Is Image Labeler App a good choice for Labelling unlabeled data?

I am supposed to develop a CNN for medical image analysis. For that, I am collecting data from hospitals and they are not labeled. But I have some labelled data obtained from online platform. Would I be able to use those labeled data to label the unlabeled data using image labeler app in matlab. Is it a good choice for labelling? Does someone had do it before and had have a good results?

Answer 1

[Jaynik]

Hi Kalhara,

The Image Labeler app is indeed a good choice for labeling unlabeled data. It provides an easy way to interactively create different shapes to mark as regions of interest (ROI) labels, which is required for training algorithms like CNNs.

For medical images specifically, you can use the Medical Image Labeler app which is specifically designed to label pixels in 2D and 3D medical images. This app allows you to export labeled data as a groundTruthMedical object to train semantic segmentation algorithms.

You can refer the following examples to learn more about using the Image Labeler app:

• Medical Image Labeler: https://www.mathworks.com/help/medical-imaging/ug/get-started-with-medical-image-labeler.html
• 3D Image Labeling: https://www.mathworks.com/help/medical-imaging/ug/Label-3-d-medical-image-using-medical-image-labeler.html
• Image Labeler: https://www.mathworks.com/help/vision/ug/get-started-with-the-image-labeler.html

For using previously labeled data to label the unlabeled data, you will need to train a basic model and apply the model to the unlabeled data using the Image Labeler app. You can check out this example to automate the labeling task:

• https://www.mathworks.com/help/medical-imaging/ug/automate-labeling-in-medical-image-labeler.html

Hope this helps!

[Kalhara]

@Jaynik Thank you so much for your detailed answer for my question.I'll definitely check them out for my research analysis.