Friday, September 11, 2009

Spectral Quality & Artefacts

Effect of spurious echoes. Insufficient amplitude of gradient crusher pulses in combination with local B0 inhomogeneities can lead to the refocusing of unwanted echoes (e.g. 2 pulse echo in a PRESS sequence). (A) The FID from a PRESS acquisition (TE 20 ms, TR 3 s) localizing developing white matter in a female preterm neonate (34 weeks gestational age). The encircled part of the FID originates from an unwanted echo. (B) The typical appearance of spurious echoes, often called ghosts, in the spectrum. Because extended phase cycling was used (phase rotation) in data acquisition, the origin of the spurious signal could be identified in a separate trace after Fourier transformation along the phase rotation dimension (C). The particular phase evolution proved the spurious signal to arise from a two-pulse echo of the initial 90 and last 180 pulse. In the current case, elimination of the ghosting artifact can easily be accomplished by zeroing the latter half of the FID. The resulting spectrum is plotted in (D).

Effect of head movements. All peaks are doubled in a spectrum from a neonate because the baby had moved its head between two distinct positions during the scan (E). The repeat examination shows single peaks with perfect shim and lineshape (F), when the baby was soundly asleep. (41 weeks gestational age, ROI in thalamus, PRESS TE 20 ms, TR 2 s, 128 acquisitions

Effect of eddy currents. All lineshapes are distorted in spectrum (G) due to eddy currents in a short-TE PRESS spectrum (TE 20 ms) of occipital GM in a 14-year-old boy. (H) The same spectrum after restoration of the lineshape using the phase information from a water reference scan.

Effect of gross head movement. An ROI was placed in the putamen on a localizer image (A). The resulting spectrum is shown in (B). However, unknown to the operator, the examined subject had turned his head a little towards his left, which was picked up from the repeated localizer scan after the MRS acquisition (C). As ROIs are prescribed in magnet coordinates, the ROI targeted at the putamen ended up in insular GM, such that the spectrum in (B) was in reality acquired from insula, not putamen. This explains the narrow lines, which are atypical for basal ganglia. A spectrum from the putamen was then acquired (D) and shows that a completely wrong diagnosis would have resulted when the spectrum from insular GM was taken as originating from putamen. Verification of proper ROI placement is crucial. (Scan parameters: 26-year-old man; MRI: fast spin echo sequence, echo train length 16, TR 3 s, TE 100 ms; MRS: 2.2 cm3 ROI, PRESS, TE 20 ms, TR 3 s, 128 acquisitions).

Signal bleed from outside the targeted ROI. Signal from outside the selected ROI can give dominating signal contributions, if the transition zone of the slice selective pulses falls into regions with large lipid content. This is illustrated for PRESS spectra obtained from a 40-year-old woman. The original ROI dimensions of 10 x 15 x 27 mm, used for spectra (A) and (B) were reduced to 10 x 15 x 22 mm for spectrum (C). This diminished the transition zone of the longest dimension of the voxel pointing towards the lipid-containing areas and the lipid contribution vanished. Just moving the ROI away from the skull, (E)–(F), did not completely eliminate the lipid contamination in the spectrum, (A)–(B). (Scan parameters: TE 20 ms, TR 3 s, 1953 Hz spectral width, 1024 points zero-filled to 2048 points, outer volume suppression pulses disabled).

Conspicuity of artifacts in MRI and MRS. If a patient leaves the magnet half-way through a scan, even a layman will refrain from interpreting the resulting image (A). If this happens in a MRS scan, even the expert will not be able to recognize this fact from the resulting spectrum (B), since only signal-to-noise and absolute concentrations will be affected. Spectra (B) (half of the acquired FIDs contain noise only) and (C) (normal acquisition) were scaled to the largest peak, resulting in an apparent signal-to-noise difference, while quantitative analysis would yield a 50% deficit for all metabolites. Unless double-checking mechanisms are put in place and plausibility arguments are used, the resulting diagnosis will be completely wrong. (Scan parameters: 38-year-old healthy woman; MRI, fast spin echo with TE 102 ms, TR 3 s, 256 x 256, 4 mm slice thickness; MRS, PRESS with TE 20 ms, TR 3 s, 6.7 cm3 ROI in periventricular GM, 128 acquisitions).

Source: Kreis R. Issues of spectral quality in clinical 1H-magnetic resonance spectroscopy and a gallery of artifacts. NMR Biomed. 2004;17:361-381.

No comments:

Post a Comment