Savvopoulos, Fotios
(2020)
CEST MRI for the characterisation of human brain tumours.
PhD thesis, University of Nottingham.
Abstract
Chemical Exchange Saturation Transfer (CEST) is a novel MRI technique that amplifies the detection of otherwise undetectable compounds. Under certain physico-chemical and experimental conditions, CEST allows the detection of compounds containing amide, amine and hydroxyl proton groups, that are in chemical exchange with free water molecules. The main focus of this work is to gain an understanding of the underlying principles of CEST, both in-vitro and in-vivo.
After introducing essential MRI principles in chapter 1, chapter 2 contains a review of the CEST theory, optimisation techniques and applications of CEST, mainly in characterising brain tumours.
Chapter 3 studies the impact of experimental parameters in generating and quantifying the APT signal, in-silico. The parameters investigated were: the B_{1} irradiation power, the APT pool size, the saturation length and finally the SNR. For each of those parameters, the relationship between the parameter and APT signal was derived and its implications in optimising CEST experiments were discussed. For the model under investigation, it was found that the optimal parameters were: B_{1} = 1µT, APT pool size of 0.01 and saturation time of 3 seconds, at most. Furthermore, it is shown that the presence of noise diminishes the ability to quantify the APT effect accurately and that this loss varies linearly with SNR. In conclusion, the model studied here should be imaged with an SNR of ~70 so that the percentage error and relative accuracy in the measurements is ~10% and 0.8, respectively.
Chapter 4 contains a feasibility comparison of measuring extracellular pH with CEST, by using three clinically approved iodinated contrast agents, at 7T, in-vitro. The contrast agents tested were: Iodixanol (brand name - Visipaque, GE Healthcare), Iopamidol (brand name - Niopam, Bracco UK Ltd) and Iopromide (brand name - Ultravist, Bayer Healthcare). Phantoms of the agents, in phosphate buffer and at different pH levels were prepared and subsequently scanned using a 2D gradient echo CEST sequence. The CEST effect was calculated by asymmetry analysis and pH maps were derived on a pixel by pixel basis. In addition, calibration curves were determined by fitting the in-vitro CEST signal of all three agents, against pH. Under the experimental conditions used, Niopam and Ultravist showed two, while Visipaque showed one, CEST exchange peaks. Consequently, Visipaque was found unsuitable in determining pH, by ratiometric CEST analysis. On the contrary, Niopam and Ultravist showed great potential in determining extracellular pH, as the ratiometric analysis produced linear calibration curves, for both agents. Moreover, Ultravist showed superior linearity to Niopam (R^{2} = 0.99, R^{2} = 0.88). Next steps could involve, measuring the CEST effects, of both Niopam and Ultravist, in-vivo. First, on a pre-clinical model and second, in humans. The latter will be at 3T, instead of 7T. Patients could be immediately scanned after routine CT procedures, where either of these contrast agents is used. Before assessing the pH, calibration curves, of both agents, will be derived at 3T, in-vitro, with experiments similar to those reported at 7T.
Chapter 5 contains the comparison between a modified single slice and volumetric CEST gradient echo sequences, in generating CEST contrast, in a series of phantoms, at 3T. The single slice sequence was successfully modified by myself, whereas the volumetric sequence was prepared by a team from Institute of Psychiatry, Psychology and Neuroscience in King’s College, London. Besides allowing for different spatial excitation, the two sequences also differ in the pre-saturation pulse type. The latter is a Fermi pulse, in the 2D sequence and a Gaussian pulse, in the 3D sequence. A common limitation of both sequences is the maximum allowed pulse duration, which is in the order of milli-seconds. However, CEST effects were successfully measured in phantoms of the iodinated contrast agent, Ultravist, creatine and glutamate, under physiological conditions (pH =7). Asymmetry analysis was used to quantify the CEST effect, in all cases. Despite post-processing corrections for B_{0} inhomogeneities, some datasets were noisy. This is perhaps due to small sample size and inefficient shimming. The Gaussian pulse of the 3D sequence was found to produce a 2-fold increased CEST effect in Ultravist and a 7-fold CEST effect increase in creatine, as compared to the Fermi pulse of the 2D sequence. Furthermore, increasing the duty-cycle of the 2D sequence did not result in greater CEST effect magnitude. Analysis of the main field homogeneity, after linear and high order shimming, showed that the latter results in frequency shifts within 0.5 ppm, whereas the shifts, after linear shimming, extended beyond that range. As the CEST effect is highly sensitive on frequency shifts, a high order shimming is more beneficial, especially when characterising CEST effects in-vivo. Future work could include studying the pH dependence of Ultravist and Niopam, in-vitro. Even if only one CEST peak, instead of the two found at 7T, can be measured at 3T, calibration curves can still be derived by implementing an RF power based ratiometric method. Subsequent translation, of measuring pH, in patients that have received iodinated contrast injection, will require further ethical approval. Besides measuring pH, typical CEST acquisitions could be optimised in more complex phantoms, in-vitro. For example, the CEST effects of three-pool model phantoms (agar gel + creatine or ammonium chloride) could be tested, using the 3D gradient echo CEST sequence. Existing analysis tools could be updated to allow for Lorentzian fitting separation of co-existing CEST effects. New tools, allowing fitting of Bloch equations to CEST data, should also be developed. In addition, besides B_{0} inhomogeneity corrections, the impact of RF inhomogeneity effects, on CEST signals, should be studied. The latter correction might be crucial in detecting weaker CEST effects at 3T. It would also be interesting to measure the relaxation compensated CEST effect at 3T. That would be possible if our sequences allow for a pre-saturation time in the order of seconds (t_{sat}>>T_{1w}). Ultimately, the efficacy of the 3D CEST sequence could be tested in healthy volunteers and finally in brain tumour patients, after acquisition of appropriate ethical approval.
Chapter 6 contains the details of a CEST study, of cerebral gliomas at 7T, where the APT and NOE signals, between low and high grade brain tumours and normal white matter, are compared. In addition, the correlation between, APT and NOE signals with apparent diffusion coefficient (ADC), is also investigated. More specifically, a 3D CEST MRI scan, targeting amide and aliphatic protons, was acquired, at 7T, in seven untreated glioma patients, who had previously undergone routine clinical 3T MRI. APT and NOE maps were produced by Lorentzian fitting of the CEST data. Significantly higher tumour APT signal was found in both low grades and glioblastomas, as compared to healthy white matter. In addition, the APT signal, was 3 to 4-fold elevated in glioblastoma versus low grades. On the contrary, significantly lower tumour NOE was found in both low grades and glioblastomas, as compared normal tissue, with the latter being ~7-fold decreased in GBMs. Further, investigation of the correlation of the APT and NOE, tumour and control signals, with ADC, did not show any particular trend, with the exception of the NOE - ADC signals in controls, which showed a weak, negative correlation. Further analysis of our data could include asymmetry analysis, as a comparison to Lorentzian fitting, in cerebral gliomas. Also, quantification of the conventional magnetisation transfer (MT), along with APT and NOE effects, in oedema, necrotic and haemorrhagic regions, that often accompany gliomas.
Chapter 7 contains an overall discussion of the work on this thesis, how it relates to the literature and presents some ideas on what should be done next.
Actions (Archive Staff Only)
 |
Edit View |
Tools
Tools
![[img]](/style/images/fileicons/application_pdf.png)