Measuring chemical exchange saturation transfer exchange rates in the human brain using a particle swarm optimisation algorithm

Andrew J Carradus,Joe M P Bradley,Penny A Gowland,Olivier E Mougin

DOI: https://doi.org/10.1002/nbm.5001

Abstract:The z-spectrum contains many pools with different exchange rates and T2 values, which can make it difficult to interpret in vivo data and complicates the design of experiments aimed at providing sensitivity to one pool. This work aims to characterise the main pools observable with MRI at 7T in the human brain. To achieve this, we acquired z-spectra at multiple saturation powers in the human brain at 7T. We used simulations to optimise the use of particle swarm optimisation (PSO) to fit these data, validating this approach using further simulations and creatine phantoms. We then used the PSO to fit data from grey and white matter for the pool size, exchange rate, and T2 of five proton pools (magnetisation transfer, amides, amines, nuclear Overhauser enhancement NOE-3.5ppm and NOE-1.7ppm in addition to water). We then devised an approach for using PSO to fit z-spectra while limiting the computational burden, and we investigated the sensitivity of the fit to T2 and k for three overlapping pools. We used this to measure the exchange rate of creatine and to show that it varied with temperature, as expected. In the brain we measured a significantly larger pool size in white matter than in grey matter for the magnetisation transfer pool and the NOE-3.5ppm pool. For all other parameters we found no significant difference between grey and white matter. We showed that PSO can be used to fit z-spectra acquired at a range of B1 to provide information about peak position, amplitude, exchange rate, and T2 in vivo in the human brain. These data could provide more sensitivity to change in some clinical conditions and will also provide key information for further experimental design.