Spinal fusion is the most commonly employed surgical procedure for managing conditions of the spine. Unfortunately, these procedures often do not produce expected results. The failure of surgical intervention in the spine to relieve symptoms often is described by the generalized term ‚”Failed Back Surgery Syndrome” (FBSS) [1].

Fusion procedures often implant metallic hardware, which generate MRI artifacts. In practice, spine surgeons often install titanium alloyed hardware components. While titanium implants induce far less artifact than other metallic alloys, easily discernible artifacts remain in conventional MRI. To elucidate the nature of these subtle artifacts, both conventional metal artifact reduction sequences (MARS) and advanced 3D multi-spectral imaging (3D-MSI) techniques were applied and analyzed on a cohort of symptomatic FBSS subjects.

Commercially available 3D-MSI sequences are not yet equipped with prospective capabilities to adjust to different implant scenarios. The lower susceptibility titanium alloys utilized in spinal fusion devices are often imaged with 3D-MSI parameters that are‚ ”overkill”‚ for the problem at hand. Therefore, in this study, a prospective calibration scan (~1 min) was utilized to calibrate each 3D-MSI sequence to the implanted fusion device [2,3]. This spectral calibration provides crucial efficiency improvements to 3D-MSI, which further allows for higher resolution and higher SNR spinal imaging, often with shorter acquisition times than conventional MARS techniques.

MRI were collected from a GE Healthcare Optima 450w GEM 1.5T system. GEM variable receiver array configurations were utilized for all spine acquisitions. A cohort of symptomatic FBSS subjects (N=21) provided written consent for an imaging research protocol approved by the MCW Institutional Review Board. Subjects were imaged with a routine MARS acquisition protocol as well as a series of T1w and T2w prospectively calibrated MAVRIC SL 3D-MSI images. The most commonly fused segments were cervical and lumbar, with 10 subjects each. One subject underwent evaluation after thoracic.

To quantify artifact reduction, a subset of subjects was identified from the study cohort (N=6, 3 L-Spine, 3 C-Spine) who previously had CT imaging of their spinal fusion. Bone CT images of titanium fusion devices have limited beam scattering artifacts and allow for easy identification of implant interfaces. After affine co-registration of CT/MARS/3D-MSI image volumes, implant or artifact volumes were identified using a combination of manual segmentation (MRI) and seeded threshold-based region growing (CT). The difference of the implant/artifact volumes between the MRI and the CT provided an estimate of residual artifact volume. Multiple-observer measurements were used to estimate the error bars in these volume-estimates. To asses the diagnostic impact of calibrated 3D-MSI, the MARS and 3D-MSI image sets were reviewed and compared with clinical records to independently identify complications in each image set.

Figure 1 provides volumetric renderings of a segmented lumbar spine fusion device. Quantitatively, the subtraction of measured and rendered volumes in this case yielded 3.9±1.6 cm3 of residual artifact in the MAVRIC SL 3D-MSI images and 58.0±12.2cm3 of residual artifact in the MARS images. It is important to note that the dominant volume displacement is occurring in the vertical dimension, which is the slice-selective dimension in this axial MARS acquisition.

Figure 2 shows that for all subjects studied, the measured artifact volume in the MARS image was significantly higher than in the MAVRIC images. The 3D-MSI residual artifact was subtle and within the measurement error. For the subject shown in Figure 1, Figure 3 shows a 3D rendering of the segmented nerve-roots from the MAVRIC and MARS images. It is clear where the expanded artifact volume of the MARS image interferes with the nerve root and can impact clinical assessments.

Clinically, 10 of the subjects exhibited normal post-surgical findings on plain-film radiography, CT, and MRI. Imaging findings identified adjacent segment disease in eight subjects; in which it was at least partially identified on both MARS and 3D-MSI. Of these cases, 3D-MSI uniquely identified foraminal pathology in three subjects, nerve root edema in one subject, epidural fibrosis in one subject, and a malpositioned screw in one subject.

Figure 4 provides sample images from this MARS preliminary clinical analysis. Foraminal pathology adjacent to a pedicle screw is highlighted in the indicated zoomed region. In the MAVRIC SL image, a compression of the nerve root is identified (dashed white arrow). This compression is obscured by artifact (dashed white arrow) in the corresponding MARS image.