Phd in PHYSICS

Thesis title: NMR applications for the diagnosis of female pathologies

Nuclear magnetic resonance (NMR) imaging (MRI) and spectroscopy (MRS) are powerful techniques that can be used to study in-vivo human body and diseases. MRS allows the quantification of the metabolites and a chemical analysis of the analyzed sample, whereas MRI is a multi-parametric technique that provides detailed, cross-sectional images of the body's internal structures using non-ionizing radiation. In particular, the Diffusion-Weighted imaging (DWI) allows the evaluation of the diffusion properties of water in biological tissues without requiring contrast agents and with excellent resolution dependent on diffusion length. These techniques applied in the brain studies have been revolutionizing the neuroimaging field, leading to important information related to the microstructure and function of brain. To extract this information, the development of mathematical models has been the topic of greatest interest of the last 20 years. Indeed, in DWI the NMR signal is the Fourier transform of the motion propagator, which is easily calculated for free water, but it’s more complex in biological tissues leading to the development of models such as the Intravoxel incoherent motion (IVIM) model and the Kurtosis representation. All these models were firstly applied and optimized for neurological application; hence the body application should be carried out considering these limitations (related to the structural and physiological differences between cerebral and ex cranial tissues) and adapting the models to the specific analyzed tissues to avoid the evaluation of parameters without physical sense. Sometimes, clinical NMR images are affected by high level of noise which imply the wrong evaluation of the NMR parameters. Hence, it is necessary to know the signal to noise ratio (SNR) of each image and perform a denoising if the SNR is too low. In this thesis, a biophysical model with two perfusion compartments was adapted for placental tissues analysis to study the diffusion and perfusion properties of water in the placenta which is characterized by three main compartments: the fastest perfusion compartment due to the fetal villous-trees, a slower perfusion compartment related to the trophoblastic activity and a diffusion compartment linked to the maternal blood in the intra-villous space. The model was applied to evaluate the characteristics of the normal placental tissues and to find any differences between normal placentas and those affected by two specific pathologies: the fetal growth restriction disease (FGR) and the placental accretism. The application of this new model allowed the identification of new biomarkers non-obtainable applying the more known and used IVIM model, which nevertheless is able to highlight important perfusion features. Indeed, the IVIM model was used to study the differences between normal placentas and those belonging to women with SARS-CoV-2 infection, pinpointing substantial differences in the values of the diffusion coefficient due to a damage to the microstructure of the tissue. DWI also has important applications in the oncology field. In fact, it has been widely used to study the complexity of tumor tissues in the brain. In this work, DWI was adopted to study the cervical cancer and the endometrial cancer in particular. Generally, the diagnosis of these kinds of pathologies is performed by the histology which is an invasive technique and it is limited by the size of the taken tissue. In this thesis, tissue complexity and tumor grade were evaluated using the kurtosis representation combined with a clusterization based on the tissue differences in the same region of interest obtaining a higher variance of the kurtosis parameters in tumor tissues than in the healthy endometrium. Since the kurtosis representation is sensitive to image noise, a denoising was performed before applying the model and the non-dependence of the parameter to the noise was verified. MRS was used to pinpoint early biomarkers in bone marrow and muscles to study the osteoporosis and osteoarthritis diseases, highlighting interesting differences in the fat lipid content and in the level of unsaturated fatty-acids in both bones and muscles, corroborating the hypothesis that these pathologies involve the whole musculoskeletal system. Although the promising results, new perspectives should be considered for the diffusion model’s selection. Indeed, the dynamics underlying the studied system is crucial for the model selection since the application of the wrong model would bring the evaluation of physically no-sense parameters. In this work a preliminary experimental study was conducted on PEG to evaluate its true dynamics and then choose the right diffusion model applying a particular “recipe” conceived and developed in the NMR laboratory of the CNR-ISC & Sapienza where I carried out my PhD work. These last results underline the importance of the knowledge of the tissue diffusion properties a priori to perform more precise and faithful quantification of diffusion parameters indispensable for more precise and early clinical diagnostic.