Functional magnetic resonance imaging studies in small animals

Résumé

This thesis is framed within the field of preclinical biomedical imaging, and specifically devoted to the study of functional magnetic resonance imaging (fMRI) technique in small animals. The experimental and technological complexity of this modality has greatly limited its use, and therefore it is not a routine imaging modality. However, it provides valuable information both at the physiological level, to study the mechanisms of normal brain during neuronal activity, and at the pathological level, to study drugs intended for different brain dysfunctions. In this work we have studied techniques and methods that intend to alleviate these difficulties and facilitate their use by the scientific community. The work includes contributions at several stages: the experimental setup, the data acquisition and reconstruction, and the quantitative image analysis. The first section addresses the problem of using anesthesia during the experiment. In order to perform functional measurements, it is necessary to establish a protocol to induce anesthetic sedation of the animal rather than a deep anesthetic state. Moreover, the use of non-toxic drugs with fast induction and recovery is desirable. In this section of the thesis we conducted fMRI experiments in rats sedated with sevoflurane, and since this agent had not been previously reported for fMRI, it was necessary to conduct strategies in order to determine the optimum dose-response and stimulation frequency. Furthermore, the signal obtained in the cerebral cortex was compared with a more traditional protocol sedation, subdermal medetomidine. The signal obtained was similar to that obtained under medetomidine, but the animal preparation time increased considerably, which constitutes a serious practical drawback for the use of sevoflurane. The second section is devoted to the study of a compressed sensing framework that allows a substantial reduction on the acquisition time without degrading image quality. The acquisition of a much reduced amount of data, thus at high rates of acceleration that violate the Nyquist-Shannon criterion, is possible by means of a wise exploitation of the temporal information redundancy and by the use of nonlinear iterative reconstruction algorithms. In this study we evaluated the performance of three compressed-sensing reconstruction algorithms that exploit temporal redundancy to recover the BOLD contrast and which have proved successful in other applications or imaging modalities such as: X-ray tomography, dynamic cardiac MRI, and resting state MRI studies. The comparison was performed in two signal-to-noise ratio scenarios and the conclusion drawn is that the algorithm which uses an a priori image (PICCS) yields the best reconstruction. The third section deals with the post-processing and image analysis. There are several open-source tools available to this purpose, but they were originally designed for human studies. Their adaptation to rodent images requires the use of additional tools or some image transformation processing that involve programming skills. Moreover, to obtain quantitative values, the user would need to use additional extensions or external software. In this work we have studied the existing tools and proposed and developed a new software, fMRat, which automatically performs a full multi-subject analysis, from the initial format conversion to the extraction of numerical values from the regions interest chosen by the user. The tool was programmed in Matlab as an extension of the existing SPM package, and was validated with 460 real rat studies. The code has been published as "open-software" in Github website and is accessible to the neuroscience community.