Präzise Volumetrie von menschlichen Fettkompartimenten durch Kombinationvon automatischen Auswertealgorithmen und angepassten MRT-Aufnahmetechniken

• Antragsteller: Professor Dr. Fritz Schick; Professor Dr.-Ing. Bin Yang
• Fachliche Zuordnung: Nuklearmedizin, Strahlentherapie, Strahlenbiologie
• Förderung: Förderung von 2011 bis 2016
• Projektkennung: Deutsche Forschungsgemeinschaft (DFG) - Projektnummer 192749169

• Erstellungsjahr: 2016

Zusammenfassung der Projektergebnisse

The key scientific outcomes of this project were initiation of development of machine-learningbased segmentation algorithms which can be applied to different organs and tissues as well as different contrast weightings. One of these methods employed the statistical shape and local appearance models to automatically delineate abdominal cavity over MR images of both Dixon and T1-weighted acquisitions. This enabled application of a common segmentation method to the images of those contrasts to compare the accuracy of the widely used imaging techniques for automatic segmentation of visceral adipose tissue at 3 Tesla. In this comparison, effects of intensity nonuniformities and partial volumes have been quantitatively addressed and compensated. To reduce intensity nonuniformities a previously proposed method of our institute has been extended to enhance the accuracy of segmentation of lean muscle mass. Thus it could be of potential application to the projects involving muscle segmentation. Another used machinelearning-based segmentation method is the random forest algorithm. This method has been applied for segmenting vertebral body in order to facilitate future follow up studies on morphometrics characteristics and composition of this organ. In addition to above techniques, for segmenting visceral adipose tissue a hybrid segmentation algorithm has been proposed. This method combines a region-based and a contour-based segmentation technique in order to enhance the capture range, convex object handling, and noise suppression. For selective segmentation of mammary adipose tissue, we have developed a scheme using fatwater MR images in which after segmentation of the entire breast by finding the outer boundary of pectoral muscle the mammary adipose tissue has been segmented from the subcutaneous adipose tissue by using the Bellman-Ford-Moore shortest path algorithm. This method is of potential interest for studies on correlation between the volume of mammary adipose tissue and breast pathogenesis. For segmenting cardiac adipose tissues, multi-gradient-echo chemical shift encoded images of the heart have been acquired successfully. Using these images, fat, water, field map, and T2* map have been quantified and epi- and peri-cardial and peri-vascular adipose tissues have been segmented in the thoracic region. These segmentations have been done by an unsupervised graph-based segmentation algorithm in which the appropriate landmark points were identified by a multi-atlas registration of the entire cardiac structure. The obtained results are promising for subjects of significant cardiac adiposity. Regarding the compromise between signal-to-noise-ratio (SNR), scan time efficiency, and spatial resolution in MR imaging, the present study have evaluated multiple super-resolution techniques for MR imaging. Also we have proposed a novel approach for a combined partial volume reduction and super resolution estimation to enhance the edges between different tissues without introducing artefactual effects. For chemical shift decomposition of species a method for simultaneous T1-weighted imaging and proton density fat fraction mapping using single-scan multi-echo spoiled gradient echo sequences has been proposed. Additionally, a new method for edge preserved decomposition of fat and water images from chemical shift encoded data is recently developed. Furthermore, multiple pulse sequences have been developed in our institute for compressed sensing partial sub-sampling of k-space in order to speed up image acquisition, to enhance the SNR, and to correct for motion artifacts in long scan time scenarios. Those implementations have shown promising results for future use in multi-gradient-echo quantitative analysis of the heart and abdomen. Additionally, some of those techniques have been applied to correct for motion artifacts in positron emission tomography (PET) via MR imaging. The initiated work on estimation and compensation of cardiac motion is of high clinical value for the cases where patient arrhythmia or other disorders hinder efficient use of electrocardiography triggering or breath hold imaging for avoidance of motion artifacts.

Projektbezogene Publikationen (Auswahl)

• "A new 3D hybrid algorithm for automatic segmentation of visceral adipose tissue", in Proceedings of 32nd Annual Scientific Meeting of ESMRMB Congress, 2015;28–29
  F. Fallah, F. Schick, B. Yang
  
• "Combining Compressed Sensing with motion correction in acquisition and reconstruction for PET/MR", in Proceedings of IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP); 2015. p. 788–792
  Küstner T, Würslin C, Schmidt H, Yang B
  
• "Comparison of T1- weighted imaging and two point chemical shift decomposition for volumetry of visceral adipose tissue", in Proceedings of 32nd Annual Scientific Meeting of ESMRMB Congress, 2015;166–167
  F. Fallah, J. Machann, P. Martirosian, M. Bongers, F. Schick, B. Yang
  
• "RF and Coil Inhomogeneity Correction in 2D Leg Images: A New Method Comparing with LEMS", in Proceedings of 23th Annual Scientific Meeting of ISMRM, 2015;1176
  F. Fallah, C. Würslin, F. Schick, and B. Yang
  
• "Simultaneous T1-weighted imaging and proton density fat fraction mapping using single-scan multi-echo spoiled GRE", in Proceedings of 32nd Annual Scientific Meeting of ESMRMB Congress, 2015;496–497
  F. Fallah, F. Schick, B. Yang
  
• "3D motion flow estimation using local all-pass filters", in Proceedings of IEEE 13th International Symposium on Biomedical Imaging (ISBI); 2016. p. 282–285
  Gilliam C, Küstner T, Blu T
  
• "A combined partial volume reduction and superresolution reconstruction for magnetic resonance images", in Proceedings of IEEE Nuclear Science Symposium and Medical Imaging Conference, 2016
  F. Fallah, B. Yang, F. Schick, F. Bamberg
  
• "Automatic separate quantification of mammary adipose tissue and subcutaneous adipose tissue in the female breast using fat-water MR imaging at 3 Tesla", in Proceedings of 33rd Annual Scientific Meeting of ESMRMB Congress, 2016;443–444
  F. Fallah, P. Martirosian, J. Machann, F. Schick, B. Yang
  
• "Automatic volumetry of cardiac adipose tissues using opposed-phase magnetic resonance imaging at 3 Tesla", in Proceedings of 33rd Annual Scientific Meeting of ESMRMB Congress, 2016;67–68
  F. Fallah, P. Martirosian, J. Machann, B. Yang, F. Schick
  
• "Comparison of T1-weighted 2D TSE, 3D SPGR, and two-point 3D Dixon MRI for automated segmentation of visceral adipose tissue at 3 Tesla", Journal of Magnetic Resonance Materials in Physics, Biology and Medicine (MAGMA), 2016;1–13
  F. Fallah, J. Machann, P. Martirosian, F. Bamberg, F. Schick, B. Yang
  
• "Evaluation of the performance of post-acquisition resolution enhancement methods for magnetic resonance images", in Proceedings of 33rd Annual Scientific Meeting of ESMRMB Congress, 2016;178–179
  F. Fallah, D. Di Franco, B. Yang, F. Bamberg
  
• "Effects of intensity nonuniformities, nonuniformity correction methods, and intensity normalizations on texture-based classification of T2-weighted aged calf muscles", in Proceedings of 33rd Annual Scientific Meeting of ESMRMB Congress, 2016;328–329
  F. Fallah, N. Schwenzer, F. Schick, B. Yang
  
• "MR image reconstruction using a combination of Compressed Sensing and partial Fourier acquisition: ESPReSSo", IEEE Transactions on Medical Imaging, 2016;PP(99):1–1
  Küstner T, Würslin C, Gatidis S, Martirosian P, Nikolaou K, Schwenzer N, et al.
  
• "Selfnavigated 4D cartesian imaging of periodic motion in the body trunk using partial kspace compressed sensing", Magnetic Resonance in Medicine, 2016
  Küstner T, Würslin C, Schwartz M, Martirosian P, Gatidis S, Brendle C, et al.
  
• "Simultaneous in-vivo respiratory and cardiac motion correction system for PET/MR", in Proceedings of 24th Annual Scientific Meeting of ISMRM, Singapore; 2016
  Küstner T, Würslin C, Schwartz M, Martirosian P, Gatidis S, Nikolaou K, et al.