CSF-Free Imaging of the Lumbar Plexus using Sub-Millimeter Resolutions with 3D TSE
Barbara Cervantes1, Houchun Harry Hu2, Amber Pokorney2, Dominik Weidlich1, Hendrik Kooijman3, Ernst Rummeny1, Axel Haase4, Jan S Kirschke5, and Dimitrios C Karampinos1

1Diagnostic and Interventional Radiology, Technische Universität München, Munich, Germany, 2Radiology, Phoenix Children’s Hospital, Phoenix, AZ, United States, 3Philips Healthcare, Hamburg, Germany, 4Zentralinstitut für Medizintechnik, Garching, Germany, 5Neuroradiology, Technische Universität München, Munich, Germany


High-resolution MRI with 3D turbo spin echo (TSE) is arising as an accurate, non-invasive method for detecting disease and injury in the nerves of the lumbar plexus. Imaging of the lumbar plexus with 3D TSE frequently faces signal contamination of the cerebrospinal fluid (CSF) within the spine. Increasing spatial resolution in 3D TSE can affect flowing signal. The present study describes numerically the effects of the imaging gradients in 3D TSE on flowing CSF and demonstrates in vivo that CSF can be completely suppressed without modifications to refocusing angle modulation when sub-millimeter voxel sizes are used with 3D TSE.


Lumbar nerves are susceptible to traumatic and inflammatory disorders, sometimes only reliably detected with high-resolution MRI [1]. T2-weighted 3D turbo spin echo (TSE) is frequently used to visualize lumbar nerve roots but faces signal contamination from high cerebrospinal fluid (CSF) within the spine [2]. Previous studies have shown that refocusing flip angle (FA) modulation in 3D TSE can reduce CSF signal in the cervical spine [3]. In 3D TSE imaging of the lumbar plexus, given that frequency encoding is commonly aligned to the direction of flow of CSF, increasing resolution can reduce CSF signal without altering FA modulation. The present work shows, for 3D TSE, 1) numerical reductions of CSF signal with increasing readout resolution and 2) complete CSF suppression in vivo in adult and pediatric subjects.


Extended phase graphs (EPG) allows numerical assessment of signal in pulse sequences with long refocusing trains [4] considering FA modulation and relaxation. Effects related to the imaging gradients can be characterized by EPG by incorporating additional information into the algorithm [5]. Motion sensitization in 3D TSE can become significant when large readout gradients are used, which occurs when the readout voxel dimension is reduced. The signal of a tissue moving with velocity v can be computed with EPG by including the phase shift implied by each readout gradient in the gradient duration $$$\Delta t$$$, $$$\Delta \phi = \gamma (\text{v} \times \text{m}_{\text{0}}) \Delta t$$$, where $$$\text{m}_{\text{0}}$$$ is in the gradient direction and $$$|\text{m}_{\text{0}}|$$$ is the gradient area. FA modulation gives different $$$\Delta \phi$$$ for different magnetization pathways, resulting in signal losses that scale with gradient area and velocity (Fig.1).


Simulations: Signal of CSF was simulated using EPG considering dephasing from coherent motion for voxel sizes from 0.3 to 1.3 mm and for CSF velocities from 0 to 4 cm/s, in agreement with observed values [6]. FA modulation optimized to yield high signal of small nerves [7] was used. CSF signal was computed with $$$\text{T}_{1}=3120$$$ms and $$$\text{T}_{2}=160$$$ms.

In vivo measurements: The lumbar plexus of all subjects was imaged using a 3T Philips scanner (Philips Ingenia, Best, the Netherlands) with the 12-channel posterior coil and a 16-channel torso coil only in the study of the healthy adult volunteer. $$$\text{T}_{2}$$$-weighted 3D TSE was used with echo spacing=4ms, TR/TE=2000/330ms, TSE factor=150. First, a resolution comparison in a healthy 29-year-old adult volunteer was performed with FOV=488×400×80$$$\text{mm}^{3}$$$, acquisition voxel = a) 1.4, b) 0.7, c) 0.4 ×1.25×1.4 $$$\text{mm}^{3}$$$, scan duration = 4m22s. Second, pediatric subjects a) 8 months, b) 2 years and c) 8 years of age were scanned with FOV = a) 180×208×50, b) 230×230×50, c) 200×231×50 $$$\text{mm}^{3}$$$, acquisition voxel = 0.7×0.7×1.4$$$\text{mm}^{3}$$$, respectively, scan duration = 4min18s. Scans a) and b) of pediatric subjects used a 32-channel head coil in combination with the posterior coil.


Simulated CSF signal is reduced by over 90% compared to the maximum simulated signal for flow velocities above 2cm/s and readout voxel sizes below 0.7mm (Fig.2). In the healthy adult volunteer, using a voxel size of 1.4mm in the readout direction yields high signal intensities of CSF within the spine (Fig3a). Using a voxel size of 0.7mm results in reduced CSF signal (Fig. 3b); further reducing the voxel size to 0.4mm results in the complete CSF suppression (Fig.3c). Imaging of the lumbar plexus of pediatric subjects using a readout voxel size of 0.7mm results in complete CSF suppression (Fig.4).

Discussion & Conclusion

Simulation results show complete CSF suppression, without FA modulation modifications, only with sub-millimeter readout voxel sizes. Although it has been shown that FA modulation alone can reduce the CSF signal [3], doing so can affect the signal of small nerves. Here it is shown that the signal of CSF can be completely suppressed while using FA modulation optimized for nerve signal. A particular readout voxel size results in different CSF signal reductions in subjects with different ages. CSF velocities in the lumbar spine in adults have been observed to reach 2cm/s [8], while those in children have been observed to be higher [9]. CSF suppression using sub-millimeter readout resolutions is therefore shown to work optimally in pediatric subjects. Complete CSF suppression is possible in adult subjects by further reducing the readout voxel size, with the obvious disadvantage of reducing SNR. The present work therefore 1) shows that complete suppression of CSF within the spine is possible in adult and pediatric subjects using sub-millimeter-resolution 3D TSE and 2) demonstrates in vivo that CSF suppression allows the visualization of nerve structures within the spine.


The present work was supported by Philips Healthcare.


[1] Soldatos, RadioGraphics 33:967, 2013

[2] Chhabra, Radiol Res Pract, 809568, 2013

[3] Busse, Proc. ISMRM 2006, p. 2430

[4] Wiegel, J Magn Reson 205:276, 2010

[5] Weigel, J Magn Reson 67:1528, 2012

[6] Bhadelia, AJNR:16, 1995

[7] Cervantes, JMRI doi:10.1002/jmri.25076, 2015

[8] Higuchi, Anasthesiology 100:106, 2004

[9] Haughton, AJNR doi:10.3174/ajnr.A4023, 2014


Figure 1: Signal loss of CSF due to motion simulated with EPG for 3D TSE with refocusing angle modulation adapted to nerve signal with a readout voxel size of 0.5 mm. The curves denote the signal evolution along the echo train in 3D TSE. For this voxel size, CSF signal is largely reduced for increasing flow velocities.

Figure 2: CSF signal (a.u.) simulated with EPG for flow velocities from 0 to 4 cm/s and for voxel sizes ranging from 0.3 to 1.3 mm. The highlighted area denotes sub-millimeter resolutions, where the signal of flowing CSF can be suppressed by the readout gradients. Blue intensities represent regions where CSF signal is reduced to more than 90% of the maximum simulated signal, corresponding to static CS and represented by yellow intensities.

Figure 3: In vivo images of the lumbar plexus of a healthy adult volunteer obtained with 3D TSE with a) 1.4 mm, b) 0.7 mm and c) 0.4 mm readout voxel sizes. CSF signal is suppressed at sub-millimeter resolutions (b and c), which leads to the visibility of the continuation of the nerve roots at and within the walls of the spinal canal.

Figure 4: CSF-free 3D TSE imaging of the lumbar plexus shown in vivo in pediatric subjects a) 8 months, b) 2 years and c) 8 years of age using a readout voxel size of 0.7 mm. Sub-millimeter voxel sizes in pediatric subjects result in the complete suppression of CSF signal, while refocusing-angle modulation maximizes nerve signal intensity.

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)