RA-08: COMPARISON OF NORMALISATION TECHNIQUES FOR T1-WEIGHTED SUBTRACTION MAPS

PURPOSE: Methods to improve tumour response assessment, such as RANO, utilize enhancing tumour size on post-Gd T1w scans (Wen,2014). Recently, volume changes of brain tumours from contrast-enhanced T1w-Subtraction maps have been suggested to improve survival prediction compared to unsubtracted enhancement-ROIs (Ellingson, 2014). Different normalisation techniques were compared to investigate which produced the highest tumour-to-background contrast on T1w-Subtraction maps: normalisation by mean values of ROIs drawn on white-matter, post-Gd nasal mucosa, pixels that didn't change between the pre and post-Gd images, two-value scaling and normalisation, shifting the mode of the pre and post-Gd images, and Gaussian normalisation (Ellingson, 2012). MATERIALS AND METHODS: T1w-Subtraction maps from 16 consented patients were used. Pre and post-Gd images were registered to FLAIR, normalised and subtracted to obtain T1w-Subtraction maps. FLAIR abnormality ROIs were used to define the whole tumour, and tumour contrast was calculated as the ratio between the T1w-Subtraction map values in the tumour ROI and the rest of the (non-enhancing) brain. RESULTS: The highest tumour contrast (1.49 ± 1.16) was obtained by normalising by the average of pixels whose values did not change between pre and post-Gd, but the difference was not significant (p > 0.05) with respect to the other methods. The exception was nasal mucosa normalisation which, due to its high enhancement threshold resulted in the cancelling out of the majority of background and artefactual enhancement, as well as low enhancement values within the tumour. CONCLUSION: These results show the normalisation threshold not a significant factor for determining enhancement within tumour ROIs in T1w-Subtraction maps.