Towards robust Diffusion Tensor Imaging of skeletal muscles via an automatic artifact removal tool.
Chiara Giraudo1, Stano Motyka1, Siegfried Trattnig1, and Wolfgang Bogner1

1Department of Biomedical Imaging and Image-guided Therapy- MR Centre of Excellence, Medical University Vienna, Vienna, Austria

### Synopsis

STEAM-DTI sequence recently provided excellent results for DTI analysis of muscle fibers (e.g., high signal-to-noise ratio, low apparent diffusion coefficient, high fractional anisotropy values) but demonstrated also to be affected by strong artifacts, which can be assumed to be due to involuntary muscle contractions. The hereby proposed automatic post-processing method, based on weighted mean of the averages for each DTI-direction and b-value, demonstrated to successfully detect and correct these artifacts, improving fiber tracking of the calf muscles.

### Purpose

Diffusion Tensor Imaging (DTI) is nowadays successfully and reliably applied to characterize microstructural changes in the brain [1-3] but its application in the musculoskeletal field still has to face several challenges mainly associated with the necessity of using prolonged diffusion times which is an essential feature to optimize the sensitivity for fiber anisotropy in muscles [4]. For spin-echo-based DTI this translates into increased echo times (TE). STEAM-based DTI sequences demonstrated recently to provide excellent results for DTI analysis of muscle fibers (e.g., high signal-to-noise ratio, low apparent diffusion coefficient, high fractional anisotropy values) [4,5]. Indeed, this sequence allows substantially longer diffusion times without the disadvantage of signal loss that is associated with long TE. Moreover, for the same diffusion weighting factor (i.e, b-value) diffusion gradients in STEAM are much lower than in spin-echo diffusion, eliminating thus any eddy current distortion. Because of the highly promising premises of STEAM-based DTI, we decided to test its performance for DTI of muscles. However, in our population, this sequence demonstrated to be accompanied by the presence of artifacts: indeed, several areas of loss of signal, randomly localized and shaped were shown, which we hypothesized to be associated with muscle contractions. The hereby-presented study aimed to elaborate a post-processing method for detection and correction of these artifacts.

### Materials and Methods

Nine volunteers (4 males and 5 females; average age 30.2 years) without any history of muscle injuries have been scanned on a 3T TIM Trio MRI Scanner (Siemens, Erlangen, Germany). In each subject DTI and anatomical reference scans of the calf muscles were acquired. The protocol included a STEAM-DTI sequence (TR/TE 6300/33ms, diffusion time 200ms, b-value 500s/mm², matrix 96x96, FoV 180x180mm², 30 adjacent slices of 3.5 mm thickness, GRAPPA-2, FatSat); 6 averages and 12 directions were acquired. Positioning-matched axial T2w TSE sequence (TR/TE 5500/102 ms, matrix 512 x 512, FoV 180x180, 3.5 mm slice thickness) was acquired as anatomical reference. To test our hypothesis that the main origin of the artifacts was associated with small involuntary muscle contractions, we instructed two of the volunteers to perform voluntary muscle contractions during the acquisition of the STEAM sequence. i) To confirm the presence of artifacts each set of DTI-images was analyzed by a musculoskeletal radiologist, who marked each single slice that was affected by artifacts. ii) The same sets of images underwent the hereby-proposed automatic post-processing method for artifact correction. In the first step all DTI images with the same contrast (i.e., all six averages) are co-registered to correction for gross motion artifacts/misalignment. Then, the automated artifact correction method calculates a weighted mean of all six averages for each DTI-direction and b-value using the following formula: $$\mu_{weighted}=\frac{\sum_i^Nw_{i}x_{i}}{\sum_i^Nw_{i}}$$ Thereby, voxel values, which are affected by artifacts, are weighted far less than those, which are much more similar. In the end, all “weighted mean”-DTI images are co-registered among different DTI-directions and b-values. The images before and after the automated correction were analyzed through DSI Studio Software and for each volunteer one area affected by the artifact was delineated by a manual ROI and fiber-tracking performance was then assessed. To fully evaluate the advantages carried by the artifacts’ correction mean number, mean length and volume of tracts were collected before and after the automatic correction; paired t-test was applied for comparison (SPSS Statistics 21.0).

### Results

All the tested volunteers (n=9) demonstrated the presence of artifacts on the STEAM images datasets, which has been detected by the manual and automatic method.An average of 545 out of 1950 (27.95%) images in each volunteer was manually marked as affected by artifacts.The mean number, mean length and volume of tracks substantially increased after the post-processing method, showing a statistically significant difference at t-test (i.e., p=0.016, p=0.008 and p=0.21 respectively) [Fig 1, 2]. The images collected from the two volunteers, to whom voluntary muscle contractions were required during the scan, demonstrated a substantial increase in the amount of artifacts [Fig 3]. This evidence supported our hypothesis that involuntary muscle contractions occurring during the prolonged echo time, distinctive of the STEAM-DTI sequence, might be responsible for the majority of artifacts.

### Conclusions

The image quality of STEAM-based DTI for investigation of muscles of the lower limb can be severely affected by artifacts. These artifacts can be successfully detected and corrected by an automatic post-processing method based on the weighted mean of the averages for each DTI-direction and b-value. This improved substantially the image quality and the obtained fiber tracking. This automatic correction is expected to guarantee accurate DTI analysis in the musculoskeletal field for application of STEAM-DTI based sequence in clinical studies.

### Acknowledgements

No acknowledgement found.

### References

1. Lee SK, Kim DI, Kim J, et al. Diffusion-Tensor MR Imaging and Fiber Tractography: A New Method of Describing Aberrant Fiber Connections in Developmental CNS Anomalies. Radiographics 2005; 25(1): 53-65.

2. Alexander AL, Lee JE, Lazar M, et al. Diffusion Tensor Imaging of the brain. Neurotherapeutics 2007; 4(3): 316-3293)

3. Assaf Y, Pasternak O. Diffusion Tensor Imaging (DTI)-based white matter mapping in brain research: a review. J Mol Neurosci 2008; 34:51-61.

4. Noehren B, Andersen A, Damon B, et al. Comparison of twice refocused Spin Echo versus Stimulated Echo Diffusion Tensor Imaging for tracking muscle fibers. JMRI 2015;41:624-32.

5. Kim S, Chi-Fishman G, Barnett AS, Pierpaoli C. Dependence on Diffusion Time of Apparent Diffusion Tensor of Ex Vivo Calf Tongue and Heart. Magnetic Resonance in Medicine 2005; 54:1387-1396.

### Figures

Figure 1.Artifact in the muscle of the calf (STEAM-DTI sequence) (red arrow in A); localization of the artifact (i.e., subtracted image (B); ROI in the region represented in B and its fiber tracking before (C) and after the automatic correction (D), this last image showing the obtained improvement.

Figure 2.Diffusion anisotropy color-coded map (A) and fiber tracking reconstruction (B) of the calf muscles obtained applying the STEAM-DTI sequence.

Figure 3.Artifacts in the muscle of the calf during voluntary contractions (A) and at rest (red arrow in B; same volunteer); fiber tracking (ROI placed in the area showed in B) during muscle contractions and at rest on the native datasets (C,D) and after the automatic correction (E and F).

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
0751