Objective To investigate the MRI and pathological features of different molecular subtypes of breast cancer.Methods The data of 202 patients who underwent primary breast cancer resection were retrospectively reviewed.All of the patients had MRI preoperatively.The molecular subtypes of breast cancer defined by immunohistochemistry were classified as basal-like,luminal and HER-2 overexpression.Morphology (including mass or non-mass like enhancement,shape and margin of masses,unifocal or multifocal masses) and enhancement characteristics on MRI,histologic types and grades of tumors were analyzed with Chi-square test,exact test,Fisher exact test,Kruskal-Wallis H test,and Wilcoxon test.Results Among the 202 patients,34 were basal-like,144 were luminal and 24 were HER-2 overexpression.The number of mass cases in each subtype was 29,133 and 19 respectively,making no significant difference (x2 =4.136,P =0.126).As for the shape of basal-like lesions,8 were round,19 were lobular and 2 were irregular,while this distribution was 23,58,52 in luminal subtype and 1,11,7 in HER-2 overexpression subtype (x2 =13.391,P ＜ 0.05).The margin was also strikingly different among three groups (smooth,spiculated,irregular):20,5,4 respectively in basal-like,27,53,53 respectively in luminal,and 4,7,8 respectively in HER-2 overexpression (x2 =28.515,P ＜ 0.01).52.6％ (10/19) of HER-2 overexpression cases were multifocal,while only 6.9％ (2/29) of luminal and 8.0％ (24/133) of basal-like ones were multifocal (x2 =16.140,P ＜ 0.01).Characteristics in dynamic contrast-enhanced MRI were statistically different,with homogeneous,heterogeneous,and rim enhancement 0,13,16 respectively in basal-like cases,28,93,11 respectively in luminal cases and 2,11,6 respectively in HER-2 overexpression cases (P ＜ 0.01).However,the difference for enhancement curve did not reach significance (P =0.457).Histologic types were significantly different among molecular subtypes (P ＜ 0.01).Luminal breast cancer consisted of various histologic types,32.6％ (47/144) of which were mixed type.The majority of the other two molecular subtypes were invasive ductal carcinoma.Furthermore,invasive ductal carcinomas with different molecular subtypes showed different histologic grades (Hc =30.014,P ＜ 0.01).Basal-like breast cancer was more likely associated with a higher grade of malignancy.Conclusions Different molecular subtypes of breast showed distinct MRI features and pathologic characteristics.MRI might be a useful tool for preoperative prediction of molecular subtypes of breast cancer.

Objective To investigate the value of MR imaging methods for the quantification of fat content in a customized lipid phantom at 3.0 T.Methods Eleven homogeneous fat-water phantoms (50 ml)with fat volume percentages from 0 to 100％ were constructed with reference to Bernard's methods.Fat tractions of the lipid phantom were acquired using water selective saturation (WS),fat selective saturation (FS),in-and out-of-phase imaging (IOP),iterative decomposition of water and fat with echo asymmetry and least-squares estimation (IDEAL) gradient echo imaging and IDEAL Quant imaging methods on a 3.0 T MRI system.For statistical comparisons,paired-sample t test,Pearson correlation test,and Bland-Ahman maps were applied.Results Evaluated fat fractions acquired by WS,IDEAL Gradient echo imaging and IDEAL Quant were (49.6±28.8)％,(46.0±28.4)％,(51.0±32.0)％,the result has no significant difference with the true fat contents(t values were-0.186,-2.218,2.713;P values were 0.856,0.051,0.055).Evaluated fat fractions acquired by FS,corrected algorithm and IOP were (64.2±26.7)％,(58.9±31.9)％ and (45.3±32.3)％,these three kinds of methods have significant difference with the true fat contents (t values were 5.168,4.273,-6.441;P＜0.01).All the chemical shift imaging methods correlated with the true phantom model fat fractions,r values were 0.977(FS),0.978 (corrected algorithm),0.982 (WS),0.99 8(IOP),0.993 (IDEAL Gradient echo imaging),0.999 (IDEAL Quant) (all P＜0.01).Each method's 95％ confidence interval of the mean difference acquired by Bland-Altman map was WS (-14.7％ to 13.8％),FS (-3.6％ to 32.0％),corrected algorithm (-4.6％ to 22.5％),IOP(-9.4％ to 0.0％),IDEAL gradient echo imaging (-15.9％ to 7.8％),IDEAL Quant(-2.0％ to 4.0％).IDEAL Quant had the best correlation and confidence with the true fat fraction.Conclusions Chemical shift imaging methods (IOP,IDEAL Gradient echo imaging,IDEAL Quant) can acquire more accurate fat quantification results than chemical saturation imaging methods (FS,Corrected algorithm,WS) in a customized lipid phantom at 3.0 T.IDEAL Quant can acquire the best fat quantification result compared with the other imaging methods.

Objective To investigate the imaging features of Ewing's sarcoma and primitive neuroectodermal tumor (ES and PNET) in sacrum.Methods Imaging data were retrospectively collected and reviewed in 17 patients with ES and PNET confirmed by pathology.X-ray was performed in all patients,CT scans in 13 patients,and MR scans in 10 patients.Results All lesions were detected solitary and lateral.X-ray and CT features were lytic bone destruction in 12 cases,2 of them with bone sclerosis; one patient showed diffuse sclerosis.Soft tissue mass in pelvic cavity can be seen in 11 cases,9 of them with adjacent large blood vessels invasive and 4 patients with sacroiliac joints invasive.On post-contrast images,mild enhancement was found in 8 cases,and 5 patients with marked enhancement.MR images showed intermediate signal intensity on T1WI in 8 cases and heterogeneous moderate to high signal intensity on T2WI/fat suppression in 10 cases.DWI showed high signal intensity in 4 cases.Unlike CT scan,contrastenhanced MR images showed significant enhancement in 7 cases.Conclusions Ewing's sarcoma and primitive neuroectodermal tumor (ES and PNET) in sacrum have some specific imaging features.Imaging examination could provide reliable evidence for clinical diagnosis.