| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1
Department of Radiology, Shinshu University School of Medicine, 3-1-1 Asahi,
Matsumoto 390-8621, Japan.
2
Department of Otolaryngology, Shinshu University School of Medicine, Matsumoto
390-8621, Japan.
Received September 22, 2000;
accepted after revision November 20, 2000.
Address correspondence to S. Takashima.
Abstract
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
SUBJECTS AND METHODS. Lesion-to-muscle magnetization transfer ratios obtained with a spoiled gradient-recalled acquisition in a steady state sequence with a 1-kHz off-resonance pulse were measured in 72 parotid masses (52 benign lesions, 20 malignant tumors). Various MR imaging findings and lesion-to-muscle magnetization transfer ratios were simultaneously assessed using a logistic model to determine the useful factors for predicting malignancy. We also studied the clinical usage of magnetization transfer ratios.
RESULTS. Of the MR imaging findings, poorly defined margins showed the highest accuracy, 81%, with 60% sensitivity and 88% specificity. Of the lesion-to-muscle magnetization transfer ratios, a ratio of greater than 0.71 was most accurate (85%), with 90% sensitivity and 83% specificity. All four recurrent tumors and 10 (91%) of 11 secondary tumors were correctly diagnosed using the magnetization transfer ratio analysis. The logistic model revealed that the margin characteristics (p = 0.084) and lesion-to-muscle magnetization transfer ratios (p < 0.001) were statistically significant predictors for malignancy. A combined criteria of poorly defined margins and a lesion-to-muscle magnetization transfer ratio of greater than 0.71 raised the accuracy to 86% and specificity to 96%, but the sensitivity decreased to 60%.
CONCLUSION. A combination of MR imaging findings and lesion-to-muscle magnetization transfer ratios was the most accurate predictor of malignancy.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Biologic tissues have three proton pools, including free water protons, water protons in proximity to macromolecules (protons of restricted motion), and protons within the macromolecules (immobile protons) [4, 5]. The protons that generate MR signal are the free water protons [4]. The amount of the free water protons is intimately related to the T1 and T2 values. Reduction in the free water protons decreases T2 signal intensity of a tissue or a lesion. On standard MR images, the restricted protons and immobile protons are not detected directly because their concentration is low and they have very short T2 values (<1 msec) to generate an MR signal compared with those of the free water protons [5, 6].
The immobile protons that are bonded to macromolecular proteins have a broader MR spectrum than the free water protons [6, 7]. The restricted protons and immobile protons play an important role in magnetization transfer. When a saturation pulse is applied to the immobile water protons, magnetization exchange between the immobile protons and the restricted protons occurs, and diffusion of the restricted protons into unsaturated free water protons takes place [5,6,7,8,9,10]. This results in a transfer of saturation to the mobile water protons, reducing the signal from the mobile water and causing an overall reduction in signal intensity. The degrees of the magnetization transfer effect are influenced by the macromolecular type, its concentration, the presence of paramagnetic substance, and the T1 of the water [8,9,10,11,12,13]. Therefore, research on the magnetization transfer effect gives insight into the molecular biology of the lesions.
Several investigators have reported promising results with magnetization transfer ratios of the lesions for distinguishing between benign and malignant tumors in the head and neck [14, 15]. However, differentiation among different types of malignant head and neck tumors was not possible [14]. In our preliminary study, we used 0.3-, 0.5-, and 1-kHz off-resonance magnetization transfer pulses from the water resonance in 53 head and neck lesions and compared the diagnostic capability of lesion magnetization transfer ratios with that of lesion-to-muscle magnetization transfer ratios at each off-resonance pulse for predicting malignancy [16]. Although the differences did not reach statistical significance, the lesion-to-muscle magnetization transfer ratios were better than the lesion magnetization transfer ratios at each magnetization transfer pulse. Of all the magnetization transfer ratios, the best diagnostic capability in the series [16] was attained with lesion-to-muscle magnetization transfer ratios at 1-kHz off-resonance pulse.
This prospective study was designed to clarify whether the lesion-to-muscle magnetization transfer ratios at 1-kHz off-resonance pulse can provide information in addition to the traditional MR imaging findings for predicting malignancy of parotid lesions, and to study the clinical usefulness of the magnetization transfer ratios.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
The masses were 43 benign tumors (15 mixed tumors, 26 Warthin's tumors, one basal cell adenoma, and one capillary hemangioma), eight inflammatory lesions, one branchial cleft cyst, and 20 malignant tumors (three squamous cell carcinomas, six non-Hodgkin's lymphomas, five adenocarcinomas, and one each of epithelial—myoepithelial carcinoma, adenoid cystic carcinoma, acinic cell adenocarcinoma, mucoepidermoid carcinoma, sebaceous carcinoma, and small cell carcinoma). Final diagnosis was obtained with surgical resection in 58 cases, surgical biopsy in one case of actinomycosis, aspiration biopsy in five cases of non-Hodgkin's lymphoma, and aspiration biopsy and radiologic follow-up in eight cases (seven inflammatory lesions, one capillary hemangioma). Among 20 cases of malignant tumors, nine cases had secondary parotid tumors with known primary tumors (six non-Hodgkin's lymphoma, two squamous cell carcinoma, one small cell carcinoma), four had recurrent tumors (three adenocarcinoma, one adenoid cystic carcinoma), and two (one adenocarcinoma, one squamous cell carcinoma) had metastatic parotid tumors with no identifiable primary tumors. Metastasis was diagnosed in the two cases of unknown primary tumors because tumors were identified pathologically in the intraparotid lymph nodes. In the case of small cell carcinoma, an initial presenting symptom was a parotid tumor, and a primary tumor was later found in the lung.
First, conventional spin-echo T1-weighted images (TR/TE, 800/13; number of excitations, 2) and fast spin-echo T2-weighted images (3000/85; 2 excitations) were obtained on a Signa 1.5-T MR unit (General Electric Medical Systems, Milwaukee, WI). Next, a single MR image with and without spoiled gradient-recalled acquisition in a steady state sequence (45/7; excitations, 2; flip angle, 20°) using a 1-kHz off-resonance radiofrequency pulse of a single-cycle sinc wave that was operated with a band-width of 110 Hz was obtained in the largest axial plane of each lesion. The duration of the magnetization transfer pulse was 18 msec, and the pulses were T1-weighted. Specific absorption rates were within safety guidelines in all the patients. After magnetization transfer images, fat-suppressed T1-weighted MR images (800/13; 2 excitations) were obtained immediately after the IV bolus injection of 0.2 mmol/kg of godopentatate dimeglumine. The images were obtained using a quadrature head coil with a matrix of 256 x 192, a field of view of 22 cm, and a section thickness of 5 mm.
For evaluation of magnetization transfer effects, the MR signal intensity of the lesions, skeletal muscle adjacent to the lesions (masseter, sternocleidomastoid, pterygoid, or digastric muscles), and background noise were measured in the images with and without magnetization transfer pulses using an electronic cursor to define the region of interest. The measurement was made by a single radiologist who was unaware of the final diagnosis. The mean area of the regions of interest was 81 ± 28 mm2 (standard deviation [SD]) for the lesions, 39 ± 12 mm22 for the muscle, and 54 ± 14 mm2 for the background noise. In general, regions of interest for the lesions were chosen to incorporate solid-appearing portions and to exclude vessels or obvious cystic areas. For an extensive cystic lesion, cystic portions were used for the measurements. From those measurements, a magnetization transfer ratio was calculated for the lesions (lesion magnetization transfer) and the muscle (muscle magnetization transfer) in each patient as follows: magnetization transfer = 1 - (Mpost/Mpre), where Mpre indicates signal intensity in the images before magnetization transfer and Mpost indicates signal intensity in the images after magnetization transfer. Signal intensity of the background noise was subtracted from that of the lesion and the muscle because we wanted to measure the signal intensity proper to the muscle and the lesion. Then a ratio of the lesion magnetization transfer to the muscle magnetization transfer (lesion-to-muscle magnetization transfer) was calculated and was used for analysis in this study.
To correlate the lesion-to-muscle magnetization transfer ratios with the amount of nuclear proteins, we measured the cell cycle fractions, including M (mitosis), G1 (gap 1), S (DNA synthesis), and G2 (gap 2) using the flow cytometric technique. Detailed methods are reported in the literature [17]. Flow cytometric DNA analyses were performed on 17 lesions (three inflammatory lesions, seven benign tumors, and seven malignant tumors). Surgically resected fresh materials or aspirate materials obtained by sonographically guided fine-needle biopsy were used for analyses. In this series, the percentage of (S + G2 + M) fraction of the cell cycle that reflects the degree to which the cells are proliferative was correlated with the lesion-to-muscle magnetization transfer ratios in those 17 lesions [18].
Without knowledge of the final diagnoses or the values of the lesion-to-muscle magnetization transfer ratios, two radiologists independently evaluated on a two-point scale the following MR imaging findings: T2-weighted signal intensity of the lesions relative to the parotid gland (hypointense or not [isointense or hyperintense]), invasion (presence, absence), and margin characteristics (poorly defined, well defined). Invasion was defined when an irregular interface between the lesions and the concerned adjacent structures with loss of the intervening fat plane was depicted on at least one of the three MR pulse sequences (unenhanced T1-weighted images, T2-weighted images, or contrast-enhanced T1-weighted images). Kappa values for the two radiologists were calculated, and final interpretations were determined by consensus of both radiologists.
We compared the accuracy of the lesion-to-muscle magnetization transfer ratios with that of one or a combination of the three MR imaging findings for diagnosing histology on the basis of the same two-point scale (benign or malignant). Next, all the parameters were simultaneously assessed using a stepwise logistic model to estimate the useful factors for predicting malignancy. Then we proposed the most accurate criteria for predicting malignancy.
To find the contributors to the lesion-to-muscle magnetization transfer ratios in benign lesions, we correlated the lesion-to-muscle magnetization transfer ratios with pathologic findings such as cellularity (epithelial cells plus nonepithelial cells) or the amount of collagen tissue in the lesions. Pathologic assessment was done by a pathologist who was unaware of the values of the lesion-to-muscle magnetization transfer ratios. Cellularity of a lesion was classified as low or high cellularity; low cellularity indicated that the area of the cellular component occupied less than 50% of the entire lesion; high cellularity meant that the area of the cellular component occupied equal to or greater than 50% of the entire lesion. The amount of collagenous proliferation was divided subjectively into no increase and increase. These analyses were possible in 20 benign lesions (one actinomycosis, one basal cell adenoma, seven mixed tumors, and 11 Warthin's tumors). The chisquare test, Student's t test, Mann-Whitney test, and Spearman's rank correlation were used for statistical analyses, and a p value of less than 0.05 was considered statistically significant.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
|
Kappa values of the two radiologists were 0.70 for margin characteristics, 0.66 for invasion, and 0.73 for T2-weighted signal intensity of the lesions. All the values were substantial [19]. Poorly defined margins were seen in six (three inflammatory lesions, three Warthin's tumors) (12%) of the 52 benign lesions and in 12 (60%) of the 20 malignant tumors. Invasion of surrounding structures such as skeletal muscle, the parapharyngeal space, the mandible, or the skull base was detected with MR imaging in three inflammatory lesions and seven (35%) of the 20 malignant tumors. Invasion was pathologically or surgically verified in six of the seven malignancies; however, no pathologic or surgical confirmation was obtained in the remaining four cases (three inflammatory lesions, one malignant lymphoma). Low signal intensity on T2-weighted images of the lesion was identified in 20 benign lesions (one actinomycosis, one basal cell adenoma, one mixed tumor, 17 Warthin's tumors) (38%) and nine malignant tumors (45%). Of the three MR imaging findings (T2-weighted signal intensity, invasion, margins), the prevalence of poorly defined margins and invasion in malignant tumors were statistically significantly greater than in benign lesions (p < 0.001 and p = 0.002, respectively).
Stepwise logistic models revealed that the margin characteristics (p = 0.084) and the lesion-to-muscle magnetization transfer ratios (p < 0.001) were the statistically significant factors for predicting malignancy. When we used a single feature on MR images, poorly defined margins had the highest accuracy—81%—with 60% sensitivity and 88% specificity (Table 1). False-positive results for this criterion occurred in three inflammatory lesions and three Warthin's tumors, and false-negative diagnoses occurred in eight malignant tumors (one acinic cell carcinoma, one epithelial—myoepithelial carcinoma, six secondary cancers). Accuracy was 78% for invasion and 58% for low signal intensity on T2-weighted images. Even when we used a combination of the three MR imaging findings, no combination of factors was better than poorly defined margins as a predictor of malignancy. When we used the lesion-to-muscle magnetization transfer ratios for predicting malignancy, a lesion-to-muscle magnetization transfer ratio of greater than 0.71 had the highest accuracy—85%—with 90% sensitivity and 83% specificity. With a threshold of 0.51, no false-negative diagnoses were included, but the specificity of the threshold was poor (55%). Using a threshold of 1.58, we found only one false-positive diagnosis (one capillary hemangioma) (98% specificity), but the sensitivity was low (15%).
|
Concordant diagnoses between the optimal MR criterion of poorly defined margins and the magnetization transfer criterion (a threshold of 0.71) were found in 55 (76%) of the 72 lesions (in 41 [79%] of the 52 benign lesions and in 14 [70%] of the 20 malignant tumors), whereas a discordant diagnosis was noted in 17 lesions (24%). Of the 17 lesions with a discordant result, a correct diagnosis was obtained with the MR criterion in seven benign lesions (five Warthin's tumors, one basal cell adenoma, one hemangioma) and with the magnetization transfer analysis in 10 lesions (four benign lesions [two inflammatory lesions, two Warthin's tumors], six malignant tumors [three malignant lymphomas, one squamous cell carcinoma, one adenocarcinoma, one epithelial—myoepithelial carcinoma]). Of the six malignant tumors with an MR-negative and magnetization transfer—positive diagnosis, five tumors were metastatic to the parotid gland and one tumor was a primary malignancy (epithelial—myoepithelial carcinoma).
Of combinations of the three MR imaging findings and the lesion-to-muscle magnetization transfer ratios, a combined criteria of the MR finding of poorly defined margins and the lesion-to-muscle magnetization transfer ratios (a threshold of 0.71) showed the highest accuracy—86%—and a specificity of 96%, but the sensitivity was 60% (Figs. 2A,2B,2C,2D,3A,3B,3C,4A,4B,5A,5B,5C). Two false-positive and eight false-negative results occurred for this optimal combined criteria. The false-positive cases included one highly cellular Warthin's tumor with slight collagenous proliferation and one actinomycosis associated with high cellular infiltration and a conspicuous increase in collagen pathologically (Fig. 5A,5B,5C). False-negative cases were the same for poorly defined margins and included six secondary malignancies (four malignant lymphomas, one squamous cell carcinoma, one adenocarcinoma) and two primary tumors (acinic cell adenocarcinoma, epithelial—myoepithelial carcinoma) (Fig. 4A,4B).
|
|
|
|
|
|
|
|
|
|
|
|
Four recurrent tumors were all correctly diagnosed both with the combined criteria and with the threshold of 0.71 for the lesion-to-muscle magnetization transfer ratios. Of the 11 patients with secondary cancer, 10 tumors (91%) exceeded the magnetization transfer ratio thresh-old, but only four tumors (36%) had poorly defined margins; one low-grade non-Hodgkin's lymphoma had well-defined margins and a lesion-to-muscle megnetization transfer ratio of 0.63. Thus, when our analyses were limited to the population of patients with secondary cancer, the highest sensitivity of 100% was obtained with a lesion-to-muscle magnetization transfer ratio of greater than 0.62, followed by the thresh-old of 0.71 (91% sensitivity). The combined criteria showed a low sensitivity of 36%.
Flow cytometric DNA analyses of 17 lesions showed that the mean (S + G2 + M) fraction in malignant tumors (15.8% [mean] ± 9.9% [SD]) was statistically significantly greater than that (3.0% ± 1.7%) in benign lesions (p = 0.001). The mean lesion-to-muscle magnetization transfer ratio in malignant tumors (1.05 ± 0.28) was statistically significantly greater than that (0.63 ± 0.38) in benign lesions (p = 0.006). Sperman's rank correlation revealed that the lesion-to-muscle magnetization transfer ratios had a significant positive correlation with the (S + G2 + M) fraction (p = 0.022).
In pathologic studies of 20 benign lesions, nine lesions (one actinomycosis, one basal cell adenoma, one mixed tumor, six Warthin's tumors) had high cellularity, and 11 (six mixed tumors, five Warthin's tumors) had low cellularity. A mean lesion-to-muscle magnetization transfer ratio of high-cellularity lesions (0.90 ± 0.34) was statistically significantly greater than that of low cellularity lesions (0.39 ± 0.13) (p = 0.008). No increase in collagen was found in 10 lesions (one basal cell adenoma, three mixed tumors, six Warthin's tumors), and increased collagen was identified in 10 lesions (one actinomycosis, four mixed tumors, five Warthin's tumors). Mean lesion-to-muscle magnetization transfer ratios were 0.62 ± 0.32 for lesions with no increase in collagen and 0.64 ± 0.27 for lesions with increased collagen. No significant difference in the mean lesion-to-muscle magnetization transfer ratios was noted between these two groups (p = 0.952).
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
We used T1-weighted gradient-echo sequences for obtaining magnetization transfer images in this study. In evaluating the magnetization transfer ratios, the T1 of a lesion should be considered in addition to the rate of magnetization transfer [13]. The moment the lesion is irradiated with magnetization transfer pulses, the T1 reduces to the apparent T1 [13]. The relationship between the magnetization transfer ratio and the T1 is given by the following equation: magnetization transfer ratio = ka x T1a, where ka indicates the rate of magnetization transfer and T1a represents the apparent T1 of the bulk water pool measured after irradiation [13]. The lesion-to-muscle magnetization transfer ratio is given by a new equation: lesion-to-muscle magnetization transfer ratio = kaL / kaM x T1aL / T1aM, where kaL and kaM are the rates of magnetization transfer for the lesion and the muscle, respectively, and T1aL and T1aM are the apparent T1 of the bulk water pool in the lesion and the muscle, respectively [13]. Previous studies revealed that the T1 of the muscle is not very different from that of carcinoma, and cancers and benign tumors have little variation in T1 [20,21,22]. Thus, we think the lesion-to-muscle magnetization transfer ratios reflected mainly kaL / kaM, regardless of the T1 of the lesion [16].
Yousem et al. [14] evaluated 54 head and neck tumors with lesion magnetization transfer ratios obtained at 2-kHz offset frequency from the water resonance on a 1.5-T MR unit. Although those researchers reported a statistically significant difference between lesion magnetization transfer ratios of malignancies and benign tumors, the series of Yousem et al. included only three parotid tumors. Markkola et al. [15] reported the diagnostic capability of lesion magnetization transfer ratios in 40 patients with head and neck tumors, including 14 salivary tumors, on a 0.1-T MR imager with a 4-kHz offset pulse. Those researchers mentioned that the highest accuracy of 80%, with a sensitivity of 94% and specificity of 65%, was achieved with a lesion magnetization transfer ratio of 0.32 or greater. When we used the lesion-to-muscle magnetization transfer ratios for predicting malignancy, a ratio of greater than 0.71 had the highest accuracy—85%—with 90% sensitivity and 83% specificity. The accuracy in our series was better than that reported by other authorities. These better results may have been caused partly by the use of the lesion-to-muscle magnetization transfer ratios instead of lesion magnetization transfer ratios, which may have corrected the T1 effect among the lesions and the variation among individuals and among anatomic locations in the [16].
Of the MR imaging findings, poorly defined margins had the highest accuracy—81%—with 60% sensitivity and 88% specificity in our study. False-positive results for this criterion occurred with inflammatory lesions and Warthin's tumors with poor definition, and false-negative diagnoses were mostly caused by a secondary cancer of the parotid gland. The addition of any other MR imaging findings to this criterion did not improved its diagnostic accuracy. Of a combination of all the predictors, an optimal criterion was the combination of poorly defined margins and a lesion-to-muscle magnetization transfer ratio of greater than 0.71. This criterion achieved the highest accuracy—86%—and improved the specificity from 88% to 96% because four benign lesions with poorly defined margins were correctly diagnosed using this criterion. However, the sensitivity remained the same, mainly because of the false-negative cases of secondary parotid cancer. We suggest that when patients have a known primary cancer other than in the parotid gland, a lesion-to-muscle magnetization transfer ratio may be helpful for screening the secondary parotid cancer because the threshold of 0.62 or 0.71 has a far better sensitivity than the combined criteria.
Yousem et al. [14] found no correlation between the lesion magnetization transfer ratios of head and neck squamous cell carcinomas and the degree of mitosis that was classified microscopically by a pathologist. However, our study showed that the cell proliferative fractions in malignant tumors were statistically significantly greater than those in benign lesions and that the lesion-to-muscle magnetization transfer ratios had a significant positive correlation with cell proliferative fractions. We used the (S + G2 + M) fraction of the cell cycle, because they represent the percentage of all the proliferative cells in a lesion [18]. Flow cytometry can rapidly, objectively, and quantitatively measure the DNA content of the nuclei of a lesion. Thus, our data support a positive correlation between the amount of macromolecular proteins in the nuclei of the lesions and the lesion-to-muscle magnetization transfer ratios.
The variation in the lesion-to-muscle magnetization transfer ratios was considerably large, which was conspicuous in benign lesions in our series. Some standard deviations of the magnetization transfer ratios were 27% for malignant tumors, 60% for all of the benign lesions, 20% for mixed tumors, and 40% for Warthin's tumors. Other authors documented a similar large variation. Markkola et al. [15] found standard deviations of 28% for malignant tumors and 52% for benign tumors in the head and neck. Of the 52 benign lesions, nine (17%) exceeded the threshold of a lesion-to-muscle magnetization transfer ratio of 0.71 in this study. Our pathologic analyses of benign lesions revealed that the high cellularity of the lesions and the resultant high macromolecular proteins had a significant positive effect on the magnetization transfer ratios, and that the magnetization transfer ratios were not relevant to the amount of collagen. Salivary gland tumors, both malignant and benign, are histologically heterogeneous, and most have both myoepithelial and epithelial components [23]. The tumors contain various kinds of intermediate filament proteins such as cytokeratins, neurofilaments, fibronectins, tonofilaments, and vimentin in various degrees [23,24,25,26]. Molecular weights of these proteins range from 40 to 150 kd [27]. Therefore, we believe that these intermediate filament proteins other than macromolecular proteins in the nuclei and cytoplasm may have influenced the magnetization transfer ratios of parotid lesions to some extent [9].
Most parotid lesions are diagnosed relatively noninvasively by fine-needle aspiration biopsy, with a high accuracy of 90-98% and at low cost [28,29,30]. MR imaging cannot replace the biopsy procedures for diagnosing parotid masses. Nonetheless, evaluation with MR imaging is useful in defining the full extent of the lesions, determining a surgical approach, indicating appropriate sites of biopsy, and predicting possible complications after surgery. We recommend an additional study with magnetization transfer for routine MR imaging examinations. Magnetization transfer ratio information may provide useful information before biopsy or surgery. We think that rebiopsy should be done in appropriate portions of the lesions that have a cytologically negative diagnosis or insufficient aspirate materials but have a lesion-to-muscle magnetization transfer ratio of greater than 0.51, because all malignant tumors in our series exceeded that threshold.
In conclusion, combined criteria of MR imaging findings and lesion-to-muscle magnetization transfer ratios were most accurate for predicting malignant parotid tumors, and magnetization transfer ratio analysis may be useful for diagnosing recurrent and secondary tumors.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
