Quantitative chemical exchange saturation transfer imaging of nuclear overhauser effects in acute ischemic stroke

Purpose In chemical exchange saturation transfer imaging, saturation effects between −2 to −5 ppm (nuclear Overhauser effects, NOEs) have been shown to exhibit contrast in preclinical stroke models. Our previous work on NOEs in human stroke used an analysis model that combined NOEs and semisolid MT; however their combination might feasibly have reduced sensitivity to changes in NOEs. The aim of this study was to explore the information a 4‐pool Bloch–McConnell model provides about the NOE contribution in ischemic stroke, contrasting that with an intentionally approximate 3‐pool model. Methods MRI data from 12 patients presenting with ischemic stroke were retrospectively analyzed, as well as from six animals induced with an ischemic lesion. Two Bloch–McConnell models (4 pools, and a 3‐pool approximation) were compared for their ability to distinguish pathological tissue in acute stroke. The association of NOEs with pH was also explored, using pH phantoms that mimic the intracellular environment of naïve mouse brain. Results The 4‐pool measure of NOEs exhibited a different association with tissue outcome compared to 3‐pool approximation in the ischemic core and in tissue that underwent delayed infarction. In the ischemic core, the 4‐pool measure was elevated in patient white matter (1.20±0.20) and in animals (1.27±0.20). In the naïve brain pH phantoms, significant positive correlation between the NOE and pH was observed. Conclusion Associations of NOEs with tissue pathology were found using the 4‐pool metric that were not observed using the 3‐pool approximation. The 4‐pool model more adequately captured in vivo changes in NOEs and revealed trends depending on tissue pathology in stroke.

Supporting Information S1 I. OBTAINING CESTR* METRICS Figure 1 provides a high-level outline of how 4-pool NOER* is obtained, based on the original definition of APTR* for amide proton transfer [1]. The CEST data are fitted to a 4-pool model which accounts for a water pool, amides, NOEs, and semisolid MT effects. The output of the model fitting process is a number of model-fitted parameters corresponding to different variables in the physical model. In order to derive NOER*, which is an isolated quantitative measure of NOEs, only two modelfitted parameters are used: NOE pool relative concentration, NOE exchange rate. A 1-pool model is used to simulate water-only saturation, and a 2-pool model is used to simulate the contribution of the NOE pool. NOER* is defined as the difference between the two simulated line shapes at the NOE resonance frequency, -3.5 ppm. An equivalent procedure is used to calculate lMTR*, where semisolid pool parameters are used at the simulation stage; 3-pool NOER*, where initial model fitting is done using only 3 pools; and APTR*.

II. IMAGE PROCESSING
The BET tool in the FMRIB Software Library (FSL) package [3] was used to remove the skull and non-brain areas in all of the collected data. All of the imaging modalities were transferred to the T 1 space; within time point image registration was performed using FSL FLIRT, and, across time point image registration using FNIRT, both available in the FSL package [4]- [6]. The different CEST frequency offsets were motion-corrected using MCFLIRT, which applied linear co-registration to achieve alignment with the unsaturated acquisition [4]. MCFLIRT non-default options were: stages=4 which internally uses sinc interpolation for more robust optimisation, and sinc-final to apply the final output transformations using sinc interpolation. The quantified CEST effects were transformed to the T 1 image space using FLIRT. The T 1 structural data were segmented using FAST into cerebrospinal fluid, GM and WM.

III. ROI DEFINITIONS IN NATIVE SPACE
In stroke patients, infarct at presentation was defined using semi-automated delineation of ADC below an externally validated threshold of 620 × 10 −6 mm 2 /s [7]. Final infarct was manually defined preferentially on the 1 week FLAIR image, or, if not available, the b = 1000 DWI at 24 hours [8]. The mask representing perfusion deficit was generated using a threshold approach where voxels with a cerebral blood flow (CBF) threshold of less than 20 ml/100g/min were identified and clustered, and then used as a guide for manual delineation by an expert clinician (GH) [9]. The ROIs used in this study were: • Ischaemic core: within both presenting and final infarct definitions.
• Infarct growth: within the final infarct, but not within the presenting infarct.
• Oligaemia: tissue present in the perfusion deficit but not the final infarct. • Mirrored contralateral mask: a contralateral ROI was obtained for each patient by non-linearly registering the union of the pathological masks to standard MNI152 space, reflection in the sagittal plane, and transforming back to CEST space.
These ROI definitions are in keeping with those used in ref. [9] but have been updated to improve ROI fidelity with tissue fate [8]. A whole slice mask, a grey matter mask, and a white matter mask, were defined as follows for the healthy subjects and stroke patients. Grey matter and white matter masks were first generated from partial volume (PV) estimates using FSL FAST [10] on the presenting T 1 -weighted scan and the images were transformed to the resolution of the CEST images. Thresholds were applied to create healthy subject masks in the data space: • Whole slice mask: voxels with a GM ∪ WM PV threshold of 50%.
• GM mask: voxels with a GM PV threshold of 70%.
• WM mask: voxels with a WM PV threshold of 90%.
For the preclinical data, pathalogical ROIs representing ischaemic core, infarct growth, oligaemic, and contralateral tissue, were defined automatically based on the differences between the 1 h and 2 h post-MCAO scans, where the ADC and CBF maps were thresholded using the values defined in the study of ref. [9]. All brain voxels were included in the analyses of animal data (no PV threshold defined).