2420
Quantitative Brain Diffusion Metrics are Fragile to Voxel Size: A Prospective Volunteer Study Calling for Protocol Standardization1Department of Imaging, Tongren Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China, 2Research Collaboration Team, Siemens Healthineers Ltd., Shanghai, China, 3Department of Radiology, Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China
Synopsis
Keywords: DWI/DTI/DKI, Diffusion/other diffusion imaging techniques, Reproducibility of results
Motivation: The quantitative brain diffusion metrics showed potential to provide functional information about diseases, but their robustness must be established before their implementation in clinical practice.
Goal(s): To evaluate the variability and reproducibility of quantitative brain diffusion metrics.
Approach: Fourteen volunteers prospectively underwent brain diffusion spectrum imagingusing a protocol designed to investigate the impact of scan-rescans, voxel size, coils on twenty-five quantitative diffusion metrics. Two observers drew thirteen regions of interests for metrics calculation.
Results: The voxel size has a greater influence on reproducibility of quantitative diffusion metrics than scan-rescans and coils. The reproducibility within an observer was higher than that between two observers.
Impact: Quantitative brain diffusion metrics, as promising tools for providing functional information of diseases, should be interpreted with caution because they are fragile to voxel size, which calls for standardization of scan protocols before prior to their implementation in clinical practice.
Introduction
Quantitative diffusion metrics have been demonstrated to be useful in the diagnosis and prognosis of brain tumors and breast cancers [1-3]. However, the robustness of these metrics must be evaluated before they can be translated from an academic research technique to a clinically practicable tool[4]. Therefore, in this study, we assess the impact of scan-rescans, voxel size, coils, and observers on the variability and reproducibility of quantitative diffusion metrics.Methods
This prospective study was approved by the ethical committee of our institution, and the written consents were obtained from all volunteers (Figure 1). We included fourteen healthy participants who volunteered to undergo conventional magnetic resonance imaging (MRI) and diffusion spectrum imaging for brain using a 3.0T system (MAGNETOM Vida, Siemens Healthcare, Erlangen, Germany). A protocol was designed to investigate the impact of scan-rescans, voxel size, and coils by changing acquisition parameters: 20-channel coil, voxel size 2×2×2 mm3, twice; 20-channel coil, voxel size 3×3×3 mm3, once; and 64-channel coil, voxel size 2×2×2 mm3, once (Table 1). A research post-processing software (NeuroDiLab, ChengDu ZhongYing Medical Technology Co., Ltd.) was used to calculate twenty-five metrics using four models: diffusion tensor imaging (DTI), diffusion kurtosis imaging (DKI), mean apparent propagator (MAP), and neurite orientation dispersion and density imaging (NODDI) (Figure 2). Thirteen regions of interest (ROI) were manually drawn using NeuroDiLab software. One neuroradiologist with 10 years of experience drawn and repeated the ROI two weeks apart, while another musculoskeletal radiologist with 5 years of experience drawn the ROI once. The variability of quantitative diffusion metrics for scans and segmentations was evaluated using coefficients of variation (CV) and quartile coefficient of dispersion (QCD). The reproducibility of quantitative diffusion metrics among scans and segmentations was measured using intraclass correlation coefficient (ICC), and concordance correlation coefficient (CCC). The variability and reproducibility of quantitative diffusion metrics were assessed using Wilcoxon signed rank test. A two- tailed P<0.05 was considered statistically significant.Results
The overall CV values ranged from 2.4% to 68.2%, and the QCD values ranged from 0.6% to 48.2% (Figure 3). The inter-scan, inter-resolution, and inter-coil variability measured by CV values were 2.4% to 26.1%, 5.7% to 68.2%, and 4.1% to 32.5%, respectively. The corresponding QCD values were 1.7% to 18.4%, 4.0% to 48.2%, and 2.9% to 23.0%, respectively. The CV and QCD values of inter-scan assessments were lower than those of inter-resolution (both P<0.001), and inter-coil (both P<0.001) assessments. The intra- and inter-observer variability, as measured by CV values were 3.4% to 40.6%, and 4.1% to 42.0%, respectively. The corresponding QCD values were 0.6% to 7.2%, and 0.7% to 7.4%, respectively. The ICC values ranged from 0.029 to 0.996, and the CCC values ranged from 0.029 to 0.996 (Figure 4). The inter-scan reproducibility was generally good to excellent (ICC 0.75-0.97, CCC 0.75-0.97). The inter-coil reproducibility was moderate to excellent (ICC 0.59-0.95, CCC 0.59-0.95), while the inter-resolution reproducibility ranged from poor to excellent (ICC 0.03-0.84, CCC 0.03-0.84). The inter-resolution reproducibility was worse than inter-scan reproducibility and inter-coil reproducibility (both P<0.001). The intra-observer reproducibility was generally moderate to excellent (ICC 0.50-0.94, CCC 0.50-0.94). The inter-observer reproducibility was generally moderate to excellent (ICC 0.56-0.91, CCC 0.56-0.91). The intra-observer reproducibility was better than inter-observer reproducibility (P<0.001).Discussion
In this study, we found that the stability of quantitative diffusion metrics varied significantly. The reproducibility was high, similar to that reported in previous studies on the brain [5-7]. The previous studies used coils with different numbers of channels for quantitative diffusion metrics acquisition [5-9]. We found that the coils increased the variability or damage the reproducibility in small. Therefore, a 20-channel coil may be a suitable choice for future participants or patients, providing a more comfortable scanning experience. The voxel size had a greater impact on the reproducibility of quantitative diffusion metrics than scan-rescans and coils. The voxel size in the protocols varied among current studies [1-3], and potentially limiting the generalizability of the results from one to another. This calls for standardization of scan protocols before prior to their implementation in clinical practice. The freehand ROI method has been demonstrated to be highly robust for DTI metrics [10]. Our study suggested that this method is suitable for research work with an experienced observer, as well as for those with only general experience in neuroradiology.Conclusion
Our results indicated that voxel size had a significant impact on the reproducibility of quantitative diffusion metrics. Therefore, we recommend that variability and reproducibility assessments of quantitative diffusion metrics should be conducted prior to the implementation of diffusion spectrum imaging protocols in clinical practice.Acknowledgements
The authors would like to express their gratitude to the volunteers for their participation in this study. The authors would like to acknowledge Dr. Shiqi Mao for his advice on data visualization, and Dr. Guangcheng Zhang for English editing.References
1. Mao J, Zeng W, Zhang Q, et al. Differentiation between high-grade gliomas and solitary brain metastases: a comparison of five diffusion-weighted MRI models. BMC Med Imaging 2020;20(1):124.
2. Mao C, Jiang W, Huang, et al. Quantitative parameters of diffusion spectrum imaging: HER2 status prediction in patients with breast cancer. Front Oncol 2022;12:817070.
3. She D, Huang H, Guo W, et al. Grading meningiomas with diffusion metrics: a comparison between diffusion kurtosis, mean apparent propagator, neurite orientation dispersion and density, and diffusion tensor imaging. Eur Radiol 2023;33(5):3671-3681.
4. European Society of Radiology (ESR). ESR statement on the validation of imaging biomarkers. Insights Imaging 2020;11(1):76.
5. Shahim P, Holleran L, Kim JH, Brody DL. Test-retest reliability of high spatial resolution diffusion tensor and diffusion kurtosis imaging. Sci Rep 2017;7(1):11141.
6. Andica C, Kamagata K, Hayashi T, et al. Scan-rescan and inter-vendor reproducibility of neurite orientation dispersion and density imaging metrics. Neuroradiology 2020;62(4):483-494.
7. Lehmann N, Aye N, Kaufmann J, et al. Longitudinal reproducibility of neurite orientation dispersion and density imaging (NODDI) derived metrics in the white matter. Neuroscience 2021;457:165-185.
8. Chung AW, Seunarine KK, Clark CA (2016) NODDI reproducibility and variability with magnetic field strength: a comparison between 1.5 T and 3 T. Hum Brain Mapp 37(12):4550-4565.
9. Thieleking R, Zhang R, Paerisch M et al (2021) Same brain, different look? - the impact of scanner, sequence and preprocessing on diffusion imaging outcome parameters. J Clin Med 10(21):4987.
10. Hakulinen U, Brander A, Ilvesmäki T et al (2021) Reliability of the freehand region-of-interest method in quantitative cerebral diffusion tensor imaging. BMC Med Imaging 21(1):144.
Figures
Figure 1 Study workflow
The workflow of this study includes following steps: participants inclusion, image acquisition, image postprocessing, ROI segmentation and statistical analysis.
Figure 2 An example of diffusion quantitative metric mappings
This example of diffusion quantitative metric mappings was generated using diffusion spectrum imaging data from a 28-year-old right-handed quadrilingual male participant with a doctorate degree. He reported drinking three cups of coffee daily, but had never smoked. He denied any history of neurological disorder, psychological disorder, traumatic brain injury, or prior neurosurgical procedure, and a conventional MRI brain scan revealed no abnormalities.
Figure 3 Variability of quantitative diffusion metrics
(A) Heatmap of CV and QCD values of quantitative diffusion metrics. (B) Category of variability according to CV and QCD values. The CV and QCD values were interpreted as follows: acceptable, <10%; moderate but still adequate, 11-20%; and too high and inadequate, ≥20%.
Figure 4 Reproducibility of quantitative diffusion metrics
(A) Heatmap of ICC and CCC values of quantitative diffusion metrics. (B) Category of reproducibility according to ICC and CCC values. The ICC and CCC values were interpreted as follows: poor, <0.50; moderate, 0.50–0.75; good, 0.75–0.90; or excellent, ≥0.90.
Table 1 Diffusion magnetic resonance imaging acquisition parameters
We designed a diffusion spectrum imaging protocol to investigate the robustness of quantitative diffusion metrics in brain using a 3.0-T system (MAGNETOM Vida, Siemens Healthcare, Erlangen, Germany): 20-channel coil, voxel size 2×2×2 mm3, twice; 20-channel coil, voxel size 3×3×3 mm3, once; and 64-channel coil, voxel size 2×2×2 mm3, once.