How to label data for data train deep learning.

Dear All,

Now, I have the origionally data J0ACSC PET 1HR

(data can dowload here: https://drive.google.com/file/d/18CCXXr80BONM5s771uw9uv5cpNh2hrmZ/view?usp=sharing).

And I want to label it for data trained as data J0ADRM Local Energy Deposition 1HR

(data can download here: https://drive.google.com/file/d/11VpJzVUB37ATdX5ex_kzSKhcp565kBI3/view?usp=sharing).

Meaning that I want to label the data from J0ACSC PET 1HR to J0ADRM Local Energy Deposition 1HR.

May I know how to label it? because the value for labeling is every single pixel.

Answer 1

[colordepth]

Unfortunately, it seems the J0ACSC PET 1HR dataset you provided does not have any labels for reference. I assume you are trying to label the voxels for the kidney-like blobs in the volume. You can use MATLAB's “Volume Segmenter” app, which includes unsupervised algorithms to help with labelling 3D volume data. To begin, refer to the guide at https://www.mathworks.com/help/releases/R2023a/images/create-semantic-segmentation-using-volume-segmenter.html.

You can start by selecting a slice from the middle of the volume and use the paint brush tool to label the regions of interest. For larger areas, the "Paint by superpixels" feature can be useful, while the regular paint brush tool can be used for more detailed work.

After labelling an initial slice, you can use one of the unsupervised labelling algorithms under the "Automate" tab to automatically label the remaining slices.

Experimenting with different algorithms and adjusting their parameters can help you achieve the desired results 😊.