Table 2.4:Abbreviated and full breast MRI protocols
First author Protocol Dynamic pre-contrast Dynamic 1st post-contrast Dynamic 2nd post-contrast Dynamic 3rd-5th post-contrast T1 TSE T2 DWI Kuhl76 Abbreviated x x Full x x x x x x Mango52 Abbreviated x x Full x x x x x x x Grimm53 Abbreviated1 x x x Abbreviated2 x x x x Full x x x x x x Heacock54 Abbreviated x x x Full x x x x x x
breviated protocol. Studies reporting on abbreviated protocols varied widely in ac- quisition times ranging from approximately 3 minutes to 15 minutes. In particular, the role of a T2-weighted sequence for screening purposes needs to be investigated. In addition, the number of patients scanned and the number of cancers detected is currently still too low to draw solid conclusions. Future larger prospective trials need to prove the non-inferiority of abbreviated protocols. Nevertheless, shortened breast MRI protocols could play a relevant role in lowering costs and allowing more widespread availability of MRI as a screening tool.
2.8
Ultra-fast breast MRI
All the abbreviated protocols discussed in the previous paragraph discard dynamic information. Only one of those investigated by Grimm and coworkers53 used the
second dynamic post-contrast scan. This is not problematic for larger malignant le- sions, which are generally well recognized based upon their morphological features. However, in particular for the classification of small mass lesions, which are typi- cal findings in breast screening, additional dynamic information is important. This implies that dynamic information is appreciated, while imaging time should not be extended.
Conventional dynamic information cannot be obtained, as this requires acquisition of the wash-out phase of contrast which takes up to 6 - 7 minutes after contrast ad- ministration. However, even in the early days of breast MRI, it was already shown that dynamic information obtained from the inflow phase had better discriminating capacity than the wash-out phase. Nevertheless, in previous years the temporal res- olution, typically in the range of 60 - 75 seconds of high spatial resolution bilateral images was not sufficient to document this inflow phase. Therefore, acquiring scans
Figure 2.4: Ultrafast axial images in a 42-years-oldBRCA2mutation carrier, using a TWIST sequence (temporal resolution 4.57s). The central slice of each of the first 12 acquisitions after aorta enhancement are shown, numbered from 1 to 12. The corre- sponding volumetric MIPs are shown at every time point, below the original central unsubtracted images. These clearly show the arrival of the contrast, first in the tho- racic vessels and the heart, subsequently in the breasts and liver. No suspicious early enhancement is visible in the breasts.
2.8 Ultra-fast breast MRI 31 at a high temporal resolution re-enables the use of contrast dynamics for the classifi- cation of suspicious breast lesions77.
Karl-Heinz Herrmann and coworkers78were among the first to describe a new ultra-
fast sequence named time-resolved angiography with stochastic trajectories (TWIST) for breast MRI (see Figure 2.4). With this technique, the outer part of k-space is heavily under-sampled and data points are shared between successive time points to increase the obtained spatial resolution to diagnostic quality. Sophisticated sam- pling patterns are used to minimize the disadvantage of data sharing. These authors showed, in a pilot study of 14 patients, that this TWIST sequence can be used to ob- tain dynamic images at a very high temporal resolution (5.7 seconds). Furthermore, they showed that benign lesions enhance at a later time point than malignant lesions. Luminita A. Tudorica and coworkers79reduced the temporal resolution to 18 s, show-
ing that the dynamic images were very comparable with the images provided by the conventional protocol. Yuan Le and coworkers80showed that a TWIST sequence can
be combined with a dual-echo (two-point) Dixon technique to obtain fat-suppressed images with a high temporal resolution. Our group55 investigated the use of maxi-
mum slope of the contrast enhancement versus time curve obtained from the TWIST sequence at a temporal resolution of 4.3 seconds as a novel dynamic parameter for the differentiation between benign and malignant lesions. The total acquisition time was 102 seconds. Of the 199 enhancing lesions included, 95 were proven benign and 104 malignant. We found that maximum slope achieved a much higher accu- racy in differentiating benign and malignant lesions than the BI-RADS curve type does, thus, solidifying the use of ultrafast breast MRI, and allowing the creation of new protocols with a short post-contrast period (∼85 seconds). While we did not evaluate TWIST for morphological features, the technique meets every breast MRI requirement that is stated in guidelines. Results of a recently presented reader study in which four radiologists evaluated 200 screening cases showed that the use of ultra- fast MRI alone was just as accurate as evaluating a full diagnostic protocol including high spatial resolution acquisitions, T2-weighted imaging, and DWI81.
Federico D. Pineda and coworkers77investigated a bilateral, fat-suppressed ultrafast
acquisition with a time resolution of 6.9-9.9 s during the first minute after contrast injection, followed by four high spatial resolution acquisitions with a time resolu- tion of 60-79.5 s. They confirmed that first-minute ultrafast dynamic imaging can add valuable information, increasing the radiologists confidence in identifying le- sions in the presence of marked background parenchymal enhancement. A hybrid
construction, where ultrafast acquisitions are interleaved in an abbreviated breast MRI protocol, allows the collection of dynamic data for lesion classification without a penalty in acquisition time.
Further improvements of ultrafast MRI are still increasing image quality. Radial imaging using a golden angle approach, as is performed in the golden-angle radial sparse parallel (GRASP) sequence, enables dynamic imaging using continuous data acquisition and retrospective reconstruction of image series with an arbitrary tem- poral resolution by grouping different numbers of consecutive radial lines into tem- poral frames. This means that, with the use of GRASP, images of every temporal resolution can be reconstructed, thus, both ultrafast and regular high spatial reso- lution acquisitions can be obtained using the same sequence, as described for liver, pediatrics, breast, and neck82. This approach can help to improve clinical work- flow by enabling data acquisition without the need for synchronization with breath- hold commands or for selection of predefined rigid temporal resolution. A recently published study83showed that the performance of the GRASP sequence in terms of
conspicuity of benign and malignant breast lesions is near-comparable to that of con- ventional volumetric imaging breath-hold examination (VIBE) imaging. Thus, tech- niques that employ compressed sensing might be used to further improve image quality of ultrafast imaging. Table 2.5 lists multiparametric protocols that include ultrafast MRI sequences recently described in literature.