Yoshihiko Fukukura^{1}, Toshikazu Shindo^{1}, Yuichi Kumagae^{1}, Koji Takumi^{1}, Hiroto Hakamada^{1}, Masanori Nakajo^{1}, and Takashi Yoshiura^{1}

^{1}Radiology, Kagoshima University Graduate School of Medical and Dental Sciences, Kagoshima, Japan

### Synopsis

**This study focused on the
potential of ADC histogram analysis on DW imaging to characterize solid
pancreatic masses. Among the ADC histogram parameters, the entropy of ADC with
every b-value combination showed the highest area under the receiver operating
characteristic curve for distinguishing neuroendocrine tumors from pancreatic
adenocarcinomas and mass-forming autoimmune pancreatitis. The entropy of ADC
might add helpful information in differentiating neuroendocrine tumors from
pancreatic adenocarcinomas and mass-forming autoimmune pancreatitis, especially
in patients with contraindication to contrast agents or with solid pancreatic
masses showing atypical findings at dynamic CT or MRI.**### Purpose

Diffusion-weighted (DW) MR imaging is used for
various aspects of the evaluation of pancreas lesions such as detection,
diagnosis, and predicting patient prognosis

^{1-3}. Apparent
diffusion coefficient (ADC) histogram analysis is a reproducible technique,
and several ADC histogram parameters are more complementarily or effectively
reflect the microstructure of tumors

^{4}. However, the utility of ADC
histogram analysis in characterizing solid pancreatic masses has not been
elucidated. Therefore, the purpose of this study was to investigate whether ADC
histogram analysis of DW imaging can characterize solid pancreatic masses.

### Methods

One hundred ten patients with histologically
confirmed 66 pancreatic adenocarcinomas (PACs), 27 neuroendocrine tumors (NETs),
and 17 mass-forming autoimmune pancreatitis (AIPs) underwent
respiratory-triggered fat-suppressed single-shot echo-planar DW 3.0-T MRI with
b-values of 0, 200, 400, and 800 s/mm

^{2}. The pulse sequence
parameters were as follows: repetition time, which was based on the respiratory
interval; echo time, 60 ms; flip angle, 90°; field of view, 350 mm; matrix, 60 x 112; number of excitations, 2 (b-values of 0,
200, and 400 s/mm

^{2}) or 4 (b-value of 800 s/mm

^{2});
sensitivity encoding acceleration factor, 4; and acquisition time,
approximately 3–4 min. Frequency-selective fat saturation was used to reduce
chemical shift artifacts. A free-hand region of interest on each equatorial
plane delineated the tumors. We evaluated
the pixel distribution histogram parameters of the ADC values derived
from b-values of 0 and 200 s/mm

^{2} (ADC

_{200}), 0 and 400 s/mm

^{2}
(ADC

_{400}), or 0 and 800 s/mm

^{2} (ADC

^{800}). The
histogram parameters (i.e., mean, coefficient of variation (CV), kurtosis, skew, and entropy) of the ADC values were compared between PACs,
NETs, and AIPs by using the Kruskal-Wallis test, followed by the Mann-Whitney U
test. Receiver operating characteristic (ROC) curve analyses for histogram
parameters of ADC

_{200}, ADC

_{400}, and ADC

_{800} were
generated to evaluate accuracy in diagnosing PACs, NETs, and AIPs.

### Results

ADC histogram results are
summarized in Table 1. Mean ADC

_{200}
was significantly higher in NETs than in PACs (P=0.005) and AIPs (P=0.022). Mean ADC

_{800}
was significantly lower in AIPs than in PACs (P=0.003) and NETs (P=0.014). Kurtosis
showed significantly lower in NETs than in PACs with all b-value combinations (P=0.038 for
ADC

_{200}, and P<0.001 for ADC

_{400} and ADC

_{800}),
and AIP with ADC

_{400} (P=0.008). Skew of ADC

_{400} and
ADC

_{800} showed significantly lower in NETs than in PACs (P<0.001
for ADC

_{400} and ADC

_{800}) and AIPs (P=0.006 for ADC

_{400}
and P=0.001 for ADC

_{800}). With
all b-value combinations, the entropy of ADC was significantly lower in NETs than in PACs (P=0.002 for ADC

_{200},
P=0.001 for ADC

_{400}, and P<0.001 for ADC

_{800}) and AIPs (P<0.001
for ADC

_{200} and ADC

_{400}, and P=0.005 for ADC

_{800}). For differentiating
PACs from AIPs, the area under the curve (AUC) for mean ADC

_{800} was 0.743. The entropy of ADC with every b-value combination showed the highest area under the ROC curve for differentiating
NETs from PACs (0.740 for ADC

_{200}, 0.736 for ADC

_{400}, and
0.758 for ADC

_{800}) and AIPs (0.781 for ADC

_{200}, 0.785 for
ADC

_{400}, and 0.765 for ADC

_{800}).

### Discussion

We used histogram
analysis to evaluate not only mean ADC, but also CV, skew, kurtosis, and entropy,
which reflect the distribution of ADC values. There have been no reports
assessing the usefulness of ADC histogram analysis for characterizing solid
pancreatic masses. In our study, a significantly lower entropy of ADC was found
in NETs. The entropy of ADC with every b-value combination showed the highest
area under the ROC curve for differentiating NETs from PACs and AIPs. Entropy describes
the variation in ADC histogram. Therefore, a lower entropy of ADC in NET could be expected as it is more homogenous compared with PAC
and AIP.

### Conclusion

ADC histogram analysis could be helpful for diagnosing solid pancreatic
masses, especially in NETs that have higher entropy characteristics with
promising potential for differentiating from PACs and AIPs.

### Acknowledgements

No acknowledgement found.### References

1. Ichikawa T,
Erturk SM, Motosugi U, et al. High-b value diffusion-weighted MRI for detecting
pancreatic adenocarcinoma: preliminary results. AJR Am J Roentgenol. 2007;
188:409-414.

2. Fukukura Y,
Takumi K, Kamimura K, et al. Pancreatic adenocarcinoma: variability of
diffusion-weighted MR imaging findings. Radiology. 2012; 263:732-740.

3. Lee SS, Byun JH,
Park BJ, et al. Quantitative analysis of diffusion-weighted magnetic resonance
imaging of the pancreas: usefulness in characterizing solid pancreatic masses.
J Magn Reson Imaging. 2008; 28:928-936.

4. Kang Y, Choi SH, Kim YJ, et al. Gliomas: Histogram
analysis of apparent diffusion coefficient maps with standard- or high-b-value
diffusion-weighted MR imaging: correlation with tumor grade. Radiology.
2011:261:882-890.