and the object of interest does notseem to have clear delineation. It seems reasonable to say that you cannot use histogram thresholding or another simple image processing method like Otsu, grayscale watershed, etc, to segment this data.

You should have a look at convolutional neural networks, which are good at solving semantic segmentation problems provided that you have enough labelled (annotated) image data. A trained neural network can be deployed in Python or C++ but you need a GPU to implement the convolutions in a reasonable time. The U-Net architecture developed by Uni of Frieburg in 2015 has been quite successful at segmenting biological images, including ultrasound.

Please recommend this answer if you find it useful.

I can recommend Segment, it is freely available for research purposes, but processing only MRI, CT and Myocardial perfusion SPECT images:

http://medviso.com/segment/

Unlike MRI/CT modalities, Sonography exhibits difficulties in finding the final delineation of a ROI. If you still want to stick to ImageJ, then you could check the "Trainable Weka Segmentation" plugin

https://imagej.net/Trainable_Weka_Segmentation

Abbas Cheddad Graham W Pulford

Thanks for the tips. I am not familiar with the world of deep learning and neural networks, but perhaps it is time to dive in. For this short term project it will not be profitable, I think that I should focus on semi-automatic ways to segment the myocardium.

If someone knows a macro/script for imageJ or any other software to realize this, I would be happy to know. I will also take a look at the trainale weka segmentation plugin.

Svitlana Alkhimova Thanks for the tip. Unfortunately I could not get it working with US images.

Glad to be of help! Weka is believed to be easy to use since it is GUI based and meant also for teaching. Good luck with the dive!