Synthetic data (illustration)

Case contributed by Dr Candace Makeda Moore
Diagram

The image appears close enough to an actual hand x-ray with foreign bodies that it could be used as synthetic data in a neural network to classify x-rays, although it was actually created by editing the image. 

Diagram

The second image is an example of radiology art, and only in certain very specific contexts would be considered synthetic data. 

Case courtesy of Dr Matt Skalski, Radiopaedia.org, rID: 43911

Case Discussion

Synthetic data is often used in training data for machine learning algorithms. The distinction between synthetic data and radiology art depends on the desired classification of a machine learning algorithm.

The second image would not be considered synthetic data for any actual radiological machine learning programs; however, if a program were made to classify x-rays as opposed to black and white photography, then the second image might be legitimate synthetic data.

Synthetic data can also be contrasted with augmented data, which can also be used to increase the size of training data sets for AI but can be created from real data with simple linear algebra functions.

PlayAdd to Share

Case information

rID: 74242
Published: 11th Feb 2020
Last edited: 11th Feb 2020
Inclusion in quiz mode: Excluded

Updating… Please wait.

 Unable to process the form. Check for errors and try again.

 Thank you for updating your details.