Introduction

Dynamic contrast-enhanced (DCE) MRI is clinically used to assess response to neoadjuvant systemic treatment. In this study, we evaluated if changes in DCE MRI acquired after single ablative dose neoadjuvant partial breast irradiation (NA-PBI) could predict pathologic response.

Methods

MRI scans were obtained from 36 low-risk breast cancer patients participating in a single dose NA-PBI trial. Patients received single dose NA-PBI with 20Gy to the gross tumor volume (GTV) and 15Gy to the clinical target volume (GTV+2cm margin). NA-PBI was followed by breast conserving surgery (BCS) after an interval of 6 or 8 months. Pathologic response was then assessed according to EUSOMA criteria¹. Patients were divided in two groups: pathologic complete response (pCR; n=15) and no pCR (n=21).
Patients underwent 3.0T MRI scans (Ingenia, Philips, Best, The Netherlands) in prone position pre-radiotherapy (RT), 1 week post-RT, and 2, 4, 6, and, if applicable, 8 months post-RT for clinical follow-up before BCS. The protocol included two DCE series: a high-spatial/low-temporal resolution series (hereafter defined as ”slow”: 1 pre- and 5 post-contrast-injection images, acquired voxel size: 1.0x1.0x1.8mm³, 1.35 minute intervals post-contrast) and a low-spatial/high-temporal resolution series (“fast”: 17 images, 2.6x2.6x6.0mm³, 5 seconds interval). The latter was acquired between the pre- and the first post-contrast-injection image of the slow series (Fig. 1).

Radiologist assessment
Breast radiologists assessed radiologic response on MRI (e.g. DCE, DWI) according to clinical practice in neoadjuvant systemic treatment. The positive predictive value (PPV, probability that radiologic complete response predicts pCR) and negative predictive value (NPV, the probability that absence of radiologic complete response predicts residual disease) were calculated for the last MRI before BCS.

Quantitative assessment
The GTV was manually delineated on the pre-RT DCE MRI. Rigid registration using Elastix^2,3 was performed to rigidly transform the delineation of the GTV on the pre-RT MRI to the post-RT MRIs. Patients were included in the analysis when transformation of the pre-RT delineation was visually acceptable on at least 3 post-RT MRIs. Rigid registration was also used to correct for intra-scan motion in both the slow and fast series. GTV delineations were expanded with a 1-voxel margin to account for delineation and registration inaccuracies.
The relative enhancement (RE) for the slow and fast series was determined as $$$RE(t) = \frac{SI(t)-SI(0)}{SI(0)}*100\%$$$; where SI is the signal intensity per voxel, t=0 is the pre-contrast-injection image and t>0 are the post-contrast-injection images.
Within the GTV, we determined the percentage of voxels with >50% RE, the wash-in (90^th percentile RE within the GTV at the first post-contrast-injection slow image and at 1 minute post-contrast-injection in the fast series) and the area under the curve for 90^th percentile post-contrast curves (AUC_slow and AUC_fast) (Fig. 1). Additionally, the distribution of enhancement curve types describing the contrast wash-out⁴ (RE difference between first and last post-contrast-injection slow images) was evaluated for voxels with >50% RE. Wash-out was divided in three types indicating the probability of malignancy: type 1 (≥+10% RE) – low, type 2 (-10% to +10% RE) – intermediate, and type 3 (≤-10% RE) – high probability of malignancy. Median and interquartile range values were computed and compared across subjects.

Results

Radiologist assessment (n=36) showed a PPV of 67% and an NPV of 76% at the last MRI before BCS. The quantitative relative enhancement parameters (n=27) showed an initial increase followed by a decrease at later time points which can also be identified from visible inspection of the MRI scans over time (Fig. 2). The median percentage of voxels with >50% RE increased from 76% pre-RT to 81% at 1 week post-RT and decreased to 47% at 6 and 48% at 8 months post-RT (Fig. 3). Median wash-in decreased from 181% (slow) and 148% (fast) pre-RT to 103% and 75% at 6 and 110% and 108% at 8 months post-RT respectively (Fig. 4). The AUC response showed a similar pattern as the wash-in. A decrease from 13% pre-RT to 3% (at 6 and 8 months post-RT) in percentage of voxels with type 3 wash-out curve was observed (Fig. 5). Quantitative parameters did not differ between pathologic response groups.

Discussion

Changes in quantitative parameters describing wash-in and wash-out in DCE MRI were observed after single ablative dose NA-PBI. The slow and fast series showed similar patterns of response. All parameters indicated an increase in relative enhancement in the first week after RT, which can be explained by acute tissue response, i.e. edema, caused by RT. At later time points a decrease in enhancement and a slower wash-out pattern was observed, which indicates tumor regression. Although no differences were observed between pathologic response groups, 1 week post-RT response in the percentage of type 3 wash-out voxels showed a promising pattern for response prediction. Therefore, we would recommend to include an early response evaluation in future research on quantitative response assessment of NA-PBI.

Conclusion

Radiologist assessment had a low predictive value for pathologic response and a moderate value for predicting absence of pathologic response. Further quantitative evaluation of DCE MRI showed no differences between pathologic response groups. Therefore, radiologic response on MRI could not predict pathologic response after single dose NA-PBI in this low-risk patient group.

Acknowledgements

The authors would like to thank BHM van der Velden, KGA Gilhuijs, JP Wijnen, and ALHMW van Lier for their contributions on methods for image processing, image registration and quantitative analysis of MRI that were used for this study.