MRSI-data frequently contains bad-quality spectra, what can prevent proper quantification and consequently lead to data misinterpretation. Machine-learning based methods have been proposed for automatic quality control of MRSI-data with performance levels identical to expert’s-manual-checking and that can classify thousands of spectra in a matter of a few seconds. Besides this, a considerable amount of time needs to be spent labelling data required to train these algorithms. Here we present a method that allows to actively select those spectra that carry the most information for the classification, allowing to reduce drastically the amount of time needed for labelling.
This abstract and the presentation materials are available to members only; a login is required.