2617

Diffusion-Weighted Imaging with Stretched-Exponential Model for Pancreatic Tumors
Takeshi Yoshikawa1, Yoshiharu Ohno2, Masao Yui3, Yoshimori Kassai3, Tatsuya Ohkubo3, Shinichiro Seki4, Katsusuke Kyotani5, and Yuji Kishida6
1Radiology, Kobe University Graduate School of Medicine, Kobe, Japan, 2Fujita Health University School of Medicine, Toyoake, Japan, 3Canon Medical Systems Corporation, Otawara, Japan, 4Hyogo Prefectural Tamba Medical Center, Tamba, Japan, 5Kobe University Hospital, Kobe, Japan, 6Konan Medical Center, Kobe, Japan

Synopsis

Stretched-exponential model enables diffusion analysis considering diffusion varieties in each voxel on abdominal DWI. Our study showed stretched-exponential model has a potential to increase diagnostic performance of pancreatic DWI. Alpha has a potential to be used for malignant lesion differentiation.

Backgrounds & Purpose

Backgrounds
  • To measure water diffusion in tissue more precisely compared to conventional ADC (mono-exponential model), several non-mono-exponential models have been proposed and evaluated.
  • One of them is stretched-exponential model. It can estimate heterogeneity of water diffusion as well as microstructures in the tissue and express them with only two parameters.
  1. Distributed diffusion coefficients (DDC)
  2. Water molecular diffusion heterogeneity index (Alpha, α)
α=1: pure homogenous intravoxel signal decay
α=0: high degree of multi-exponential signal decay
Equation : Sb/S0 = exp{-(b×DDC)α}　　　 cf. Sb/S0 = exp(-b×ADC)

Purpose
The purpose of this study was to assess DWI with stretched-exponential model in evaluation of pancreatic and peripancreatic diseases.

Methods and Materials

Patients& Lesion
  • 79 patients (44 men and 35 women, mean: 68.9years), who were suspected to have pancreatic and per-ipancreatic malignancy and underwent 3T-MRI, were retrospectively analyzed.
  • 61 malignant (pancreatic CA: 19, pancreatic metastasis: 4, IPMC: 3, MCN: 3, pNET: 3, bile duct CA: 12, papillary CA: 4, node metastasis: 13) and 40 benign (IPMN: 17, SCN: 4, LEC: 1, retention cyst: 7, WON: 1, pancreatitis: 8, papillary adenoma: 2) lesions were confirmed in the pancreas and peri-pancreatic region and evaluated.
Imaging Technique, Post-Processing, & ROI placement
  • All patients underwent MRI at a 3T scanner (Vantage Titan 3T; Canon Medical Systems, Otawara, Japan).
  • Source DWIs were obtained with SE-EPI sequence (TR/TE/FA = 3000-6000/66/90, b values: 0, 500, 1000, matrix: 96 ×128, thickness: 7mm, NEX: 2, scan time: 10-12min, PASTA+SPAIR, PI: 2, MPG: x, y, z) as one of routine sequences in our institution.
  • ADC, DDC, Alpha images were calculated by using mono-exponential and stretched-exponential models on a workstation (OleaSphere, Olea Medical).
  • Oval ROIs were placed in the pancreas and focal lesions and mean ADC, DDC, and Alpha were measured.
Quantitative Analysis
  • Mean values of malignant and benign lesions were compared for each parameter.
  • Correlations coefficients among the three parameters were assessed in the pancreas and focal lesions.
  • Lesion-to-pancreas contrasts (=(lesion-pancreas) / (lesion+pancreas)) were calculated and compared among the parameters.
  • Lesion characterization was compared using ROC among the parameters and their combinations.
  • Mean values in malignant lesions were compared for evaluation of lesion differentiation capability.

Results

  • Malignant lesion ADC and DDC were significantly lower than benign ones and no significant difference was found in Alpha (fig. 1).
  • DDCs showed similar trends to ADCs and were significantly correlated. Correlations of lesion ADCs and DDCs were significantly lower than pancreatic ones (fig. 2).
  • Malignant lesion contrast had non-significant trend toward higher on DDC than on ADC (fig. 3).
  • For lesion characterization, no significant difference was found between ADC and DDC, and Az had non-significant trend toward highest at DDC+Alpha (fig. 4).
  • For differentiation of malignant lesions, 6 significant differences were found in ADC, 4 in DDC, and 3 in Alpha. Significant differences were found only in Alpha between pancreatic cancer and bile duct cancers or node metastasis (p=0.008, p=0.003, fig. 5).
  • Representative cases are shown in figures 6 & 7.

Summary & Conclusion

Summary
  • Stretched-exponential model enables diffusion analysis considering diffusion varieties in each voxel and has a potential to increase diagnostic performance of liver DWI.
  • DDC can be used as an alternative to ADC.
  • Alpha has a potential to be used for malignant lesion differentiation.
Conclusion
  • Stretched-exponential model has a potential to increase diagnostic performance of pancreatic DWI.
  • Alpha has a potential to be used for malignant lesion differentiation.

Acknowledgements

No acknowledgement found.

References

Previous reports in the pancreas

  • Klaassen R, et al. Int J Radiat Oncol Biol Phys 2018.

Alpha: normal pancreas>PDA

  • Li J, et al. Magn Reson Imaging 2019.

Alpha: PNET>PDA

Figures

Figure 1. Mean ADC, DDC, Alpha of Lesions

Figure 2. Correlation Coefficients among the Parameters

Figure 3. Lesion Contrast on Images

Figure 4. Lesion Characterization -Malignant vs Benign-

Figure 5. Malignant Lesion Differentiation

Figure 6. A Case of Pancreatic Cancer

Figure 7. Malignant Lesion Differentiation on Alpha

Proc. Intl. Soc. Mag. Reson. Med. 28 (2020)
2617