Research Presentation Session: Physics in Medical Imaging

RPS 1913 - Magnetic resonance imaging (MRI)

March 5, 12:30 - 13:30 CET

8 min
SPIO nanoparticles-based phantom for standartistion of non-invasive MRI methods for iron overload assesment
Petr Алексеевич Bulanov, Moscow / Russia
Learning Objectives
Author Block: P. А. Bulanov, E. Manzhurtseva, P. Menshchikov, D. Kupriyanov, G. Novichkova, G. Tereshchenko; Moscow/RU
Purpose or Learning Objective: Liver iron overload is a pathological condition in which extra iron increase in tissues and organs, mainly in liver. It can lead to toxic damage and organs dysfunction. A biopsy is a common method for iron concentrations assessment, but it has a number of negative factors, such as risks of infection. The alternative, a non-invasive, method is T2* mapping, because T2* values depend on magnetic susceptibility which changes with an iron present. Earlier we carried out a study on the correlation of T2* values (ms) with LIC (mg/ml) values obtained from a liver biopsy from children with varying degrees of liver iron overload. However, this formula can be reliably used only with those MRIs on which it was obtained, because T2* values are very sensitive to various scanning parameters. The creation of T2* obtaining standardisation method is the main goal of the current study.
Methods or Background: We used an MR-compatible phantom which consisted of 28 test tubes with different solution concentrations of paramagnetic iron oxide nanoparticles. To check the results for repeatability, the phantom was scanned throughout the year in 1-week increments on our Philips Achiva 3T and Signa GE 1.5T MRI scanners. We further verified reproducibility on other two scanners in other medicine centers.
Results or Findings: The T2* values showed good repeatability on both our scanners. As a result of working with other centres we formulated the standardisation protocol: (1) scanning phantom in another centre, (2) calibration of scanning parameters and T2* mapping or introducing additional calibration coefficients into the recalculation formula, and (3) using our formula to convert T2* into LIC.
Conclusion: Thus, to be able to use formulas, it is enough to check the correspondence of T2* values using the phantom we created.
Limitations: Not applicable.
Ethics committee approval: Not applicable.
Funding for this study: Not applicable.
8 min
Varying MRI-contrast in solid 3D printed structures without adding contrast agents
Alejandra Valladares, Vienna / Austria
Learning Objectives
Author Block: A. Valladares, G. Oberoi, T. Beyer, A. Berg, E. Unger, I. Rausch; Vienna/AT
Purpose or Learning Objective: The development of MRI phantoms presents several challenges in terms of materials and is mainly limited to water-filled compartments and gel-based components. Here, we present a novel design to additively manufacture solid objects with adjustable MRI contrast based on the integration of MRI-visible supporting material.
Methods or Background: Eleven homogeneous cubic probes and one spherical object mimicking a necrotic tumour were designed and printed using additive manufacturing. Each probe comprises a solid internal structure and a specific percentage of commonly-used support material. The models were scanned using standard spin- and gradient-echo pulse sequences on a 3T PET/MRI system. We evaluated MRI visibility and T1 and T2 relaxation rates over time.
Results or Findings: For most probes, measurable signal intensities were obtained. T2 values ranged from 22 ms to 40 ms, comparable to reported values for some body tissues at 3T. T1 values were relatively short compared to similar reports with relaxation times of 109 ms to 131 ms. T1 and T2 relaxation times were stable over three months.
Conclusion: The proposed design allowed the adjustment of imaging contrast in MRI, based mainly on the spin density contribution of the MRI-visible supporting material within a voxel. Furthermore, we demonstrated the possibility of creating a one-compartment object to mimic necrosis in tumours, which is advantageous to create complex models for tumour heterogeneity in MRI and PET/MRI.
Limitations: The current design does not allow mimicking the varying contrast with different T1 and T2- weighing MRI sequences.
Ethics committee approval: Not applicable.
Funding for this study: This work has received funding from the Horizon 2020 programme under the Marie Skłodowska-Curie grant agreement No.764458. AM infrastructure was supported by the Austrian FFG within the M3dRES project (Grant-No.858060).
8 min
Comparisons between simulated and experimental data in quantitative susceptibility maps in a 3.0T MRI-scanner
Enrique Gustavo Cuña, Montevideo / Uruguay
Learning Objectives
Author Block: E. G. Cuña1, E. Tuzzi2, L. Biagi3, P. Bosco3, M. Garcia1, J. Mattos1, K. Scheffler2, M. Tosetti3, G. E. Hagberg2; 1Montevideo/UY, 2Tübingen/DE, 3Pisa/IT
Purpose or Learning Objective: To compare quantitative susceptibility quantification between simulated and experimental phantom containing iron. This would help to adjust processing pipeline parameters to obtain quantitative susceptibility mapping (QSM).
Methods or Background: A phantom was constructed by inserting small vials into a cylindrical container. Vials contained FeCl2 at four different concentrations (range 0.22-1.79mM) in two forms: clustered and free. The phantom was scanned at 3.0T (Siemens Germany). Multi-echo-GRE-images (TE=6:6:30ms; TR=53ms; nominal FA=18°; voxel=600x600x600μm; GRAPPA=2). For simulations, the same geometry as the real phantom was used. Susceptibility values were assigned to three different regions: background (1ppm), body (0ppm) and iron vials. A molar susceptibility of 1ppm/mM was assigned. For clustered iron vials, random noise (with a uniform probability distribution, from 0 to 1) was generated and multiplied by the corresponding iron concentration. Phase images at TE=6ms were obtained by convolution of the phantom geometry with the macroscopic unit-dipole function. Morphology enabled dipole inversion (MEDI)2 was applied to the earliest echo-time phase images (6ms) to obtain QSM for experimental and simulated data using lambda1=1000 and lambda2=10 for background regularisation.
Results or Findings: For measured QSM images, clustered iron led to highly localised field effects, also captured by simulations. For experimental data, best curve fittings were obtained after applying a Laplacian-based unwrapping pre-processing, with determination coefficients >0.88. In the simulations, QSM-values in both clustered and free iron showed a linear increase with iron (determination coefficient >0.99). In both cases, estimated molar susceptibility was lower with clustered iron.
Conclusion: Our simulation method captures the effect of iron clustering in QSM calculations as seen in experimental phantom acquisitions.
Limitations: QSM processing pipelines need to be refined to achieve higher accuracy for local field effects, as also seen in Alzheimer’s beta-amyloid plaques.
Ethics committee approval: Not applicable.
Funding for this study: Validation EU-LACH #16/T01-0118.
8 min
Validation of a novel segmentation tool (segfatMR) for semiautomatic volumetry of adipose tissue in MR images
Anna Linder, Leipzig / Germany
Learning Objectives
Author Block: A. Linder, T. Eggebrecht, N. P. Linder, T. Denecke, H. Busse; Leipzig/DE
Purpose or Learning Objective: Subcutaneous (SAT) and visceral adipose tissue (VAT) are considered as biomarkers for a variety of clinical disciplines and questions. Segmentation and volumetry are generally time-consuming, creating a need for rapid but accurate analysis tools. This work aims to present a new segmentation tool (developed under a custom framework, DicomFlex [RS]) that combines the speed of automatic pre-segmentation with the reliability of a brief interactive revision. Results were validated against a widely used commercial tool (SliceOmatic).
Methods or Background: Validation was performed on MRI datasets (1.5 T Achieva XR, Philips Healthcare) from 20 patients (10 women/men), 25.1-63.1 (mean 48.5) years old, with BMIs between 28.3 and 58.8 (mean 36.8) kg/m2. Two independent readers analysed the abdominopelvic datasets (30-40 slices, mean 35.8 ) with both tools (segfatMR, SliceOmatic). Coefficients of determination (R2) as well as bias and limits of agreement (Bland Altman) were determined.
Results or Findings: Segmentation performance (R2 between methods) was excellent for both readers for SAT (> 0.99) and very high for VAT (around 0.90). The novel method was almost twice as fast as the reference – 19 (R2) and 25 (R1) s/slice vs 34 and 40 s/slice. On a subjective level, readers appreciated the intuitive workflow resulting from the standardised interface.
Conclusion: The open-source semiautomatic PC segmentation tool enables a fast and accurate quantification of whole abdominopelvic adipose tissue volume. Development under a standardised software framework (Dicomflex) allows for further extensions or adaptations.
Limitations: Limitations include single-centre-design, limited number of readers (n=2) and specific patient selection based on BMI(>25kg/m²).
Ethics committee approval: MRI study was under IRB approval.
Funding for this study: There was no specific funding for this study.
8 min
Skin temperature changes in pregnant women during foetal MRI
Laura Wachholz, Jena / Germany
Learning Objectives
Author Block: L. Wachholz, H-J. Mentzel, A. Heinrich, H. Proquitte, E. Schleussner, P. Schlattmann, C-H. Cho-Nöth, U. K. M. Teichgräber; Jena/DE
Purpose or Learning Objective: Foetal MRI will be performed as an adjunct to sonography in foetal pathologies. Generally, MRI is related with heat, caused by radiofrequency waves produced during MRI scans. The purpose of this prospective study was to confirm that there is no heating in pregnant women during MRI.
Methods or Background: 20 pregnant women (aged between 19 and 45 years) in a range of 20 to 35 pregnancy week underwent foetal MRI for various indications. Imaging was performed on a 1.5 T MRI (AERA, SIEMENS, GERMANY) without sedation or contrast media. Imaging time was between 30 and 45 minutes. Two probes Tesla M3 (Mammendorfer Institut für Physik und Medizin, GERMANY) were used for measuring the skin temperature: one on the leg outside the field of view (FOV) (TL) and the second one next to the belly button within the FOV (TFOV). Before and after each MRI investigation the body core temperature was estimated by evaluation of the ear temperature (Tc).
Results or Findings: Both probes demonstrated in all 20 pregnant women an increase of the skin temperature of 0.1°C per imaging minute. TFOV increased 3.6°C (mean; min 1.8, max 5.6°C), TL 1.7°C (mean; min 0.1°C, max 2.6°C), Tc 0.3°C (mean; min -0.3°C, max 1.2°C).
Conclusion: We observed a significant increase in temperature in pregnant women during MRI. We don’t know if there is any side effect on the foetus. Therefore, possible health consequences, also for the fetus, should be checked.
Limitations: The limitations are the missing online rectal and intrauterine temperature measurements.
Ethics committee approval: The ethics committee approval was obtained.
Funding for this study: Not applicable, institutional study.
8 min
Ultrafast total body EPI-DWIBS with compressed sensing
Petr Menshchikov, Moscow / Russia
Learning Objectives
Author Block: I. Karpov, P. Menshchikov, D. Kupriyanov, E. Patrikeev, A. Morozov; Moscow/RU
Purpose or Learning Objective: Compressed Sensing (CS) is a novel and promising method for the accurate reconstruction from sparsely sampled k space data that allows significant noise reduction compared with parallel imaging methods (SENSE). The study's main aim was to use CS to speed up EPI DWIBS for the whole body.
Methods or Background: Axial EPI-DWIBS images (voxel-2.5×2.5×6 mm, b-factor-1000 s/mm, 30 slices) were acquired both for the phantom and 8 healthy subjects in the abdomen with a 1.5T Philips Ingenia MRI scanner. The phantom contained 8 tubes with pure water, fat and water solution with PVP 10-20-30-40-50-60-70%. Phantom acquisition: Sense/CS factors varied from 1 to 3 with 0.5 increment; NSA=4 and 6. For the in vivo study the combinations of the SENSE/CS and NSA were as follows: SENSE=2, NSA=6; CS=2, NSA=6; CS=1.5, NSA=6; CS=1.5, NSA=4.
Results or Findings: Phantom DWIBS images with CS acceleration show significantly higher SNR than images with appropriate SENSE CS both for NSA=4 and 6. SNR of phantom DWIBS image with SENSE=2, NSA=6 (standard implementation) is equal to the CS=1.5, NSA=4. The same result was obtained for the in vivo studies. Comparing in vivo images with NSA=6, CS=1.5 and CS=2 shows significantly higher SNR (on 14% and 8%, respectively).
Conclusion: CS acceleration shows a higher efficiency than SENSE for EPI-DWIBS based on the SNR analysis. Combination CS=1.5, NSA=4 allows to acquire images with the same quality as for the standard acquisition SENSE=2, NSA=6, reduces the scan time from 1:37 to 1:03 min per one stack resulting in 11 min for full-body DWIBS.
Limitations: The decrease of the accelerating factor is accompanied by increasing the EPI-factor, which may cause additional EPI distortions. For all subjects, no EPI distortions on DWIBS images were found.
Ethics committee approval: The study was approved by NMRCTO N.N.Priorova.
Funding for this study: No funding was received for this study.
8 min
Analysis of slice profile effects for super resolution thin-slice multi-slice MRI
Stephen J. Riederer, Rochester / United States
Learning Objectives
Author Block: S. Riederer, E. borisch, R. Grimm; Rochester, MN/US
Purpose or Learning Objective: Application of super-resolution methods to the slice select direction attempts to recover through-plane resolution finer than the slice thickness. Critical in this process is the correction for the slice profile. However, conventional “rect” profiles have zeroes in the kZ passband, causing undesirable attenuation of high spatial frequencies in the slice direction. The purpose of this work is to investigate alternative slice profiles which allow improved fidelity.
Methods or Background: First, several alternative RF profiles were identified with no zeroes in the kZ passband. These are a Gaussian and a (1+cos) (raised cosine, or “rCos”). Second, experimental images using these were compared with a standard “rect” profile using a phantom with resolution patterns varying from 0.15 to 0.50 lp/mm. Third, sagittal images were acquired of the prostate in vivo from several subjects. Axial sampling using 3mm thick slices with the various slice profiles was simulated using several levels of additive noise, super-resolution reconstruction performed, and results compared with the original sagittal image. Finally, results were compared with sagittal reformats made from the actual axial acquisition.
Results or Findings: For 3mm thick slices, both Gaussian and rCos RF pulses allow full restoration of spatial frequencies up to 0.50 1/mm using SR reconstruction, equivalent to 1mm slice thickness. rCos excitation consistently provides improved sharpness vs Gaussian over all noise levels. Sagittal reformats of SR-reconstructed axially acquired data well resemble direct sagittal acquisition.
Conclusion: Alternative (non-rect) RF slice profiles such as Gaussian and rCos allow improved fidelity of high spatial frequencies in the slice direction with resolution finer than that in the acquired slices.
Limitations: The number of subjects studied is limited. Finer (<1mm) slice resolution is desirable.
Ethics committee approval: Approved by Institutional Review Board (IRB).
Funding for this study: Funding for this study was institutional.