| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Service d'Imagerie Médicale, Hôpital Henri Mondor, 51 Avenue du
Maréchal de Lattre de Tassigny, Créteil 94010 Cedex,
France.
2 Department of Biostatistics, Centre Hospitalo-Universitaire Henri Mondor,
Créteil 94010, France.
3 Department of Pathology, Centre Hospitalo-Universitaire Henri Mondor,
Créteil 94010, France.
4 Department of Plastic Surgery, Centre Hospitalo-Universitaire Henri Mondor,
Créteil 94010, France.
Received March 31, 2003;
accepted after revision September 26, 2003.
Address correspondence to A. Luciani.
Abstract
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
SUBJECTS AND METHODS. We studied 16 women undergoing axillary lymph node dissection after combined bilateral breast MRI and high-resolution MRI of the axilla with a maximum in-plane resolution of 0.6 x 0.4 mm. MRI was performed using a standard double breast coil and a 15-cm round flexible surface coil adapted to the axilla. High-resolution axillary sequences, including inversion recovery T2- and spin-echo T1-weighted sequences, were performed before and after gadolinium chelates bolus injection. Axillary image analysis focused on nodal morphology including size, contour regularity, cortex and hilar appearance, signal intensity, and enhancement parameters. Axillary MRI findings were compared with the final pathogic results from axillary lymph node dissection in all patients. Patients were divided into groups according to the final pathologic axillary status. Differences in MRI lymph node features across the groups were tested using a t test for quantitative data and the chisquare test or Fisher's exact test for binary data.
RESULTS. The features of the axilla on high-resolution MRI that best discriminated between patients with positive pathologic findings and those with negative pathologic findings were the presence of nodes with irregular contours (p < 10–4), high signal intensity on T2 sequences (p < 10–3), marked gadolinium enhancement (p < 10–3), and round hila and abnormal cortexes (p < 0.05).
CONCLUSION. Breast tissue and axillary lymph nodes both can be analyzed on MRI in a one-step process using a bilateral breast coil combined with a surface coil. Morphologic features observed on high-resolution MRI of the axilla can improve the identification of metastatic nodes.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
The aims of this study were to develop an MRI protocol for the simultaneous imaging of both breasts and axilla using a standard bilateral breast coil combined with an axillary surface coil and to determine whether high-spatial-resolution MRI of the axilla can be used to identify features of axillary lymph nodes that allow differentiation of benign nodes from those with tumor involvement.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Clinical Assessment
Clinical breast cancer and axillary staging according to the TNM
classification system
[8–10]
was performed by a breast surgeon before surgery and imaging.
MRI Procedure
MRI was performed on a 1.5-T magnet with multichannel capability (Symphony,
Siemens) using a standard double breast coil combined with a flexible surface
coil. All patients were examined in the prone position with bilateral shoulder
extension. A 15-cm-diameter round flexible axillary surface coil
(Fig. 1) was precisely fitted
and tightly taped to the patient's axilla ipsilateral to the affected breast;
additional foam cushions were used when necessary. This yielded the closest
contact between the surface coil and the axilla and provided the best possible
comfort for the patient. The imaging protocol was composed of four sequences
for breast analysis and three high-resolution sequences for simultaneous
axillary assessment. These sequences included T2-weighted imaging and
T1-weighted imaging without fat saturation before and after gadolinium
chelates bolus injection (Table
1).
|
|
Image Analysis of the Axilla
All digital MR images of the axilla were assessed by two radiologists
without knowledge of the clinical assessment of the axilla.
Quantitative analysis.—In all patients, the diameters of all lymph nodes with a short axis greater than 3 mm were measured with calipers on digital images by one radiologist. Short and long axes were measured to calculate each node's longest axis–to–shortest axis ratio and to determine how many nodes had a short axis greater than 5 or 10 mm. In nodes in which a hilum was identified, the thickness of the cortex and the long and short axes of the hilum were measured to calculate the longest axis–to–shortest axis ratio of the hila.
Qualitative analysis.—The qualitative interpretation of all axillary MR images was performed by two radiologists in consensus without knowledge of the breast images. If the reviewers disagreed, a consensus interpretation took place with a third reviewer.
The following MRI lymph node features were analyzed: the number of nodes showing irregular contours was determined; the signal intensity on spin-echo T1-weighted and inversion recovery T2-weighted images and the visual enhancement of each node after gadolinium chelates injection were noted; and the morphology of the cortex in nodes in which a hilum could be identified was assessed. The morphology of the cortex was assessed according to recently reported criteria [5]: The number of nodes with a normal external thin C-shaped cortex and the number of nodes with an abnormal cortex (> 2 mm, irregular, or eccentric) were determined. T1 signal intensity of the hila was characterized relative to that of adjacent fat as either high (signal intensity identical to adjacent fat) or low (signal intensity lower than adjacent fat). T2 signal intensity of the hila was defined as high or low relative to muscle signal intensity. Lymph nodes were classified as showing either intense or moderate enhancement or exhibiting absent visible enhancement by comparing unenhanced and contrast-enhanced images.
Image Analysis of the Breasts
The greatest diameter of the largest breast tumor was measured independent
of axillary evaluations on delayed subtracted T1-weighted images after
contrast injection. Measurements were performed by a reviewer who was both
unaware of the axillary findings and not in charge of the axillary
assessment.
Histopathologic Analysis
The triangular en bloc resected axillary lymph node specimens from
dissection were oriented on surgical removal using three surgical threads to
identify both the superior margin and the two inferior margins. Each specimen
was then sent immediately to the pathology department. The orientation of all
specimens was cautiously respected by pathologists to allow the most precise
correlation possible with MRI findings. All lymph nodes retrieved were
included in toto, and 1-mm-thick slices were obtained, numbered, stained with
H and E, and analyzed. Pathologic results of all mastectomy or lumpectomy
specimens were used to determine the histologic type and grade of the primary
breast malignancy.
Statistical Analysis
Patients were divided into two groups according to the final pathologic
axillary status. Differences in MRI lymph node features across the groups were
tested using a t test for quantitative data and the chi-square test
or Fisher's exact test for binary data. After an adjustment for multiple
testing, p values of less than 0.0045 were considered statistically
significant. The false-positive and false-negative results of a combination of
node features encountered possibly more frequently in patients with metastatic
nodes were also assessed. All statistical analyses were performed using SAS
software (version 8.0, SAS Institute).
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
One hundred eight lymph nodes were identified on MRI (lymph nodes per patient: mean, 6.8; range, 3–11; SD, 2.4). The features of the lymph nodes on high-resolution MRI are summarized in Table 2. Fifty nodes were observed in the patients with positive findings at pathology, and 58 nodes were seen in the patients with negative findings at pathology. Normal-appearing lymph nodes with a thin C-shaped cortex and absent visual enhancement on contrast-enhanced T1-weighted images were observed in both groups of patients (Fig. 2A, 2B, 2C). The mean long and short axes of the lymph nodes were significantly greater in patients with positive pathologic findings. However, the number of nodes showing a short axis greater than 5 mm did not differ significantly for the two patient groups. The percentage of nodes with irregular contours (p < 10–4), central high signal intensity on inversion recovery T2-weighted images (p < 10–3), or intense enhancement (p < 10–3) was significantly higher in patients with positive pathologic findings than in those with negative pathologic findings. Extracapsular invasion was identified in five patients, all of whom had at least one node with irregular contours on high-resolution MRI (Fig. 3A, 3B, 3C).
|
|
|
|
|
|
|
The assessment of the cortex and hila in nodes with identified hila is summarized in Table 3. A hilum was identified in 84 nodes and was encountered significantly more frequently in patients with negative pathologic findings (p < 0.01). The longest axis–to–shortest axis ratios for the hila were significantly smaller in patients with positive pathologic findings (p < 0.05) (Fig. 4A, 4B, 4C, 4D). Mean cortex thickness was higher in patients with positive pathologic findings, but differences were not significant when compared with patients with negative pathologic findings. Thickened, eccentric, or irregular abnormal cortexes were identified significantly more frequently in patients with positive pathologic findings (p < 0.05), although these features were also seen in patients with negative pathologic findings (Fig. 5).
|
|
|
|
|
|
Table 4 shows the number of patients presenting with at least one node showing a combination of features possibly encountered more frequently in patients with positive pathologic findings; these features were short axis longer than 5 mm, irregular contours, hyperintensity on inversion recovery T2-weighted images, intense enhancement, longest axis–to–shortest axis ratio for hila less than 1.5, and abnormal cortex. A combination of four or more of these features for a single node gave the best prediction of tumor involvement with 12% false-positive and 13% false-negative rates (Fisher's exact test, p < 0.01). Of the three patients with negative pathologic findings and at least one node showing only one MRI sign described, two had a node showing a short axis greater than 5 mm with no other specific feature.
|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Patient comfort is a critical issue in breast MRI. The use of a surface coil combined with the standard bilateral coil is beneficial: First, the main support of the patient's body weight is provided by the well-padded bilateral breast coil. Second, the flexibility of the surface coil allows precise fitting to each patient's breasts, thus accounting for a wide variation in body type. In addition, shoulder extension is possible, thus providing adequate exposure of the axilla with the closest possible contact between the skin and the coil, which is not achieved with a patient positioned with the arm along the body.
Visualization of the axilla is achievable using a standard receive-only bilateral breast coil encompassing both breasts and axilla [12] but at the expense of nonuniform images and partial coverage of the axilla [13]. Further reducing the slice thickness is usually associated with diminished signal-to-noise ratio [13]. The surface coil we selected, one that encompasses the entire axilla, provided an in-plane resolution of 0.6 x 0.4 mm with 2-mm slice thickness owing to the 307 x 512 matrix.
Our technique enabled identification and analysis of 108 nodes—accounting for 63% of all lymph nodes retrieved at axillary lymph node dissection—without altering the simultaneous bilateral breast examination. Although a comparison is difficult, the number of nodes we detected is higher than what is usually reported in the literature [14]. Nodes not seen on MRI in our study could correspond to nodes showing large central hila with high signal intensity (similar to axillary fat) and with normal thin C-shaped cortexes, which are difficult to distinguish from subcutaneous fat streaks. Also, two adjacent lymph nodes could have been mistaken for a single node.
Our high-resolution MRI protocol allows the identification of morphologic changes affecting metastatic axillary nodes, some of which, including node enlargement or spiculated contours, have been already reported extensively on mammography or sonography [15]. However, mammography remains insensitive to small lymph nodes [16]. Moreover, an inherent advantage of cross-sectional techniques such as CT or MRI is that the visualization of small nodes is not altered even in deeply located nodes, which is not the case with sonography [4]. Furthermore, high-resolution MRI study of the axilla provides morphologic information not only regarding the size and architecture of a node but also regarding its content, including the hila and cortex. Hence, increasing spatial resolution using a surface coil allowed the high-resolution MRI assessment of features of lymph nodes that have been reported on high-resolution CT in vitro, which could help differentiate benign and malignant nodes [5].
MRI evaluation of the axilla has previously been performed without combined breast study and without T2 signal analysis and enhancement evaluation [14]. Our study confirms that size is a valuable diagnostic criterion but suggests that features observed on high-resolution MRI—including irregular contours, high signal intensity on inversion recovery T2-weighted images, intense enhancement, round hila, and abnormal cortexes—are associated with malignancy. Such features were not previously reported on MRI, especially in small nodes, because of insufficient spatial resolution.
Our study has limitations. First, only a small number of patients were included, mostly because of limited access to the MRI unit in our department. Patients older than 75 years were primarily excluded because management strategies of breast cancer in elderly women usually require less frequent and less aggressive therapies [17]. Exclusion of these patients could have resulted in the selection of patients with higher tolerance for prolonged examination in the prone position. We deliberately selected patients with only T1 or T2 tumors according to TNM staging; this criterion resulted in the exclusion of patients referred to our institution for breast MRI evaluation before chemotherapy, for example. However, including patients with large primary breast lesions could have biased our results by increasing the prevalence of lymph node involvement, especially in patients with large palpable nodes.
Another limitation of our study is that an exact correlation between MRI-identified nodes and nodes retrieved at dissection and analyzed at pathology could not be systematically guaranteed, especially for nodes smaller than 5 mm. Surgical threads allowed the identification of all superior and inferior margins, but pathologists found few structures within the en bloc resected axillary specimens that could provide a physical reference for the retrieved nodes, which could have provided a higher confidence in the radiopathologic correlation. As a result, we believe that providing sensitivity and specificity figures on a node-to-node basis is not reliable and would require ex vivo axillary imaging.
Preoperative axillary evaluation requires not only a high sensitivity in order to limit the rate of false-negative examinations, which could lead to misdiagnosed metastases, but also a high specificity in order to limit the rate of unnecessary invasive procedures. Our study results show that features of axillary lymph nodes could possibly allow discrimination of patients with benign nodes from those with malignant nodes. We found that each of the following nodal features was present in less than 25% of patients with positive pathologic findings: a short axis greater than 5 mm, longest axis–to–shortest axis ratio less than 1.5, irregular contours, hyperintensity on inversion recovery T2-weighted images, intense enhancement, hila longest axis–to–shortest axis ratio less than 1.5, and abnormal cortex. However, our results suggest that a combination of four or more of these features in a single node allows optimized 12% false-positive and 13% false-negative rates on a per-patient basis. These figures nevertheless confirm that, to date, MRI examinations cannot replace axillary lymph node dissection.
Sentinel lymph node dissection has been proposed as an alternative to axillary lymph node dissection in breast cancer patients to reduce the number of patients with negative pathologic findings from axillary dissection [18–20]. The rationale for sentinel lymph node dissection is twofold: first, early metastases occur in the first, or sentinel node, to receive lymphatic drainage from the tumor; second, skip metastases during lymph node extension are rare. Thus, identification and assessment of this sentinel node by means of limited surgical dissection could be sufficient for breast carcinoma staging [18]. However, sentinel lymph node dissection remains an invasive procedure, and although mild, adverse effects have been reported [21]. In contrast to axillary lymph node dissection, sentinel lymph node dissection alone does not provide extensive control of localized disease.
Several authors have recently reported the impact of the use of ultrasmall superparamagnetic iron oxide (USPIO) agents in the MRI assessment of lymph node status and identification of sentinel nodes [22, 23]. We believe that a complementary study using high-resolution MRI with T1 sequences (for the assessment of morphology) and T2 sequences after local or peritumoral USPIO injection (for the assessment of tumor involvement) could further improve the diagnostic value of high-resolution MRI of the axilla. Furthermore, high-resolution MRI of the axilla after the injection of a USPIO agent could be helpful in assessing the distribution of the iron particles within the nodes.
Hyperplastic or reactive nodes were depicted at pathology in three of the 16 axillary lymph node specimens, all of which were from patients with positive pathologic findings. Just as we regard sensitivity and specificity figures on a node-to-node basis, we do not believe that a precise correlation between MRI findings (e.g., hyperintensity on T2-weighted images) and pathologic results (e.g., necrosis, inflammation, massive cellular infiltration) can be performed without ex vivo imaging.
Our study included only women with breast cancer. We acknowledge that additional studies are mandatory to assess whether MRI can be used to distinguish metastatic from inflammatory nodes or from nodes involved by other neoplasms. USPIO agents have been shown to be of important clinical value in the setting of head and neck cancer with regional node involvement [24, 25]. Additional studies of USPIO-enhanced high-resolution MRI of the axilla could further improve the distinction of hyperplastic and metastatic nodes. Furthermore, micrometastases, which have been increasingly reported after sentinel lymph node dissection procedures [26], were not assessed at pathology in this study because extensive axillary lymph node dissection was performed. Additional studies comparing high-resolution MRI evaluation and intensive histopathologic workup of sentinel nodes is mandatory to evaluate the performance of MRI in the depiction of micrometastases.
The mean short axis of identified nodes was 5.1 mm in our study. In such small nodes, the identification of a hilum can be difficult. For instance, the mean short axis of nodes with identified hila was 6.2 mm (range, 4–8 mm). This could explain the significantly higher incidence of nodes with identified hila among the nodes with positive pathologic findings, as nodes with a short axis above 5 mm are more frequently encountered in metastatic nodes. However, our results show that the presence of irregular shaped cortexes can be identified on high-resolution MRI, and could be predictive of malignancy. These features could be related to the preferential metastatic involvement of the periphery of lymph nodes, close to afferent lymphatics [27].
Kvistad et al. [14] and Murray et al. [7] have shown that the quantitative assessment of the enhancement of lymph nodes could be predictive of malignancy. However, in these two studies, simultaneous MRI breast study was not performed. We believe that high-resolution MRI of the axilla could provide additional morphologic information before dynamic contrast injection. The simultaneous dynamic study of the breasts and ipsilateral axilla can be performed using our protocol of combined surface coil and three distinct sagittal acquisition boxes: one positioned on each breast and the third box encompassing the axilla. However, the dynamic follow-up of lymph node enhancement can be time-consuming when numerous nodes (mean of approximately seven nodes per patient in our study) are identified on MRI and must be individually selected and studied. Dynamic study of the axilla would also probably require fat-saturation techniques to be combined with the T1-weighted sequences to favor the detection of subtle enhancement. In the protocol we used fat-suppression was not performed. Optimal selective fat saturation requires a homogeneous magnetic field. Furthermore, appropriate localized shimming on the axillary region is difficult to obtain because the axilla is eccentrically located in the MRI magnet. Additional studies using fat-suppressed dynamic contrast-enhanced sequences are mandatory.
Several authors have recently reported that the use of USPIO agents on MRI could yield up to 83% sensitivity and 97% specificity in the assessment of lymph node status [6, 28]. However, using USPIO agents requires an additional MRI examination for the bilateral breast assessment. The combination of the morphologic features suggestive of malignancy indicated by our study results and the absence of USPIO uptake could provide higher confidence in the diagnosis of metastatic nodes.
Our preliminary results warrant larger studies using high-resolution MRI of the axilla. The multichannel capability of MRI devices allows the simultaneous high-spatial-resolution study of the ipsilateral axilla together with conventional breast MRI examinations. Radiologists should be aware of this possibility when performing preoperative breast MRI assessment. Furthermore, high-spatial-resolution axillary MRI allows the morphologic analysis of lymph nodes, which could improve the identification of metastatic nodes.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
