| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Département d'Imagerie, Institut Curie, 26, rue d'Ulm, Paris 75005,
France.
2 Department d'biostatistique, Institut Curie, Paris, France.
Received January 2, 2004;
accepted after revision September 15, 2004.
Address correspondence to C. El Khoury.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. Nineteen patients with breast T2 or T3 tumors were enrolled in a previous study for tumor response evaluation during chemotherapy based on dynamic contrast-enhanced MRI. We retrospectively used the dynamic acquisition data to produce parametric images representing the washout pattern. Two radiologists unaware of the final pathologic results measured the volume of pixels exhibiting washout within the tumor before chemotherapy (volume 1), after two courses of chemotherapy (volume 2), and before surgery after four courses of chemotherapy (volume 3). The interobserver variability and intraobserver variability were calculated to evaluate the reproducibility of our method with the Pearson's correlation coefficient and the concordance correlation coefficient. We correlated the washout changes by means of a Student's t test and noted the histopathologic final outcome.
RESULTS. A washout pattern was present in all patients on the initial MR study. The quantification method of the washout changes was reproducible with good interobserver agreement (r = 0.85, p < 10–5) and an excellent intraobserver agreement (r = 0.94, p < 10–5). A significant decrease of the washout volume was observed after two courses of chemotherapy (p = 0.004), whereas no significant modification was observed between two and four courses of chemotherapy (p = 0.52).
CONCLUSION. Quantification of the washout variation in breast tumor based on the use of parametric images is feasible and reproducible. It may add information to the evaluation of tumor response to preoperative therapy.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Monitoring response to neoadjuvant chemotherapy is based on physical examination, mammography, and sonography of the breast. MRI has been shown to be the most reliable technique for evaluating residual tumor size, especially in cases of multifocal disease [4–6]. Analysis of dynamic contrast enhancement improves the accuracy of MRI and can be used to produce parametric images that provide information about vascular permeability and functionally distinct areas of the lesion [7, 8].
Early modifications of the time–intensity curves of tumors after neoadjuvant chemotherapy have been reported in the literature such as flattening or "plateau" of washout time–intensity curves, suggesting an unequivocal pathologic effectiveness for neoadjuvant chemotherapy [9, 10]. Washout is usually identified in a region of interest (ROI) placed within the most enhancing part of the tumor. However, time–intensity curve analysis does not allow quantitative evaluation, and the choice and placement of the ROI are observer-dependent.
Among the dynamic patterns of enhancement usually described, the washout has a high level of specificity for malignant lesions (85–100%) and a sensitivity of 50–85% [11–13]. We believe it is interesting to be able to detect easily this sign without the use of ROIs and to be able to assess its evolution under medical treatment with a quantitative or semiquantitative method.
The objective of the study was to assess the feasibility of MR washout quantification in breast tumors. We used this method in 19 tumors treated by primary chemotherapy with postoperative pathologic correlation.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The imaging protocol was accepted by patients seen at the medical consultation with a T2 or T3 operable breast tumor, in whom chemotherapy was proposed before local treatment. All patients had pathologically proven breast cancer and were treated by neoadjuvant chemotherapy (four courses of 5-fluorouracil, cyclophosphamide, and epirubicin) followed by local treatment. All patients underwent surgery—either mastectomy or lumpectomy. Radiation therapy was performed after surgery in all patients.
Twenty-two patients completed the three MR examinations: MR 1 was performed before surgery, with a mean interval of 11 days after core biopsy (SD, 11 days); MR 2, after the second course of chemotherapy, with a mean interval of 14 days (SD, 6 days); and MR 3, after the last course of chemotherapy, with a mean interval of 10 days (SD, 10 days). Three patients were excluded from the study because of motion artifacts.
MRI was performed at our institution on a 1.5-T imager (Signa, GE Healthcare) with a dedicated breast receiver coil. An IV catheter was placed in an antecubital vein before imaging. A bolus of gadopentetate dime-glumine (Magnevist, Schering), 0.1 mmol/kg of body weight, was administered by hand over 5–10 sec and was followed by a 20-mL saline flush.
The imaging protocol consisted of five data acquisition series. Two localization sequences (TR/TEeff, 45.5/1.6) were performed. The next sequence was a transverse T1 3D fast spoiled gradient-recalled echo sequence (TR/TE, 11/4.2; time of acquisition, 85 sec; number of excitations, 1; section thickness, 3 mm; and matrix size, 512 x 192 with a field of view of 40 x 40 cm). A sagittal fast spoiled gradient-recalled echo sequence (11/4.2; time of acquisition, 45 sec; number of excitations, 1; section thickness, 4 mm; matrix size, 512 x 192; field of view, 24 x 19.2 cm) was performed before the administration of the contrast agent and was repeated 10 times after the beginning of the bolus injection starting at the following points: 45 sec, 1 min 30 sec, 2 min 15 sec,... 7 min 30 sec. Eighteen sections were acquired and covered the fibroglandular tissue of the breast. In one case, 20 sections were needed to cover the fibroglandular tissue of the breast, with an acquisition time of 48 sec. A transverse delayed postcontrast fast spoiled gradient-recalled echo sequence (11/4.2; time of acquisition, 85 sec) was also performed.
For the purposes of the present study, we used only the dynamic sagittal acquisition series.
Image Analysis
The sagittal fast spoiled gradient-recalled echo sequence was analyzed with
the Functool algorithm on the Advantage Workstation (GE Healthcare) by two
senior radiologists experienced in breast MRI and not informed about the final
pathologic results. To test the intraobserver variability, one of the two
radiologists performed the same process twice at an interval of 1 year without
reviewing his first results; Figure
1A,
1B,
1C,
1D,
1E,
1F,
1G summarizes the various steps
of volume calculation. The algorithm was used to create the washout
(W) images resulting from a pixel-by-pixel subtraction of the final
dynamic series (Sfinal) from the maximum signal intensity
in each pixel (Smax):
 | (1) |
 | (2) |
|
|
|
|
|
|
|
The early enhancement images were used by radiologists to locate and measure the greatest axis of the enhancing tumor. We visually checked the shape of the signal-to-time curve using a 1-pixel ROI that the operator placed on the parametric images within the washout signal—that is, rapid enhancement followed by a decrease of the signal (Figs. 1D and 1E). Washout images were analyzed using the 3D algorithm (Volume Analysis, GE Healthcare) of the same workstation (Advantage): In cases of multifocal disease, the operator isolated from the rest of the breast the main tumor or the unique tumor in cases of a single lesion by outlining the region with the washout pattern (Fig. 1F). The radiologist selected a minimum cutoff value for washout on the first MR examination: This value was decreased until the region covered by the cutoff corresponded visually to the washout zone. The algorithm was then asked to exclude pixels below the cutoff value (Fig. 1F). Vessels were visible on the parametric washout images because the enhancement within vessels increases rapidly and decreases after a maximum point. In some cases, large vessels close to the tumor were delineated and excluded manually from the volume before using the cutoff. The same cutoff value was used for the three examinations for each subject.
Cutoff values and tumor washout volumes were recorded by each radiologist.
Three volumes were obtained for each patient: volume 1, before chemotherapy;
volume 2, during chemotherapy; and volume 3, before surgery
(Fig. 1G). We calculated the
percentage of residual washout on the second MR examination with respect to
the first MR examination for each patient:
 | (3) |
We assessed the intraobserver variability by comparing the cutoff values selected and the percentage residual washout on the second MR examination [(volume 2 / volume 1) x 100] processed by the same radiologist 1 year later (reading 1a and reading 1b 1 year later) without reviewing the first reading and the interobserver variability by comparing the results from the two radiologists by means of Pearson's correlation coefficient. The agreement between observers was also calculated by means of the concordance correlation coefficient to evaluate reproducibility [13].
Comparisons of volume 1 with volume 2 and volume 2 with volume 3 for the three readings (1a, 1b, and 2) of the two radiologists were performed using a Student's t test.
Surgical and Pathologic Analysis
The type of surgery was recorded, and the pathologic response was
classified from 1 to 4 according to the percentage of viable and nonnecrotic
tumor cells in the global tumor mass, as previously described
[14,
15]. Class 1 was defined as
complete or nearly complete response (no residual active cells, or < 5% of
viable tumor cells with no mitosis). Class 2 was considered to be complete
response with persistence of ductal carcinoma in situ. Class 3 was defined as
major response (< 20% of viable tumor cells), and class 4, as no response
(> 20% of viable tumor cells). The microscopic size of the residual tumor
and the type of pathologic response are given in
Table 1.
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Washout Quantification
Values of volume 1, volume 2, and volume 3 calculated by each observer are
summarized in Table 2. From the
washout image analysis, the following results were observed: The mean volumes
measured by the first observer (observer 1a) were volume 1, 2.6 mL (range,
0.12–5.9); volume 2, 1.1 mL (range, 0–6.2); and volume 3, 1.1 mL
(range, 0–7.3). The mean values measured by the second observer were
volume 1, 3.7 mL (range, 0.05–18); volume 2, 1.1 mL (range,
0–7.3); and volume 3, 0.9 mL (range, 0–5.8).
|
Residual washouts after two courses of chemotherapy—(volume 2 / volume 1) x 100—as determined by the two observers are shown in Table 3. Cutoff values used by each observer are listed in Table 3. The mean cutoff value used by the first observer (1a) was 33 ± 10 (SD) relative units. The second observer used a mean cutoff value of 33 ± 8 (SD) relative units.
|
Different washout patterns were observed in all cases. The variability of the washout pattern within tumors was supported by the fact that various cutoff values were used and various volumes were processed at the initial MR studies. The first observer (observer 1a) reviewed a group of four patients with a cutoff value that was lower than the mean value he used (< 33 relative units) and a volume less than the mean processed value for the other 19 patients (< 2.6 mL). In a second group of six patients, a cutoff value of more than 33 relative units and a washout volume of less than 2.6 mL were used. A third group of four patients had a cutoff value of less than 33 relative units and a washout volume of more than 2.6 mL, and a fourth group of five patients had a cutoff value of more than 33 relative units and a washout volume of more than 2.6 mL.
Good intraobserver agreement (r = 0.94) and interobserver agreement (r = 0.75) were noted; Pearson's correlation coefficient and concordance correlations results are detailed in Table 4. A significant difference was seen between volume 1 and volume 2 using the Student's t test. (p = 0.004 for observer 1a, p = 0.0085 for observer 1b, and p = 0.009 for observer 2), but no significant difference was observed between volume 2 and volume 3 (p = 0.99 for observer 1a, p = 0.91 for observer 1b, and p = 0.52 for observer 2).
|
Surgical and Pathologic Results
Nine patients underwent lumpectomy, and 10 patients underwent mastectomy.
Pathologic examination revealed 11 ductal invasive carcinomas, four lobular
invasive carcinomas, two undifferentiated carcinomas, and one mixed ductal and
lobular carcinoma; in one patient, no residual carcinoma was found. In this
patient, the initial core biopsy revealed poorly differentiated carcinoma
(grade 3 Elston Ellis).
Lymph node involvement was observed in 11 patients. Sixteen patients were classified pathologically as nonresponders (class 4), two patients (cases 8 and 12) had a major response (class 3), and one patient (case 17) had a complete or nearly complete response (class 1). The mean estimated size of the residual tumor was 23 ± 10 (SD) mm (range, 0–40 mm).
Washout Quantification and Pathologic Results
The mean residual washout in the 16 nonresponders (class 4) was 48%
± 45% (SD), with values ranging from 1% to 124%. In two cases of major
response (class 3), the residual washout was 0% and 2%. No residual washout
was present in the complete responder case (class 1). No statistical
correlation was possible with the washout quantification because of the single
case of complete response.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
With the method used in our study, we found a washout pattern in all patients with, in some cases, a few pixels exhibiting intense washout, which corresponds to a high cutoff level and a low volume, and, in other cases, a few pixels with moderate to slight washout, which corresponds to a low cutoff value and volume. In such cases, the weak washout sign may not be seen on visual assessment of the time–intensity curve.
The exact significance of washout is unclear. Its presence appears to be related to various vascular conditions, such as arteriovenous shunting or high vessel density. In a correlative study of MR images obtained after contrast enhancement with immunohistopathologic findings, the time–intensity curve type A (with washout) was significantly correlated with the highest microvessel density grade [11]. Because tumor signal is related to the presence of gadolinium chelate molecules in the intravascular and extracellular spaces of the tumor, the presence of vessels is responsible for rapid influx and efflux of contrast medium into the tumor.
The size, density, and permeability of capillaries are directly related to tumor angiogenesis. Immunohistochemical staining of vessels has shown that there are a greater number of capillaries in invasive ductal carcinoma in the periphery than in the center of the tumor, which may explain the frequent peripheral washout pattern [18]. Moreover, the presence of a pressure gradient between the tumor and the vascular space and between the periphery of the tumor and its center may also explain the peripheral sign [19, 20].
The parametric view rapidly shows all regions of the breast with a decrease of enhancement after a maximum point. This technique can be integrated into a routine dynamic study with no need to perform other sequences, such as T1 mapping, or to administer a second dose of contrast agent. It can be performed simply on a workstation without having to place an ROI in the most enhancing part of the tumor, as usually described in the literature, and is therefore less operator-dependent. Indeed, the variability due to the choice of ROI size and its placement within the tumor has already been reported as a major source of interobserver variations [21].
In this study, a large part of the variability was inherent to the subjective choice of the cutoff value for washout quantification, which explains differences in volume values between observers. However, there was good interobserver correlation (r = 0.59, p = 0.007) and concordance correlation (r = 0.58, p = 0.004) between observers and a better intraobserver correlation (r = 0.7, p = 0.0008) and concordance correlation (r = 0.67, p = 0.0005). Moreover, when we studied the washout variation from MR 1 to MR 2 [(volume 2 / volume 1) x 100], we found an excellent intraobserver correlation (r = 0.94, p < 10–5) and concordance (r = 0.94, p < 10–6) and an excellent interobserver correlation (r = 0.85, p < 10–6) and concordance (r = 0.78, p < 10–6).
The main condition determining success of this technique is to avoid movement artifacts, which made volume processing impossible in three cases. In the other 19 cases, washout volume was successfully calculated on the basis of the main tumor by using the entire dynamic breast imaging. We noted a statistically significant decrease of washout after two courses of chemotherapy, which is in agreement with the study by Makris et al. [22], who reported a decrease of microvessel density on immunohistochemical quantification after neoadjuvant chemotherapy for breast cancer. Such findings could be secondary to tumor regression or due to a direct effect on angiogenesis.
In this study, a complete pathologic response was observed in one patient (class 1), which is in agreement with the previously reported rate of pathologic response to neoadjuvant chemotherapy [3]. In that patient, we observed a complete regression of the washout after two cycles of chemotherapy, whereas in the nonresponders, the mean washout residue [(volume 2 / volume 1 x 100)] was 48%. However, there is an important SD in the washout residue with a range of 1–124% in the nonresponders group. Because of the unique case of complete response, no statistical relation could be established with the washout variation after two cycles of chemotherapy.
Further studies are needed to assess the additional value of washout quantification in the prediction of pathologic response to neoadjuvant chemotherapy. We believe this type of quantification method could be applied to breast tumors under endocrine neoadjuvant treatment, as Marson et al. [23] have reported the antiangiogenic effect of tamoxifen. The antiangiogenic effect of tamoxifen could be secondary to tumor regression or to a direct effect on angiogenesis, because animal experiments have shown the antiangiogenic effect of tamoxifen on tumors with negative estrogen receptors [24]. Changes in angiogenesis can be a marker of tumor response to chemotherapy, and the washout quantification may constitute a surrogate for microvessel counting with immunohistochemical techniques that requires repeat biopsy. It seemed of interest to be able to evaluate its variations and, to our knowledge, no such quantification has been previously reported.
In conclusion, washout imaging can be performed rapidly on a routine dynamic enhanced MRI acquisition. We noted a significant decrease of washout volume after two cycles of chemotherapy. Quantification of the washout changes is easily reproducible, and its additional value in monitoring breast cancer response to neoadjuvant chemotherapy needs to be evaluated in larger series.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  L. Moy and C. L. Mercado Interpretation and Clinical Applications of Breast MRI: Self-Assessment Module Am. J. Roentgenol., December 1, 2008; 191(6_Supplement): S60 - S67. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. E. Loo, H. J. Teertstra, S. Rodenhuis, M. J. van de Vijver, J. Hannemann, S. H. Muller, M.-J. V. Peeters, and K. G. A. Gilhuijs Dynamic Contrast-Enhanced MRI for Prediction of Breast Cancer Response to Neoadjuvant Chemotherapy: Initial Results Am. J. Roentgenol., November 1, 2008; 191(5): 1331 - 1338. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. K. Kuhl Current Status of Breast MR Imaging * Part 2. Clinical Applications Radiology, September 1, 2007; 244(3): 672 - 691. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
