• medical imaging analysis
• musculoskeletal MR imaging
• relaxation time
• segmentation
• sequence

The work described in this thesis has been carried out in the framework of a research topic devoted to the study of NeuroMuscular Diseases (NMD) by Magnetic Resonance Imaging (MRI) as reliable and non invasive instrument for diagnosis and follow up. In particular, this preliminary study has been realized to assess the opportunities offered by musculoskeletal imaging with MRI (MSK-MRI) at Ultra High Field (UHF). The first and unique MR scanner at 7T in Italy for human studies has been recently installed by the IMAGO 7 Foundation (Calambrone, Pisa). In the diagnostic study of NMD, MRI has become an important imaging method able to detect muscle involvement and to evaluate the severity of the disease. It already exists a pediatric protocol routinely used in clinical MRI at low magnetic field (1.5T) in which lower extremities are examined in the same session and the total acquisition time is about 30 minutes. The evaluation of these images is now only visual and software tools for quantitative analysis do not exist. Standardization procedures and quantitative methods could be very useful instruments to optimize the diagnostic performances. Moreover, in the past few years static magnetic fields of increasing strength have been employed to increase Signal to Noise Ratio (SNR), to improve image contrast and spatial resolution, and to decrease the acquisition time. Increasing the magnetic field the behavior of spins in the sample changes, and in particular change the relaxation times of tissues which are related to the magnetic field strength. These modifications should be studied in order to optimize and implement acquisition parameters for muscle MRI at 7T. In fact, the sequences parameters used at low magnetic field to extract different information from the sample can not be employed at Ultra High Field.In this context, the thesis work has been focused on these two main topics: the development of a software for the quantitative analysis of the clinical muscle MRI images currently available and the realization of an acquisition protocol to broadly characterize the tissues of interest in NMD (fat and muscle) in ex-vivo samples at 7T and acquire the first in-vivo images of human volunteers.After a description of the physical principles and technological equipment of nuclear magnetic resonance and magnetic resonance imaging, a method to develop an automatic algorithm of segmentation has been exposed, able also to quantifying fat infiltration percentage in muscles of the leg. Such algorithm has been preliminary optimized and validated by means of a total of 26 MRI scans (10 healthy subjects and 16 NMD patients) already available at the MR laboratory of IRCCS Stella Maris and acquired with a clinical 1.5T MR scanner Signa GE Medical Systems. The scans of healthy subjects have been studied in order to extract and quantify from the MR images a reference standard, absent in literature, of anatomical parameters as geometry and signal intensity of muscle and fat tissues. The results have then been used as reference parameters to evaluate the grade of disease of NMD patients. The resulting automatic system has been developed and implemented as an original hierarchical module network in the MeVisLab medical image analysis framework. The separation between fat and muscle done at single voxel (volume elements) level allowed a quantitative automatic evaluation of fat infiltration percentage. This result has then been related to the grade of muscle impairment in the NMD patients. The scans of 8 of the 16 NMD patients have been selected by the child neuropsychiatrist involved in this study to form the gold standard for 4 grades of effective muscle involvement. By applying this method to the scans of the 8 NMD patients of the validation set, a great agreement with respect to the diagnosis of the child neuropsychiatrist has been obtained.About the topic related to image acquisition at 7T, a complete characterization and optimization of acquisition sequences to realize muscle and fat images has been conducted on the GE 7T Human research MRI system of IMAGO7 foundation. Relaxation times T1, T2 and T2* at 7T
for ex-vivo muscle and fat tissues have been measured in pork leg samples respectively acquiring images with Inversion Recovery (varying Inversion Time), Spin Echo (varying Echo Time) and Gradient Echo (varying Echo Time) and then calculating the desired parameters. Moreover, the Dixon technique has been studied for a possible use at 7T, which is based on chemical shift between water and fat. The signals acquired at different Echo Time (TE) are used to realize two images with different information, with water and fat spins in phase and out of phase, respectively. By means of a numerical processing of these images is possible to obtain other two images, one with the fat signal suppressed ("water image") and one with the water signal suppressed ("fat image"). The implementation as additional feature in Spoiled Gradient Recalled Echo (SPGR) sequence has been tested and optimized by evaluating the Ernst angle, which is the flip angle that maximizes the echo signal. Taking into account the results of this preliminary work, the differences between ex-vivo and in-vivo samples and the data available in the literature, a first attempt of protocol for human calf has been implemented and used to acquire images of a first healthy volunteer at 7T. In conclusion, the software for automatic and quantitative analysis of muscle MR images resulted a promising instrument both in the characterization of muscle and fat patterns in healthy subjects and in the classification of NMD patients for the clinical images currently available, acquired at 1.5T. The data acquisition for this work is still in progress in order to obtain a statistically significant dataset. Moreover, a preliminary protocol suitable for human in-vivo muscle acquisitions at 7T
has been realized and the first images of human calf have been acquired.