The overall objective of the present work is the translation of advanced qMRI techniques from the research environment into the field of clinical neuroimaging. In this context, qMRI is defined as the application of absolute quantitative measures that are extracted from in vivo MRI data. These can be used to describe biophysical characteristics and processes and thereby enhance the diagnostic power of qualitative, “weighted” imaging that is primarily used in the clinical setting. The feasibility, usefulness, and limitations of five qMRI techniques were investigated in different CNS pathologies (brain tumours, ischaemic stroke, migraine, brain/skull malformations) and in the description of normal brain maturation in infants and young children. The translation of new imaging methods from “bench to bedside” involves several steps, and the presented studies are located at different stages in this process.
Studies 1 and 2 are examples of a relatively early stage. At the time of publication, pH-weighted APT imaging had been tested preclinically and in smaller cohorts of patients, but not in acute stroke, where anaerobic glycolysis and tissue acidosis is highly prevalent. In study 1, it was postulated that APT imaging could be a novel approach to demonstrate oligaemia in hyperacute stroke, allowing a more detailed description of tissue at risk. For acceleration purposes, sequence parameters were optimised by using computer simulations and subsequently validated in healthy subjects. Ten acute stroke patients were included (7 < 4 hours, 3 < 24 hours after symptom onset). As expected, the APT effect was significantly decreased in ischaemic regions compared to normal white matter (p=0.03) and APT values tended to be lower in the final infarct volume (p=0.10). In study 2, APT imaging was moved to a different pathology, also characterised by hypoperfusion, tissue hypoxia, and anaerobic glycolysis. Here, the metabolic changes during the migraine aura of a patient with FHM were investigated for the first time using APT imaging. The patient developed clear tissue acidosis and blood flow disturbances in the absence of ischaemia in the affected cerebral hemisphere, possibly caused by CSD, i.e. the state of neuronal inhibition that is supposed to be the pathophysiological basis of migraine aura. The studies were not designed to provide a statistical conclusion, but to identify technical strengths and weaknesses of this imaging technique.
Study 6 also represents an early phase of clinical translation. Here, a new postprocessing approach was developed to achieve absolute metrics for the measurement of dynamic processes on CINE MRI, a time-resolved method to visualise moving structures in vivo, e.g. in cardiac, bowel, or foetal imaging. Usually movement is evaluated qualitatively and to date objective quantitative approaches are missing. In this study, a measuring method (voxel intensity distribution method, VIDM) for subtle movements was developed and applied in 27 children with Chiari and other brain/skull malformations, where cerebellar tissue herniates dynamically through the foramen magnum following CSF pulsatility. The degree of movement was compared using VIDM and visually derived, clinically accepted linear measurements on CINE sequences. In 85% of the patients, VIDM showed significantly more cerebellar displacement (p=0.002) compared to simple visual assessments, although this did not correlate with the clinical outcome parameters (hydrocephalus or syringomyelia; Pearson’s correlation coefficient -0.28; p=0.16). It is suggested that VIDM might be a valuable tool to detect and measure subtle dynamic processes in the CNS, but extracranial applications are also very likely.
Study 3 and 7 represent validation studies of methods that have been presented in clinical data before. In study 3, 2HG MRS was used in 35 patients suspected for cerebral gliomas to determine the IDH mutational status that today is an integral part of the WHO brain tumour classification system. For this study, a dedicated MRS sequence was used and the routine imaging protocol was extended by only 6 min. The sensitivity/specificity for determining the IDH mutational status was 89.5% and 81.3%, respectively. It could be concluded that 2HG MRS is an easily applicable supplement to standard imaging protocols that allows presurgical diagnostics and opens up for more detailed assessment during treatment.
In study 7, T1 maps were generated from clinical MRI data using the MP2RAGE sequence, a technique extensively applied in neuroscience, but little in the clinical setting. The technical parameters were adapted to find a balance between short acquisition times, high signal-to-noise, and reliable T1 values to quantify myelin maturation in 94 children up to the age of 6 years. The assessment of adequate myelination is a central part of paediatric imaging diagnostics, but is to date done by evaluating images qualitatively. The aim was to validate the MP2RAGE-based T1 mapping technique for the assessment of normal myelination, and data were compared to those of children with various CNS pathologies. Additionally, the diagnostic power of the MP2RAGE was pointed out for the qualitative assessment of regular myelination and brain pathologies.
The purpose of study 4 and 5 was to improve the diagnostic confidence of perfusion-weighted DCE maps. DCE is a well-established technique outside the CNS, but is used less in neuroimaging due to a number of technical issues. Here, postprocessing was addressed with the aim to reduce noise in the resultant parameter maps. Two curve-fitting methods, the Levenberg-Marquardt (LM) algorithm and a Baysian method (BM), were compared in digital phantoms and in 42 glioma patients applying two compartmental models (extended Toft’s, ETM, and 2-compartment- exchange model, 2CXM). The image quality was assessed with regard to tumour discrimination and overall impression of the images. Moreover, the diagnostic performance to differentiate high-grade from low-grade gliomas was investigated. The image quality of parameter maps generated by BM was significantly improved compared to LM (p<0.001), and the 2CXM- based maps were higher rated, regardless of the fitting method. The diagnostic performance to differentiate tumour grades was excellent for Ktrans and Vp (p<0.001). This was not affected by the fitting method for the leakage parameter Ktrans, whereas Vp was improved when using BM. These studies suggest that using BM to derive perfusion parameters from DCE data are superior to LM, hopefully leading to higher diagnostic confidence and acceptance in the clinical community.
Clinical imaging diagnostics benefits without doubt from the integration of quantitative information gained by qMRI, thereby increasing reproducibility and reliability and enabling the objective comparison to normative and patient databases. Each step of the clinical translation process is essential to show opportunities, identify areas of optimisation, and to reveal challenges and limitations. After further development APT imaging is today available on standard MRI platforms, and BM-based curve fitting of perfusion data has been implemented in postprocessing software programmes. T1 maps of normal myelination in children are made publicly available and may be a first step towards an automated tool to detect myelination disorders more efficiently.