Evaluation of the diagnostic utility on 1.5T and 3.0T 1H magnetic resonance spectroscopy for temporal lobe epilepsy | BMC Medical Imaging

Participants and experimental design

This study was approved by the ethics committee of the Second Affiliated Hospital of Xiamen Medical College. Figure 1 shows the data collection process, we retrospectively collected single-voxel ¹H-MRS of bilateral temporal poles from 2017 to 2021. There were 23 TLE patients diagnosed with the 1.5T scanner (15 males and 8 females, age: 29.52 ± 13 [Mean ± Standard Deviation (SD)]), 29 TLE patients diagnosed with the 3.0T scanner (20 males and 9 females, age: 28.2 ± 9.0), and 17 healthy controls (HCs) (11 males and 6 females, age: 23.35 ± 4.11) were scanned with both 1.5T and 3.0T MR scanners. To minimize the impact of multi-factor experimental conditions on outcomes, data were collected using scanners with different field strengths on the same healthy volunteer. To reduce the effect of time, HCs was scanned in batches in July 2021, with each batch scanned at both 1.5T and 3.0T in the same evening.

¹H-MRS data collection flowchart of the study. The left and right temporal pole were used as regions of interest (ROI) and represented by red boxes to obtain metabolite concentrations of Cho, Cr, and NAA of subjects. The outer volume suppression (OVS), which are not parallel to the voxel edges, are manually added to enhance the saturation effect

The diagnosis procedure of TLE seizures was as follows: First, the neurologist identifies whether the patient was having a seizure according to clinical symptoms, and then employed auxiliary monitoring, such as blood tests, neurological examinations, and electrocardiograms, to rule out causes of non-brain abnormalities; Second, TLE seizures were confirmed by recording epileptiform discharges through an electroencephalogram; Finally, by analyzing imaging (MRI and ¹H-MRS) and genes, neuroradiologists ascertained the cause of temporal lobe epileptic seizures (TLES). The whole process was completed by a neuroradiologist and an experienced neurologist, and the diagnosis result was finally given by an authoritative neurologist.

The ¹H-MRS data were acquired from the bilateral temporal poles of all subjects, and four control experiments were designed (Fig. 2) as followed: (1) 1.5T TLE group vs. 1.5T HCs; (2) 3.0T TLE group vs. 3.0T HCs; (3) the power analysis between the 3.0T scanner and 1.5T scanner based on the statistical test of the TLE and HCs; and (4) 3.0T HCs vs. 1.5T HCs. These comparisons were aimed at evaluating the differences in the diagnostic utility of TLE at 1.5T and 3.0T, both of which were the most widespread magnetic resonance fields.

Schematic diagram of four control experiments. The number on the left side of the box indicates the sample size in an experiment

MRI and ¹H-MRS acquisition

MRI acquisition parameters

1.5T MR scanner was equipped with GE SIGNA HD medical system, and the specific imaging parameters were as follows: (1) T2 FLAIR sequence: TR = 8600ms, TE = 120ms, TI = 2100ms, FOV = 240 × 240mm², matrix = 288 × 160, slice thickness = 5 mm, NEX = 1; (2) FRFSE T2WI sequence: TR = 4760ms, TE = 102ms, FOV = 240 × 240mm², matrix = 320 × 256, slice thickness = 5 mm, number of slices = 19, NEX = 1; (3) DWI: b value = 1000s/mm², TR = 6000ms, TE = 56ms, FOV = 240 × 240mm², matrix = 128 × 128, slice thickness = 5 mm, number of slices = 19, NEX = 2.

3.0T MR scanner was equipped with GE Discovery Silent MR (750 W) medical system, and the specific imaging parameters were as follows: (1) T2 FLAIR sequence: TR = 8600ms, TE = 140ms, TI = 2100ms, FOV = 240 × 240mm², matrix = 288 × 224, slice thickness = 5 mm, number of slices = 22, NEX = 1; (2) T2WI: TR = 4425ms, TE = 90ms, FOV = 240 × 240mm², matrix = 384 × 384, slice thickness = 5 mm, number of slices = 22; NEX = 1, (3) DWI: b value = 1000s/mm², TR = 4250ms, TE = 56ms, FOV = 240 × 240mm², matrix = 128 × 128, slice thickness = 5 mm, number of slices = 22, NEX = 2.

¹H-MRS acquisition parameters

As shown in Table 1, PRESS sequence was applied to 1.5T scanner with the following sequence parameters: voxel size = 20 × 20 × 20mm³, chemical shift imaging layer thickness = 20 mm, NEX = 128, TR = 2000ms, TE = 144ms. PRESS sequence was also applied to 3.0T scanner with the following sequence parameters: voxel size = 20 × 20 × 20mm³, chemical shift imaging layer thickness = 20 mm, NEX = 128, TR = 1500ms, TE = 144ms. High-resolution T2W was used to localize single voxels in regions of interest (ROI) in the temporal lobe that coincides with the seizure onset zone determined by electroencephalography (EEG).

Data pre-processing

After the MRS data of bilateral temporal lobes are collected, technicians use the built-in MRS software package (SAGE7.1) of GE scanner to quantify the data with the default mode to obtain the metabolite concentrations, and then calculate the metabolite concentration ratios, including NAA/Cr, NAA/Cho, NAA/(Cho + Cr) and Cho/Cr.

Statistical analysis

The variables were the ratios (NAA/Cr, NAA/Cho, NAA/ (Cho + Cr), Cho/Cr) of the ¹H-MRS metabolites of the bilateral temporal poles at 1.5T and 3.0T. The data were grouped according to different magnetic field strengths and subjects’ characteristics. The Mann-Whitney U Test model was used for the 1.5T TLE group vs. 1.5T HCs, and 3.0T TLE group vs. 3.0T HCs. Each healthy volunteer was scanned using both 1.5T and 3.0T scanners, providing a set of paired and correlated data. Therefore, the Pair T-Test was used to compare 3.0T HCs with 1.5T HCs. The difference is considered statistically significant if p < 0.05. The power analysis was used to compare diagnostic capability differences between 1.5T and 3.0T scanners, and the probability of error, whose value depends on the significance criterion (α), the sample size (N), and the population effect size (ES) [23] was assessed by the G*Power (version 3.1.9.7). As shown below [24], the power value can be calculated:

where H₀ means the null hypothesis, H₁ means the alternative hypothesis, and β means the probability of the error. The smaller power value suggests that β value is larger, and thus, the probability of the misjudgment increases, i.e. it is misjudged, even though the metabolite ratio is statistically different between 1.5T and 3.0T.