The g-factor in sensitivity encoding (SENSE) PI
The availability of an optimized coil array is an essential prerequisite for parallel imaging techniques, as PI makes use of differences in the spatial sensitivity distribution of receiver coil elements to reduce the number of necessary phase-encoding steps required for magnetic resonance imaging, thus shortening the scan time. Most commonly used PI algorithms are SENSE, SMASH and GRAPPA [3] and are subdivided in two domains (k-space/image-space) in which processing is mainly performed. For this project we used the SENSE method described as an image domain reconstruction technique. The reconstruction of images acquired by parallel imaging require the resolution of an inverse problem, and the g-factor appearing as a consequence of the resolution describes the ability with the used coil configuration to separated pixels superimposed by aliasing. In practice it is a rough estimate of the noise amplification in the spatial domain, and is related to the SNRp of SENSE reconstructed images and the SNRf, which would be obtained without acceleration, by
Fig.: Relation between SENSE SNR and non-accelerated SNR
References: Pruessmann et al. Magn Reson Med 1999; 42:952-62.
where R denotes the acceleration factor and gp is the g-factor. Therefore an optimization of gp would be beneficial in term of SNR. The calculation of gp is done from [4]
Fig.: g-factor equation
References: Pruessmann et al. Magn Reson Med 1999; 42:952-62.
where S is the ncxnp sensitivity matrix, SH is the transposed complex conjugate of S and Ψ is the ncxnc noise covariant matrix, nc is the number of channels and np is the number of overlapping pixels.
High-sensitivity 16 channels breast coils
A 16 channel variable coil geometry breast imaging system (Sentinelle Medical, Toronto, Canada) consists of coils embedded in a fixed medial coil housing and two L/R and A/P adjustable lateral coils, figure 1.
Fig.: 16 channels variable geometry coil system design to accommodate a range of breast sizes and shapes. The coil can be adjusted in the LR and AP direction.
The two lateral coils can be moved to gently compress the breast tissue during scanning. This serves to immobilize the breasts during acquisition to minimize motion artifacts and also to reform the breasts to fit into the most sensitive volumes of the coil. These adjustable coils have been shown to provide high SNR and perform well for SENSE acceleration [5].
Customized phantom
A breast shape-like phantom was designed and manufactured to fit within the 16 channel coil array (figure 2), and to represent average breast dimensions as determined by measurements performed on a sample of previous breast images. The choice of the phantom design was important in that it had to be of a shape that was appropriate for realistic breast imaging. The phantom simulates both breasts and a chest wall component and was chosen to emulate bilateral imaging.
Fig.: Customized phantom designed to mimic the breast shape
The phantom was filled with distilled water doped with MnCl2 at a concentration of 0.245mM to achieve a T1 relaxation time of approximately 500ms, as well as NaCl at a concentration of 150mM to simulate the conductivity of breast tissue and give realistic coil loading.
Phantom scans were used to generate g-factor maps for quantitative evaluation of parallel imaging performance. Coil sensitivity data was acquired using an axial 3D SPGR sequence (FOV=36x36cm, matrix: 64x64x86, TR/TE/θ = 6.4ms/3.0ms/15o, BW = 31.25 kHz). RF excitation was disabled for noise coupling measurements using 2D SPGR (matrix: 256x256, BW = 31.25 kHz). Various degrees of acceleration were simulated by sub-sampling the original data in the LR and SI directions.
3D SPGR volunteer images were obtained using: TR/TE = 4.5/2.0ms, θ = 12o, BW = 167 kHz, FOV = 35 cm, 124 slices, 1.4 mm slice thickness, matrix = 320x320, and IDEAL fat suppression. The coils were tested on a 3T scanner (General Electric Inc. Waukesha,WI). All data were analyzed off-line using Matlab (Mathworks Inc. Natick, MA).