| Part I Clinical feasibility of CDT-VIBE for breast DCE-MRIObjectivesTo evaluate the image quality and morphology characterization of a novel T1-weighted sequence combining Controlled Aliasing in Parallel Imaging Results in Higher Acceleration (CAIPIRINHA) imaging with Dixon fat-separation and Time-Resolved Imaging with Interleaved Stochastic Trajectories (TWIST) view-sharing (CAIPIRINHA-DIXON-TWIST(CDT)-VIBE,[CDT-VIBE]) for breast dynamic contrast enhanced (DCE)-MRI.Materials and methods1. Phantom imaging:Scanning was performed on a3.0T MR system (MAGNETOM Skyra, Siemens Healthcare Sector, Erlangen, Germany) with a4-channel phased-array breast coil. Two spherical phantoms (filled with dimethyl silicone) were placed inside a4-channel breast coil with holders. CDT-VIBE sequences with an acceleration factor (AF) of4and TWIST A/B both of20%(temporal resolution12s,16phases in total) and conventional gradient-recalled echo (GRE) sequences with GRAPPA AF of2(without fat suppression, temporal resolution68s,15phases in total) were used for phantom imaging.2. Breast DCE-MRI:All examinations were performed on a3.0T MR system (MAGNETOM Skyra, Siemens Healthcare Sector, Erlangen, Germany) with a4-channel phased-array breast coil for58women with breast lesions. Routine MRI scanning was consisted of axial T1WI, axial fat-suppression T2WI and sagital fat-suppression T2WI. Dynamic CDT-VIBE sequences with the same parameters used for phantom imaging were repeated for40measurements and followed by a single phase of conventional T1weighted GRE sequence with the same parameters used for phantom imaging but with fat-suppression. FOV covered bilateral breasts.3. Data processing and measurements:All data processing and measurements were performed directly on a workstation (Siemens Healthcare) without extra post-processing. Signal intensity and image noise of right and left phantoms on15phases of CDT-VIBE fat-only images (began with the second phase) and on conventional GRE images were measured for the calculation of SNR. The SNRs of10central slices were calculated. On patients’last phase of CDT-VIBE water-only images and conventional GRE images, circular region of interest (ROI) were set manually on the most enhanced tumor region and non-enhanced normal parenchyma for the calculation of lesion/parachyma signal ratio (LPSR). Image quality comparison between the last set of water-only CDT-VIBE images and the conventional GRE images was scored by two first-year radiology residents in consent based on a5-point scale in terms of the following parameters:PAT artifacts, edge sharpness, lesion conspicuity, internal structure clarity and overall quality. Before the comparison, two readers were trained by a senior radiologist on the interpretation and scores of each parameter. Two radiologists with more than5years experiences in breast MRI and the application of American College of Radiology Breast Imaging Reporting and Data System (BI-RADS) were asked to depict morphology characteristics of lesions observed on the last set of CDT-VIBE water-only images. One month later, the same evaluation was repeated for conventional GRE images. Two radiologists were aware of the position of suspicious lesion but blinded to image acquisition technique. The morphology characteristics were depicted in accordance with BI-RADS MRI Lexicon.4. Statistical analysis:Single sample K-S test was used to test whether SNR and LPSR were normally distributed. If they satisfied normal distribution, independent sample t-test was used to identify significant differences in SNRs between CDT-VIBE and GRE phantom images, and paired t-test was used to determine if LPSR of the last CDT-VIBE DCE images were significantly different from those of the conventional GRE images. Otherwise, nonparametric test was used. The Wilcoxon Signed-Rank Test was used to investigate statistically significant differences in the qualitative scores of the two image sets. The proportion of intra-reader agreement in morphology characterization was calculated. Kappa statistic (quadratic weighted) was used to assess inter-reader agreement of two radiologists on CDT-VIBE morphology characterization. All statistical analyses were conducted by using Medcalc (version12.7.0). P<0.05was considered statistically significant.Results1. SNR:On CDT-VIBE and conventional GRE images, mean SNR[SD] were74.00[0.901] and76.02[0.689] of the left phantom and55.71[0.891] and55.79[0.603] of the right phantom. There were no significant differences in SNR between the CDT-VIBE and conventional GRE images (left:P=0.229, right:P=0.653). The differences in SNR between left and right sides of phantom were not compared because they might not be caused by scanning sequence.2. LPSR:Mean LPSR[SD] of CDT-VIBE and conventional GRE image were2.222[0.607] and2.248[0.540]. There was no significant difference in LPSR (P=0.548) between the CDT-VIBE and conventional GRE images.3. Image quality:The scores of edge sharpness and lesion conspicuity on the CDT-VIBE images were very close to those on conventional GRE images (P=0.782,0.491). The overall image quality of CDT-VIBE was significantly lower than that of conventional GRE (P<0.001) because of more apparent PAT artifact (P<0.001) and mild blurry internal structure (P=0.046).4. Morphology characterization:The best proportion of intra-reader agreement of100%was seen in the description of lesion type and distribution modifiers of non-mass-like enhancement. The lowest, but moderate, proportion of intra-reader agreement of80%was for internal enhancement pattern of non-mass-like enhancement. The proportions of other items were above90%. On CDT-VIBE images, the best weighted Kappa value of inter-reader agreement was achieved in lesion type, mass shape and distribution modifiers of non-mass-like enhancement (1.000), and the internal enhancement pattern also generated the lowest but moderate Kappa value of0.745。ConclusionDespite apparent PAT artifacts and mild blur, CDT-VIBE sequence allows of high temporal resolution, dynamic imaging of breast with well preserved morphology characterization which is in good agreement with conventional GRE sequence. Therefore, CDT-VIBE is a promising tool for clinical breast DCE-MRI, especially for pharmacokinetic analysis, which has a high requirement of temporal resolution to acquire enough dynamic data in a short time. Part II Influence of temporal resolution on pharmacokinetic parameters estimation and diagnostic performance in breast benign and malignant lesionsObjectivesTo evaluate the influence of temporal resolution on pharmacokinetic parameters estimation and diagnostic performance in breast benign and malignant lesions.Materials and methods1. Clinical information:CDT-VIBE based breast dynamic contrast enhanced (DCE) MRI was performed on53patients with suspected breast lesions. All lesions were proved by biopsy or surgical pathology.2. Breast DCE-MRI:same with Part I.3. Data processing:The image set that showed contrast medium filled heart and its arteries were marked as the1st post-contrast phase. All original phases (including pre-contrast phases) were used to form dynamic data sets with temporal resolution of12s. Different phases of dynamic CDT-VIBE sequences were selected to simulate dynamic sets with varying temporal resolution, like pre-contrast phases and the1st,3rd,5th,7th,9th...... post-contrast phases to simulate temporal resolution of24s, pre-contrast phases and the1st,4th,7th,10th,13rd......to simulate temporal resolution of36s, pre-contrast phases and the1st,5th,9th,13rd,17th......to simulate temporal resolution of48s and pre-contrast phases and the1st,6th,11st,16th,21st...... to simulate temporal resolution of60s. In this way,5dynamic data sets with temporal resolution of12s,24s,36s,48s and60s were generated.4. Quantitative analyses:The quantitative analyses were done in Tissue4D with input images of T1map and dynamic data sets with different temporal resolution. Pharmacokinetic evaluation was based on the Tofts model and medium population average arterial input function provided by Tissue4D. The median of four pharmacokinetic parameters, including Ktrans (volume transfer constant between plasma and extravascular extracellular space [EES], min-1), ve (EES volume, min-1), kep (constant flux rate between EES and plasma) and iAUC (area under the contrast concentration curve at initial60s) were recorded for all patients. Finally,5sets of parameters were generated.5. Statistical analysis:Single sample K-S test was used to test whether Ktrans, kep, ve and iAUC were normally distributed. If these parameters satisfied normal distribution, paired t-test with Bonferroni correction was used to identify significant differences in5sets of pharmacokinetic parameters for benign and malignant breast lesions, respectively. Otherwise, nonparametric test would be used. Receiver operating characteristic curve (ROC) analysis and pairwise comparison of area under curve (AUC) with Bonferroni correction was used to assess diagnosis efficacy of parameters. Statistical analysis was conducted using was conducted using Medcalc (12.5.7). P<0.05was considered as statistical significant.Results1. Pathology diagnosis:Of55lesions identified in53women (two women had two lesions),29were malignant and26were benign. Of the29malignancies,18were invasive ductal carcinoma and11were ductal carcinoma in situ (with or without microinvasion). Of the26benign lesions,14were fibroadenoma,7were fibrocystic changes,4were papilloma and one was plasma cell mastitis.2. Influences of temporal resolution on parameters:2.1Ktrans:In benign lesions, mean Ktrans calculated from12s,24s and36s dynamic data sets were stable (0.147min-1,0.147min-1and0.148min-1), but significantly increased with reduced temporal resolution of48s and60s (0.164min-1and0.170min-1). The differences among benign Ktrans from varying temporal resolution data sets were not significant (corrected P>0.05). The mean Ktrans of malignant lesions were relatively stable with12s and24s temporal resolution (0.373min-1and0.364min-1), but significantly decreased with36s,48s and60s (0.315min-1,0.295min-1and0.259min-1). The differences in malignant Ktrans between12s,24s vs.36s,48s and60s were significant (corrected P<0.01), and the difference between36s vs.60s was significant (corrected P=0.024).2.2kep:In benign lesions, mean kep calculated from12s,24s and36s dynamic data sets were stable (0.321min-1,0.328min-1and0.336min-1), but significantly increased with reduced temporal resolution of48s and60s (0.382min-1and0.433min-1). The differences in benign kep between12s,24s and36s vs.60s and the difference between12s and36s vs.48s were significant (corrected P<0.05). The12s mean kep of malignant lesions were0.929min-1, and kep showed significant decrease from24s to60s (0.811min-1,0.769min-1,0.651min-1and0.581min-1). The differences of among all malignant kep values were significant (corrected P<0.05).2.3ve:The changes of mean ve with varying temporal resolution were mild and nearly identical in benign and malignant lesions. Compared with12s ve,24s ve were slightly higher (benign:0.452and0.468; malignant:0.436and0.468) and other three ve of36s,48s and60s were slightly lower (benign:0.435,0.415and0.427; malignant:0.418,0.421and0.424). The differences among benign ve values were not significant (corrected P>0.05). For malignant ve, the differences between12s vs.24s and24s vs.36s,48s and60s were statistical significant (corrected P<0.05).2.4iAUC:The mean iAUCs calculated from12s and24s were close (benign9.192and9.254; malignant20.221and19.832), but from36s to60s, iAUC showed significant decrease (benign8.709,8.402and7.388; malignant16.258,14.329and12.850). The differences between benign iAUC of12s,24s,36s and48s vs.60s were statistical significant (corrected P<0.05). In malignant lesions, all paired iAUCs were statistically different (corrected P<0.001) except12s vs.24s (corrected P=0.394).3. ROC analysis:The AUCs of Ktrans calculated from varying temporal resolution showed consistent decrease from0.887(12s) to0.754(60s) and the difference between12s and60s was significant (corrected P=0.048). The AUCs of12s,24s and36s kep were0.939,0.924and0.915. However, the AUC decreased to0.847with48s kep and to 0.749with60s kep. The differences in AUCs between12s,24s and36s vs.60s were significant (corrected P<0.05). The AUCs of12s and24s iAUC were close (0.860,0.877) and other three AUCs of36s,48s and60s were also close but slightly lower (0.838,0.836and0.839). Compared with previous three parameters, ve featured low AUC values ranging from0.511to0.598. For iAUC and ve, the differences among AUC values from varying temporal resolution were not significant (P>0.05).ConclusionTemporal resolution of dynamic sequence in breast DCE-MRI has strong influence on pharmacokinetic parameters. Ktrans, kep and iAUC of12s has the best discrimination between benign and malignant lesions. With the deterioration of temporal resolution, the differences between benign and malignant parameters are weakening, as well as parameters differential ability. Considering the close AUC values of12s and24s, a temporal resolution higher than24s might be adequate for clinical purpose. However, further decreasing temporal resolution might compromise the discrimination and diagnosis performance of pharmacokinetic parameters. Part Ⅲ Influence of scanning duration on pharmacokinetic parameters estimation and diagnostic performance in breast benign and malignant lesionsObjectivesTo evaluate the influence of scanning duration on pharmacokinetic parameters estimation and diagnostic performance in breast benign and malignant lesions.Materials and methods1. Clinical information:same with Part Ⅱ.2. Breast DCE-MRI:same with Part Ⅰ.3. Data processing:Pre-contrast sequences used to form the baseline were consisted of one35s full K-space sequence and3～4phases of12s sequences. After the injection of contrast medium, the image set that showed contrast medium filled heart and its arteries was marked as the1st post-contrast phase. All pre-contrast CDT-VIBE sequences and different phases of post-contrasted sequences, including1～10,1～15,1-20,1-25,1-30,1-35post-contrast phases were used to form7sets of dynamic series with the scanning duration of1-7min after the contrast injection.4. Quantitative analysis:The quantitative analyses were done in Tissue4D with input images of T1map and dynamic series with different scanning duration. Processing procedures and parameters calculation were the same with Part Ⅱ. Finally,7sets of quantitative parameters were generated.5. Statistical analysis:Single sample K-S test was used to test whether Ktrans, kep, ve and iAUC were normally distributed. If these parameters satisfied normal distribution, paired t-test with Bonferroni correction was used to identify significant differences in7sets of parameters for benign and malignant breast lesions, respectively. Otherwise, nonparametric test would be used. Receiver operating characteristic curve (ROC) analysis and pair-wise comparison of area under curve (AUC) with Bonferroni correction was used to assess diagnosis efficacy of parameters. Statistical analyses were conducted by using Medcalc (12.5.7). P<0.05was considered as statistical significant.Results1. Pathology diagnosis:same with Part Ⅱ.2. Influence of temporal resolution on parameters:2.1Ktrans:The observed change tendency of benign and malignant mean Ktrans was:Ktrans calculated from1min (benign:0.149min-1, malignant:0.383min-1) is higher than that from2min (0.126min-1,0.276min-1). From2min to7min, there were slight increases in both benign and malignant Ktrans (7min:0.147min-1,0.313min-1). The differences of benign Ktrans between lmin vs.2min and3min were significant (corrected P<0.001, corrected P=0.02). In malignant lesions, all paired Ktrans revealed statistical differences (corrected P<0.05) except2min vs.3min and4min and all pairs among5min,6min and7min (corrected P>0.05).2.2kep:The mean kep calculated from1min (benign:0.769min-1, malignant:1.255min-1) is higher than that from2min (0.399min-1,0.699min-1). The decrease extended to3min but became moderate (0.328min-1,0.644min-1). From3min to7min, there was a slight increase in both benign and malignant lesions (7min:0.363min-1,0.747min-1). In benign lesions, the differences among1min vs. all other scanning durations (corrected P<0.001) and between2min and3min (corrected P=0.03) were significant. In malignant lesions, all paired kep revealed statistical differences (corrected P<0.05) except2min vs.4min～7min,3min vs.4min and6min vs.7min (corrected P>0.05).2.3ve:The changes of mean ve were on the contrary to that of Ktrans and kep. From1min to4min, benign ve increased from0.226to0.466and malignant ve from0.352to0.480. With longer scanning duration, benign and malignant ve gradually decreased and reached0.459and0.445by7min. The difference of ve between benign and malignant diminished with longer scanning duration. In benign lesions, there were significant differences among lmin vs. all other scanning duration,2min vs.4min,5min and7min (corrected P<.05). In malignant lesions, significant differences were revealed among lmin vs. all other scanning durations,2min vs.3min,2min vs.4min and4min,5min vs.6min,7min (corrected P<0.001).2.4iAUC:Mean iAUC showed good consistency with varying scanning duration in both benign and malignant lesions. There was no significant difference in all iAUC values (P>0.05).3. ROC analysis:Longer sampling durations had no positive effect on the diagnosis performance of Ktrans. It was observed that the highest AUC value of0.932was for1min scanning duration and the AUC values decreased with longer scanning time and reached0.885by7min. The AUCs of kep showed consistent increase from0.824of1min to its highest0.913of5min, but decreased to0.893and0.888with6min and7min. The highest AUC of0.703for ve was also seen in lmin, whereas AUCs dramatically drops to below0.6from2min. For iAUC, the AUCs showed good consistency with varying scanning duration. There was no significant difference among AUCs pairs of Ktrans, kep, ve and iAUC (P>0.05).ConclusionDynamic scanning duration of breast DCE-MRI has a significant impact on pharmacokinetic parameter. The dynamic data collected within2min after contrast injection plays a crucial role in the calculation of benign quantitative parameters. In malignant lesions, besides the initial2min wash-in data, the delayed wash-out during3min-6min also produced a significant influence on parameter calculation. However, the ability to differentiate between benign and malignant breast lesions might not be significantly influenced by scanning duration. As scanning above6min does not add any improvement, a scanning time of5min for adequate dynamic information and patient’s comfort might be a reasonable choice. |
PDF Full Text Request