0947

Quantitative non-Gaussian diffusion and IVIM MRI: Correlation between synthetic parameters and breast cancer biomarkers

Mami Iima^1,2, Masako Kataoka¹, Shotaro Kanao¹, Natsuko Onishi¹, Makiko Kawai¹, Akane Ohashi¹, Rena Sakaguchi¹, Ayami Ohno Kishimoto¹, Masakazu Toi³, and Kaori Togashi¹

¹Department of Diagnostic Imaging and Nuclear Medicine, Graduate School of Medicine, Kyoto University, Kyoto, Japan, ²Hakubi Center for Advanced Research, Kyoto University, Kyoto, Japan, ³Department of Breast Surgery, Graduate School of Medicine, Kyoto University, Kyoto, Japan

Synopsis

The association of IVIM/non-Gaussian diffusion MRI parameters with biological feature or subtypes in breast cancer was evaluated. For 144 malignant lesions, IVIM (fIVIM,D*) and non-Gaussian diffusion (ADCo,K) parameters were estimated from DWI series with 16 b values (0-2500sec/mm²), as well as syntheticADC (sADC) (b=200,1500sec/mm²) and ADC (b=0,800sec/mm²). sADC and K values were significantly different between ER,PgR,andHer2 status (p<0.05,0.01,0.05 for sADC and p<0.05,0.05,0.05 for K). There was a significant difference of ADC values between PgRandHer2 status (p<0.05,0.01). No significant difference of IVIM was found. ADCo, sADC, and ADC showed the statistical significance in differentiating subtypes of breast cancer (p<0.05,<0.01,<0.01).

Introduction

Diffusion MRI and apparent diffusion coefficient (ADC) widely used in breast MRI (1) are based on monoexponential model. Non-Gaussian diffusion parameters are relatively new DWI-derived parameters providing important information on tissue microstructure beyond ADC (2,3). However, their accurate estimation requires time-consuming fitting of the DWI signal with multiple b values using biophysical models (e.g.; kurtosis model) with long acquisition time. To reduce acquisition and processing times, we have recently introduced a “synthetic ADC” (sADC) derived from the signal acquired at a few “key b values”, which intrinsically include both non-Gaussian diffusion and IVIM effects, while maximizing information on tissue structure, without actually estimating diffusion parameters (4). We evaluated if sADC as well as IVIM and quantitative non-Gaussian diffusion MRI parameters can be used to differentiate biological features or subtypes in breast cancer.

Materials and Methods

405 patients were prospectively enrolled and 144 malignant lesions were included in this study. Breast MRI was performed using a 3-T system (Trio, B17; Siemens Healthcare) equipped with a dedicated 16-channel breast array coil. The following DWI (WIP) images were obtained: single shot EPI with 16 b values of 0-2500 sec/mm²; repetition time/echo time, 4,600/86 ms; field of view, 160×300 mm²; matrix, 80×166; slice thickness, 3.0 mm; and acquisition time, 3min55s; The algorithm was implemented in Matlab (Mathwork, Natick, MA) and comprised the following steps:

1/The signal acquired with b>200 s/mm² was fitted using the kurtosis diffusion model to estimate ADCo and K:

S/So = {exp[-bADCo + K(bADCo)²/6]+NCF}^1/2 [1]

where S0 is the theoretical signal acquired at b=0, ADCo the virtual ADC which would be obtained when b approaches 0, K the kurtosis parameter and NCF (noise correction factor) a parameter which characterizes the “intrinsic” non-Gaussian noise contribution within the images (3).

2/Then, the fitted diffusion signal component was subtracted from the corrected raw signal acquired with b<200s/mm² and the remaining signal was fitted using the IVIM model to get estimates of the (T1,T2-weighted) flowing blood fraction, fIVIM, and the pseudodiffusion, D* (3).

Additionally a synthetic ADC encompassing both Gaussian and non-Gaussian diffusion effects (4), sADC200-1500, was defined using only 2 b values as:

sADC200-1500 = ln [Sn(b200)/Sn(b1500)]/1300 [2]

ADC0-800 was also defined using standard monoexponential ADC model. IVIM and non-Gaussian diffusion parameters in the status of histological biomarkers (ER, PgR, Her2, Ki-67) were compared using Mann-Whitney test. IVIM and non-Gaussian diffusion parameters in different breast cancer subtypes (Her2-positive(H+), Luminal A(LA), Luminal B and Her2-negative (LBH-), Luminal B and Her2- positive (LBH+), and Triple Negative (TN)) were compared using Kruskal–Wallis test with a post-hoc analysis.

Results

Patient characteristics are summarized in Fig.1and 2. Comparison of non-Gaussian diffusion MRI parameters with the breast cancer biomarker and subtype status are shown in Fig 3 and 4. ER and PgR positive tumors showed significantly lower sADC value (p < 0.05 and 0.01) and higher K value (p< 0.05 and 0.05). PgR positive tumors had significantly lower ADC value (p < 0.05).Significantly higher sADC, ADC and lower K values were noted in Her2 positive status (p = 0.013, < 0.01,< 0.05). There was no significant difference of IVIM parameters depending on the histological biomarker status. ADCo (p<0.05), sADC (p<0.01), and ADC (p<0.01) showed the statistical significance in differentiating subtypes of breast cancer as shown in Fig 4. Representative non-Gaussian diffusion MRI and IVIM parameters maps are shown in Fig 5. The difference of ADCo, K and sADC values between Luminal B (Her2 negative) and Her2 postive breast cancers are remarakable.

Discussion

Only sADC and K, not ADC, showed the significant difference with ER status in our study. This indicates that non-Gaussian DWI parameters might provide additional information on the histological biomarkers. Lower ADC value in ER positive tumors was found as in the literature (5-7). Her2 positive tumors showed higher sADC or ADC values than Her2 negative, as in the previous studies (9, 10). There was significant difference of K values between positive and negative ER, PgR and Her2 status, and this has not been reported elsewhere. Luminal B and Her2 negative cancer showed the smaller ADCo, sADC and ADC values than luminal A, which was in agreement with the literature (11).The combination of these Gaussian and Non-Gaussian diffusion parameters, or synthetic ADC, might be a surrogate biomarker of the receptors expression as well as subtypes of breast cancer.

Conclusion

Non-Gaussian diffusion parameters, especially sADC and Kurtosis, showed a good correlation with biological factors and subtypes of breast cancer. There parameters might provide useful information in identification of breast cancer biological factors and molecular subtypes without the need for contrast agents.

Acknowledgements

This work was supported by Hakubi Project of Kyoto University and MEXT KAKENHI Grant No. 15K19786.

The authors would like to thank Dr. Thorsten Feiweier from Siemens Healthcare for providing WIP sequence.

References

1. Partridge S, DeMartini W, Kurland B, et al. Differential diagnosis of mammographically and clinically occult breast lesions on diffusion-weighted MRI. Journal of Magnetic Resonance Imaging. 2010;31:562-570

2. Le Bihan D et al. Diffusion Magnetic Resonance Imaging: What Water Tells Us about Biological Tissues. PLoS Biol. 2015 Jul; 13(7): e1002203

3. Iima M et al. Quantitative Non-Gaussian Diffusion and Intravoxel Incoherent Motion Magnetic Resonance Imaging: Differentiation of Malignant and Benign Breast Lesions. Investigative Radiology 2015:50:205-11

4. Iima M et al. Clinical Intravoxel Incoherent Motion and Diffusion MR Imaging: Past, Present and Future. Radiology 2016;278:1

5. Jeh SK, Kim SH, Kim HS, et al. Correlation of the apparent diffusion coefficient value and dynamic magnetic resonance imaging findings with prognostic factors in invasive ductal carcinoma. J Magn Reson Imaging. 2011;33(1):102-9

6. Kamitani T, Matsuo Y, Yabuuchi H, et al. Correlations between Apparent Diffusion Coefficient Values and Prognostic Factors of Breast Cancer. Magnetic Resonance in Medical Sciences. 2013;12(3):193-9

7. Choi SY, Chang YW, Park HJ, Kim HJ, Hong SS, Seo DY. Correlation of the apparent diffusion coefficiency values on diffusion-weighted imaging with prognostic factors for breast cancer. The British Journal of Radiology. 2012;85(1016):e474-e9

8. Black R, Prescott R, Bers K, Hawkins A, Stewart H, Forrest P. Tumour cellularity, oestrogen receptors and prognosis in breast cancer. Clinical oncology. 1983;9(4):311-8

9. Kim EJ, Kim SH, Park GE, et al. Histogram analysis of apparent diffusion coefficient at 3.0 t: Correlation with prognostic factors and subtypes of invasive ductal carcinoma. Journal of Magnetic Resonance Imaging. 2015;42(6):1666-78

10. Martincich L, Deantoni V, Bertotto I, et al. Correlations between diffusion-weighted imaging and breast cancer biomarkers. European radiology. 2012;22(7):1519-28

11. Kato F et al. Differences in morphological features and minimum apparent diffusion coefficient values among breast cancer subtypes using 3-tesla MRI. Eur J Radiol. 2016 Jan;85:96-102

Figures

Figure 1: Patient Characteristics according to Histological Type

Figure 2: Patient Characteristics according to Histological Biomarker Types and Subtypes

Figure 3: Distribution of non-Gaussian Diffusion MR Parameters according to the Histological Biomarker Types and Subtypes

Figure 4: Non-Gaussian diffusion MR parameters of each subtype. Connecting lines indicate the statistical significance (*: p<0.05, **: p<0.01) Kruskal-Wallis test showed statistical significance in ADCo (p<0.05), sADC200_1500 (p<0.01), and ADC0_800 (p<0.01). ADCo, sADC200_1500 and ADC0_800 values in LBH- were significantly lower than those in LA or H+ breast tumors. In addition, ADCo and sADC200_1500 values in LBH- were significantly lower than those in TN breast tumors. ADCo and ADC0_800 values in LBH- were significantly lower than those in LBH+ breast tumors.

Figure5: fat-suppressed T2WI, DWI, and representative non-Gaussian diffusion MRI and IVIM parameters maps. Higher ADCo and sADC200_1500 values as well as lower K values are observed in Her2 positive compared with Luminal B (Her2 negative) breast cancer.

Proc. Intl. Soc. Mag. Reson. Med. 25 (2017)

0947