Quantitative DCE-MRI Assessment of Breast Cancer Therapeutic Response : How Long Is the Acquisition Time Necessary ?

Purpose: Quantitative dynamic contrast-enhanced (DCE) MRI has been shown to be an important imaging biomarker for early prediction of breast cancer response to neoadjuvant chemotherapy (NACT) (1,2). The commonly adopted DCE-MRI acquisition time (Tacq) is 8 min or longer. However, breast cancer patients undergoing NACT are generally quite ill with compromised immune system. Consequently, the relatively long Tacq causes significant amount of patient discomfort, often resulting in motion artifacts in the acquired images and low-quality data. Therefore, without sacrificing the ability of DCE-MRI for discriminating responders from nonresponders, shorter Tacq should be used to reduce patient discomfort and obtain good-quality data. A recent simulation study (3) shows that a DCE-MRI Tacq value of 4.6 min is usually sufficient to derive stable K and ve pharmacokinetic parameters from DCE-MRI signal intensity time-course data of most curve features. In this study, by analyzing DCE-MRI data with shortened Tacq generated from the acquired data with a long Tacq, we sought to determine, for the purpose of evaluating breast cancer response to NACT, if the DCE-MRI Tacq can be significantly reduced. Methods: Eleven women with locally advanced breast cancer underwent NACT before definitive surgeries. They consented to research DCE-MRI studies that were performed at time-point zero (TP0) before NACT, at TP1 after first NACT cycle, at TP2 – midpoint of NACT (usually after three NACT cycles), and at TP3 after NACT completion. Axial bilateral DCE-MRI images with fat-saturation and full breast coverage were acquired using a 3T Siemens scanner, with a 3D gradient echo-based TWIST sequence (4), 10 flip angle, 2.9/6.2 ms TE/TR, a parallel imaging acceleration factor of two, 30-34 cm FOV, 320x320 matrix size, and 1.4 mm slice thickness. The total DCE-MRI Tacq was ~ 10 min for 32-34 image volume sets with 18-20 s temporal resolution. Gadolinium contrast agent (ProHance) IV injection (0.1 mmol/kg at 2 mL/s) was carried out following acquisitions of two baseline image volumes. Tumor ROIs were drawn by experienced radiologists. The ROI and pixel-by-pixel (within the ROI) DCE timecourse data were then subjected to both the Standard (Tofts) Model (SM) (5) and the Shutter-Speed Model (SSM) (6) analyses as previously described (7) to extract pharmacokinetic parameters. The SSM accounts for the finite intercompartmental water exchange kinetics (6). The whole tumor mean parameter values were calculated as the weighted (by ROI pixel number) averages of the ROI values from each of the image slices covering the entire tumor. Pathology analyses of the surgical specimens and comparisons with preNACT biopsy specimens revealed that three patients were pathologic complete responders (pCRs) while the other eight were non-pCRs. We found that the tumor mean K (SM and SSM) and kep (SM and SSM) values after the first NACT cycle (at TP1), as well as their % changes (TP1 relative to TP0), were able to discriminate the 3 pCRs from the 8 non-pCRs. To determine whether DCE-MRI acquisitions with shorter Tacq are still effective in predicting therapy response, data points from the actually acquired DCE-MRI time-course data of one pCR and one non-pCR were incrementally dropped from the far end of the time course before undergoing the same SM and SSM analyses, simulating a Tacq range of ~ 10 min to ~ 3 min. The raw data from the two patients were selected for these analyses because their tumor mean K and kep values derived from the full Tacq data were close to the means of the pCR and non-pCR groups, respectively, and thus were representative. Results: Fig. 1a shows the column graphs of TP1 tumor mean K (SM and SSM) and kep (SM and SSM) values of the pCR (blue) and non-pCR (red) at varying Tacq, while Fig. 1b shows the corresponding % changes (TP1 relative to TP0) of these metrics. The vertical scales are kept the same in Fig. 1a for each parameter between the two models. Some K (SSM) and kep (SM) values are out of scale at small Tacq. The TP1 K (~0.04 min) and kep (~0.06 min) parameter values of the pCR were nearly identical between the SM and SSM analysis: characteristic for tumor with diminished angiogenesis (7). These values remain stable for Tacq as short as 3-4 min. For the non-pCR, the K (SSM) value (~0.08 min) derived from the full Tacq acquisition was substantially greater than the K (SM) value (~0.05 min). This is expected for a tumor with presumably still elevated angiogenesis (7) after one NACT cycle. As a result, K (SSM) provides better separation of the pCR from the non-pCR. The non-pCR K and kep parameter values remain stable for Tacq as short as 5.5-6 min before being artificially increased with further Tacq shortening. The latter actually “artificially” enhances discrimination of the two responses albeit with increased uncertainty in data fitting. Fig. 2 shows the pCR and non-pCR TP1 tumor K (SSM) color maps obtained at varying Tacq, demonstrating clear difference between the two breast tumors even at Tacq as short as ~ 2.4 min. The % changes in K (SM and SSM) and kep (SM and SSM) remain robust measures of therapy response in the entire simulated Tacq range (Fig. 1b), with no less than 20% difference between the pCR and the non-pCR. Discussion and Conclusion: A recent review (8) indicates that the K parameter is the dominant DCE-MRI biomarker that is useful for assessment of cancer therapeutic response. The results from this study suggest that for typical responder and non-responder breast tumor K or kep values after one NACT cycle, a DCEMRI Tacq of 5-6 min is sufficient to achieve the goal of early prediction of therapeutic response when either parameter is used as the discriminatory biomarker. Therefore, any breast DCE-MRI acquisition time beyond the 6-min mark is probably unnecessary for the purpose of assessing therapy response. A shorter Tacq can speed-up DCE-MRI protocol, reduce patient discomfort, and minimize the possibility of motion. Grant Support: NIH: U01-CA154602. References: 1. Marinovich et al. The Breast 2012;21:669-677. 2. Ah-See et al. Clin cancer Res 2008;14:6580-9. 3. Wang et al. Proc ISMRM 2010;18:4811. 4. Tudorica et al. Magn Reson Imaging 2012;30:1257-67. 5. Tofts et al. JMRI 1999;10:223-32. 6. Li et al. Magn Reson Med 2005;54:1351-9. 7. Huang et al. Radiology 2011;261: 394-403. 8. O'Connor et al. Nat Rev Clin Oncol 2012;9:167-77.