Interval changes of histogram-derived diffusion indices predict treatment response to induction chemotherapy in head and neck cancer: a feasibility study

BACKGROUND: This retrospective study investigated whether the interval change of apparent diffusion coefficient (∆ADC) [baseline and after the first cycle of induction chemotherapy (ICT)] can be used as a valid predictive imaging biomarker of the treatment response to ICT in head and neck cancer (HNC). METHODS: A total of 19 consecutive patients with HNC who underwent diffusion-weighted magnetic resonance imaging (DWI) at baseline and after the first cycle of ICT were included. Whole-tumor ADC histogram parameters (mean, median, kurtosis, skewness, entropy, minimal, maximum, 25th percentile, and 75th percentile) were obtained. The correlations of ∆ADC histogram parameters, volume, T-stage, N-stage, and age with the treatment response were examined using the Mann–Whitney U test. The predictive value of histogram parameters was examined using receiver operating characteristic (ROC) curves. RESULTS: Responders showed significantly higher values of ∆ADC(25) (0.19±0.23) and ∆ADC(min) (1.78±2.98) than non-responders (−0.09±0.15 and −0.73±0.36; P=0.035 and 0.009, respectively). When ∆ADC(25) and ∆ADC(min) were used for predicting the treatment response, the area under the ROC curve was 0.850/0.933 with a sensitivity of 73.3%/80.0% and specificity of 100%/100% (P=0.036 and 0.009, respectively). CONCLUSIONS: ∆ADC(25) and ∆ADC(min) derived from whole-tumor histogram analysis are valuable imaging biomarkers for the early prediction of the ICT response in HNC.