Synopsis
Using the intravoxel
incoherent motion (IVIM) effect, one can characterize the tumor
microenvironment in terms of vascularity and cellularity. Combined with
histogram analysis of these IVIM biomarkers, these metrics are compared to
clinical responders and nonresponders of neoadjuvent treatment (NAT) in breast
cancer patients. We examine the prognostic capabilities of these IVIM metrics
and find that (1) certain IVIM parameters significantly differentiate between
responders and nonresponders to NAT and (2) IVIM parameters change between pre-
and post-treatment MRI scans. This data shows IVIM MRI to be a potentially
powerful prognostic tool in breast cancer.Purpose
Diffusion weighted imaging
(DWI) characterizes cancerous tissue cellularity, particularly in breast cancer
1-4. Through DWI, biomarkers have been developed that are sensitive to
microvascular flow via the intravoxel incoherent motion (IVIM) effect
5. Using IVIM, first proposed by LeBihan, one
can quantify the tumor hypervascularity and hypercellularity, and these markers
have been shown in a range of breast cancer studies
6-11. Histogram analysis of the spatial distribution of IVIM parameters
12 can also provide additional information on
each parameter's distributions, giving values of skewness and kurtosis
13, 14, for potentially enhanced characterization of the cancer
microenvironment
15. While it has been shown that IVIM parameters can
distinguish between benign and malignant lesions
8, 9, 11, we hypothesize that a comprehensive evaluation of IVIM (including
metrics of average, maximum, minimum, kurtosis, and skewness) can potentially
have value by providing predictive biomarkers of treatment response. Here, we
examine breast cancer patients using pre- and post-treatment imaging and
compare the metrics from intravoxel incoherent motion (IVIM) histogram analysis
with clinical response results.
Methods
This IRB approved,
HIPAA-compliant retrospective study observed 31 breast cancer patients with 32
lesions (31 invasive ductal carcinoma and 1 invasive lobular carcinoma). All
patients underwent MRI scans during April 2011 to March 2013 followed by neoadjuvant
chemo treatment (‘pre-treatment scans’). Additionally, 6 patients underwent an
additional MRI scan 12-14 weeks after the initial scan and 1-2 cycles of treatment
(‘post-treatment scan’). All patients underwent a standard bilateral contrast-enhanced
(CE) MRI along with DWI in a 3T MRI scanner (Discovery MR750; GE Healthcare,
Waukesha, WI) with a 16 channel breast coil (Sentinelle Vanguard, Sentinelle
Medical, Toronto, Canada) and included fat-suppressed T2-weighted imaging, DW
and post-contrast T1-weighted imaging. The DWI protocol consisted of a single
shot spin echo EPI sequence (TR/TE = 4000/85.3 ms; 4 averages; FOV = 28 x 28 to
36 x 36 cm
2; slice thickness: 4–5 mm; acquired matrix: 128x128,
interpolated to 256x256; 19-35 slices; and 10 b values i.e. b = 0, 30, 60, 90,
120, 250, 400, 600, 800, 1000 s/mm
2). Analyses for IVIM average and
histogram (maximum, minimum, skewness, and kurtosis) metrics were derived from
custom data analysis (Igor Pro 6, Wavemetrics, Portland, OR) using a
biexponential model
5. A single operator drew ROIs around the outer
tumor border limiting IVIM analysis to the tumor region. Monoexponential
analysis was performed to generate ADC maps of the entire lesion. Segmented
biexponential IVIM analysis was performed to estimate D
t, f
p,
and D
p. An additional filter was generated to select voxels of
highly vascular tumor tissue (VTT) and exclude necrotic or normal tissue
regions
16, the percentage of which among all lesion
voxels was labeled as VTT%. Clinical data was collected including histology of
biopsy or surgical specimens as well as clinical response to treatment, defined
by tumor size reduction by RECIST criteria (2 patients were excluded as there was
no recorded response data). For analysis of pre-treatment IVIM-DWI parameters, we conducted
Mann-Whitney tests to test for differences between responders and nonresponder. P values were not adjusted for multiple comparisons.
Results
Of all 30 lesions studied, 5
were classified as nonresponders; among the 6 patients scanned twice, only 1
did not show treatment response. Average, kurtosis, and skewness of the D
p
parameters differentiate between responders and nonresponders to treatment
(Table 1). Other
IVIM metrics did not show any significant differences when compared to response
to treatment results. In addition, we observe that IVIM parameters change when
comparing between pre- and post-treatment MRI scans as seen in the parametric
maps for an individual patient with invasive ductal carcinoma (Figure 1) and in
all patients (Figure 2). ADC and D
t values generally increased post-treatment
in responders.
Discussion/Conclusion
Among baseline measurements,
D
p and VTT% were most prognostic, with high vascularity, slow and
heterogeneous pseudodiffusion offering poor prognosis; analogously, decreases
in D
p occurred in all dual-scanned responders. Our results showed
that baseline values of ADC or D
t were not predictive of response. Low
baseline ADC trended towards significance as a response predictor (p=0.074,
consistent with other studies
17). Also, early ADC and D
t increases
were largest in dual-scanned responders, indicating a possible decline in
cellularity, as previously correlated with response
18. While a larger study will help to confirm these trends on a
larger scale, this work suggest that both mean and heterogeneity metrics from
IVIM analysis have prognostic value in the setting of neoadjuvant breast cancer
therapy.
Acknowledgements
This research was partially
supported an NIH Core Grant P30 CA008748 and P41 EB017183.
References
1. S. Sinha, et al. J Magn Reson Imaging 2002. 15 (6): 693-704.
2. H. Rahbar, et al. Radiology 2012. 263 (2): 374-82.
3. S. C. Partridge, et al. Magn Reson Med 2011. 65 (6): 1759-67.
4. S. H. Kim, et al. J Magn Reson Imaging 2009. 30 (3): 615-20.
5. D. Le Bihan, et al. Radiology 1988. 168 (2): 497-505.
6. S. Suo, et al. Journal of Magnetic Resonance Imaging 2014.
7. G. Y. Cho, et al. Magn Reson Med 2015.
8. M. Iima, et al. Investigative radiology 2015. 50 (4): 205-11.
9. L. Bokacheva, et al. J Magn Reson Imaging 2014. 40 (4): 813-23.
10. E. E. Sigmund, et al. Magn Reson Med 2011. 65 (5): 1437-47.
11. C. Liu, et al. European journal of radiology 2013. 82 (12): e782-9.
12. B. Gaing, et al. Investigative radiology 2015. 50 (3): 144-52.
13. Y. S. Song, et al. Anticancer research 2014. 34 (5): 2135-44.
14. J. Wiederer, et al. Magn Reson Mater Phy 2014. 27 (2): 185-93.
15. G. Y. Cho, et al. Eur Radiol 2015.
16. G. Schwarz Ann. Stat. 1978. 6 (2): 461-4.
17. R. Prevos, et al. Eur Radiol 2012. 22 (12): 2607-16.
18. H. J. Shin, et al. NMR Biomed 2012. 25 (12): 1349-59.