ABSTRACT
AIM: To evaluate the performance of normalised apparent diffusion coefficient (ADC) values for prostate cancer assessment when performed by independent observers blinded to histopathology findings. MATERIALS AND METHODS: Fifty-eight patients undergoing 3 T phased-array coil magnetic resonance imaging (MRI) including diffusion-weighted imaging (DWI; maximal b-value 1000 s/mm(2)) before prostatectomy were included. Two radiologists independently evaluated the images, unaware of the histopathology findings. Regions of interest (ROIs) were drawn within areas showing visually low ADC within the peripheral zone (PZ) and transition zone (TZ) bilaterally. ROIs were also placed within regions in both lobes not suspicious for tumour, allowing computation of normalised ADC (nADC) ratios between suspicious and non-suspicious regions. The diagnostic performance of ADC and nADC were compared. RESULTS: For PZ tumour detection, ADC achieved significantly higher area under the receiver operating characteristic curve (AUC; p=0.026) and specificity (p=0.021) than nADC for reader 1, and significantly higher AUC (p=0.025) than nADC for reader 2. For TZ tumour detection, nADC achieved significantly higher specificity (p=0.003) and accuracy (p=0.004) than ADC for reader 2. For PZ Gleason score >3+3 tumour detection, ADC achieved significantly higher AUC (p=0.003) and specificity (p=0.005) than nADC for reader 1, and significantly higher AUC (p=0.023) than nADC for reader 2. For TZ Gleason score >3+3 tumour detection, ADC achieved significantly higher specificity (p=0.019) than nADC for reader 1. CONCLUSION: In contrast to prior studies performing unblinded evaluations, ADC was observed to outperform nADC overall for two independent observers blinded to the histopathology findings. Therefore, although strategies to improve the utility of ADC measurements in prostate cancer assessment merit continued investigation, caution is warranted when applying normalisation to improve diagnostic performance in clinical practice.