Can DWI improve the diagnostic performance of the existing NI-RADS MRI descriptors?

Abstract Background The current ACR NI-RADS MRI lexicon does not include data about DWI. We hypothesised that incorporating DWI into the existing NI-RADS MRI descriptors could enhance its diagnostic performance. Methods This was a prospective study of 52 head-and-neck cancer patients who came for contrast-enhanced MRI surveillance 8-10 weeks after completing their curative treatment. Image analysis assessed the primary tumour site using the current ACR NI-RADS MRI morphologic criteria. Then, DWI was evaluated, and NI-RADS rescoring was performed based on qualitative DWI findings. An upgrade or downgrade of the existing NI-RADS categories was done, and diagnostic accuracy was calculated. Results Local tumour recurrence occurred in 30 patients (57.7%) while 22 patients (42.3%) had post-treatment changes. Final diagnosis was confirmed by histopathology in 26 patients (50%). Using morphologic data, 13 lesions were assigned the NI-RADS 1 category, 16 were assigned the NI-RADS 2, and 23 lesions were assigned the NI-RADS 3. When we added DWI data, eight NI-RADS 2 lesions were upgraded to NI-RADS 3, and seven of them were proven to be recurrent tumours. One NI-RADS 3 lesion was correctly downgraded to NI-RADS 2, which was proven on follow-up imaging to be post-treatment changes. Conclusions The combination of both morphologic ACR NI-RADS MRI criteria and DWI for primary tumour site assessment improves the overall diagnostic performance in predicting the presence of tumour recurrence.