Amide Proton Transfer Imaging in Stroke

Abstract

Amide proton transfer (APT) imaging, a variant of chemical exchange saturation transfer (CEST) MRI, has shown promise in detecting ischemic tissue acidosis following impaired aerobic metabolism in animal models and in human stroke patients due to the sensitivity of the amide proton exchange rate to changes in pH within the physiological range. Recent studies have demonstrated the possibility of using APT-MRI to detect acidosis of the ischemic penumbra enabling the assessment of stroke severity and risk of progression, monitoring of treatment progress, and prognostication of clinical outcome. This paper reviews current APT imaging methods actively used in ischemic stroke research and explores the clinical aspects of ischemic stroke and future applications for these methods.

1. Introduction

Ischemic stroke occurs when there is hypoperfusion of an area of the brain, which is typically caused by a vascular occlusion. Distal to the occlusion, the affected vascular territory is often characterized by a smaller ischemic core that suffers infarction, which is surrounded by a larger area of marginally perfused tissue, typically referred to as the ischemic penumbra. Different types of penumbras have been defined over the years. Classically, neurons within the penumbra are unable to maintain a resting potential and become dysfunctional, but are not yet irreversibly damaged, such as in cases where the blood flow supplied by collateral arteries is adequate to prevent cellular membrane breakdown and mitochondrial death^1–4. However, unless rapidly reversed, ongoing ischemia may result in metabolic derangements and neuronal death. Ultimately, the infarct core may grow and progressively replace the ischemic penumbra. Thus, the major goal of intervention in acute ischemic stroke is to salvage the ischemic penumbra^1,5–8. Although a number of sensitive imaging methods have been developed to evaluate the physiological state of brain tissue, the current standard of care in most hospitals is time-dependent, which limits imaging to a computed tomography (CT) scan to exclude hemorrhage. Unfortunately, this approach ignores the variability that each subject shows and rejects the possibility of treatment for about 90% of stroke patients whose clinical course does not fit a fixed time window^9–12. This under-use of imaging methods due to time constraints is unfortunate because the identification of a true ischemic penumbra could lead to more rational treatment decisions, as well as active successful outcomes outside conventional time windows. The rate of change in the size of the ischemic penumbra varies from patient to patient, depending on collateral circulation. Two randomized “late window” mechanical thrombectomy trials showed that patients with salvageable tissues had less disability and were more likely to be functionally independent at 90 days after initiation of endovascular thrombectomy between six and 16 hours or up to 24 hours after the patient was last known to be well^13–15. Similarly, a late time window thrombolysis trial using perfusion imaging showed an opportunity to personalize treatment selection and extent the time window for treatments is a subset of patients¹⁶. These trials suggest that, as long as patients are robustly phenotyped for the presence of an ischemic penumbra, treatment decisions can be personalized, rather than being based on arbitrary time windows, expanding the treatable population. This situation strongly emphasizes the urgent need for improved and accessible imaging techniques to readily define the salvageable penumbra from the irreversibly damaged infarct core at the level of the individual.

Positron emission tomography (PET) first demonstrated the ability to image the ischemic penumbra in vivo, defined by quantitative PET parameters, such as cerebral blood flow (CBF), and is considered the ‘gold standard’ in the early evaluation of acute stroke pathophysiology^2,17–19. However, the use of quantitative PET is unfortunately limited by a lack of availability in emergency settings, the need for arterial catheterization, the administration of radioactive tracers, and long scan times, making it impractical in the hyperacute stroke setting with a time-limited opportunity for intervention. Clinically, CT and CT angiography (CTA) are the primary imaging modalities for the emergency assessment of suspected acute stroke, mainly to identify vascular occlusion, exclude hemorrhagic stroke prior to thrombolytic treatment, assess collateral circulation in the territory of the occluded artery, and assess irreversibly damaged tissue during very early ischemia, albeit in a less detailed way than is possible using PET or magnetic resonance imaging (MRI)^20–22. While such approaches have been partially successful due to a heavy reliance on the speed of treatment, they currently do not provide information about changes in physiological tissue status related to pH, metabolism, and acute cell depolarization, all of which may provide more sensitive and specific markers of tissue viability. Although CT perfusion (CTP) is promising for the identification of a perfusion-based ischemic penumbra, measures of perfusion parameters are not directly translatable to events at the cellular level, and thus, can provide only an indirect estimate of whether tissue is viable.

During the past few decades, MRI techniques have gained a small role in the acute management of stroke²³. Typically, diffusion-weighted imaging (DWI) is used to visualize early membrane depolarization through changes in the local diffusion properties of water in the ischemic lesion, while perfusion-weighted imaging (PWI) provides quantitative information about abnormal cerebral blood flow or volume. The DWI/PWI spatial mismatch has been used to identify the presence of a flow-based ischemic penumbra and, over the past two decades, been tested as a selection marker for thrombolysis^5,24,25. Unfortunately, the use of the DWI/PWI mismatch concept has proven to be limited in routine clinical application due to variable sensitivity, specificity, and high false-negative rates^26–29. One of the main issues is that the mismatch area often overestimates the size of the ischemic penumbra and includes regions of benign oligemia. This is particularly true in the setting of chronic vascular occlusions that may appear as penumbra with DWI/PWI, but which are stable at the physiologic level. Preclinical animal studies indicate that a more appropriate ischemic penumbra would be the region in which oxidative metabolism is impaired, but no diffusion changes have occurred^30,31, and reviews by leading experts have suggested that this should be the required criterion^2,18,19,30. Attempts to better specify such a true penumbra through the addition of mean transit time cutoff criteria on PWI are promising, but not yet conclusive^32–36. Alternatively, previous clinical studies (e.g., Wake-Up Stroke trial) demonstrated that DWI/fluid-attenuated inversion recovery (FLAIR) mismatch can be used as a surrogate marker to identify patients eligible for intravenous thrombolysis^37,38. However, the concept of mismatch of DWI/FLAIR is thought to indirectly estimate the onset time in acute ischemic stroke patients, not specifically penumbra identification and quantification.

Previous papers reviewed the basic chemical exchange saturation transfer (CEST) or amide proton transfer (APT) principles, recent technical developments, and current and emerging clinical applications including stroke imaging^39–42. Herein, we highlight current APT imaging technologies actively used in ischemic stroke research, and important limitations and challenges of current APT-based penumbra imaging, and end with possible solutions and future directions.

2. Brain tissue acidosis in ischemia

A simplified overview of the approximate CBF thresholds^2,18,19,30 and their relationship to MR parameters⁴¹ is shown in Fig. 1. The original concept of the ischemic penumbra was defined by the thresholds of: 1) electrical failure, encompassing functionally impaired, but structurally intact neurons, which corresponds to an approximate CBF range of 8~18 ml/100g/min; and 2) membrane failure, encompassing functionally and structurally impaired neurons (likely irreversible damage, i.e., infarction), which corresponds to a CBF of <~8 ml/100g/min. The intermediate zone (orange-colored area in Fig. 1a) is referred to as the classic ischemic penumbra, a region with neuronal dysfunction, but no cellular energy failure. Eventually, below a CBF threshold of about 8 ml/100g/min, anoxic depolarization occurs, leading to intracellular accumulation of sodium, calcium, and water (cytotoxic edema). Although reversible changes in water shifts have been measured after membrane failure^43,44, as confirmed by recent clinical imaging results, anoxic depolarization mostly leads to infarction, except in the case of very short ischemic durations. Thus, further investigation of the DWI-based core with physiologic imaging may better define the inner edge of the ischemic penumbra as well.

Figure. 1.

Simplified depiction of the relationship between (a) CBF-based ischemic neurological and (b) MRI parameter thresholds. Thresholds are approximate (PET-based normal gray matter CBF ~50 ml/100g/min) and might shift to a higher CBF with a longer ischemic duration. The metabolic penumbra is indicated in the yellow shade. At the bottom, zones of perfusion-DWI and pH-DWI mismatch are shown, with the difference being benign oligemia (Reproduced, in part, with permission from Leigh et al. J Cereb Blood Flow Metab 2017; 38: 1500–1516).

The current understanding of a true ischemic penumbra is based on the concept that tissue can be impaired because of a deficit in oxidative metabolism, but can be still viable. Such tissue surrounds and is contiguous with an area of irreversible cerebral infarction^1,2. Acute cerebral ischemia causes a shift to anaerobic glycolysis, resulting in the accumulation of lactic acid and a concomitant decrease in intracellular pH⁴⁵. Therefore, tissue acidosis is the earliest sign that tissue is at risk. If ischemic acidosis can be imaged, salvageable tissue at risk of infarction could be visualized by subtracting the extent of the diffusion deficit (cellular depolarization) from the pH deficit extent (anaerobic metabolism).

3. APT preclinical research

Based on the CEST-MRI principle, it is possible to detect the MRI signal of tissue mobile proteins and peptides, such as those in the cytoplasm, through chemical exchange between exchangeable amide protons (-NH) in the protein peptide bonds and tissue water^39,46–48. While the effect of mM-concentrations of amide protons on the water resonance (~110 M protons) is not detectable, the cumulative effect of repeated saturation and chemical exchange, in which saturated amide protons move to water and are replaced by unsaturated water protons from the large water pool that are then saturated, allows sensitivity enhancements of 100–1000-fold. Thus, specific molecular information can be obtained indirectly through the bulk water signal.

Amide protons in the backbone of solute proteins resonate around +3.5 ppm downfield of the water signal, and interact with water protons through chemical exchange⁴⁹. The exchange rate of these protons is highly pH-dependent because their exchange rate is base-catalyzed in the physiological pH range, and typically decreases ten-fold per unit of pH drop⁵⁰. Therefore, the APT signal is reduced in the ischemic area since intracellular pH is lower in the ischemic lesion than in normal brain tissue. Typically, pH-sensitive APT effects can be detected by acquiring a so-called Z-spectrum, in which the ratio of water signal intensity with saturation (S_sat) and without saturation (S₀) is plotted as a function of saturation frequency. The APT signal intensity at 3.5 ppm from water is significantly decreased after global ischemia⁴⁶, as shown in Fig. 2. Notably, the reduced APT effect can be clearly seen in the MT ratio (MTR = 1 – S_sat/S₀) spectrum, as shown in rat middle cerebral artery occlusion (MCAO) models at two hours after occlusion. The most commonly used metric with which to quantify the APT effect is defined as:

$$MT{R}_{\mathit{\text{asym}}}\left(3.5\text{ppm}\right)=MTR\left(+3.5ppm\right)-MTR\left(-3.5ppm\right)$$ [1]

This asymmetry measure corrects for the influence of direct water saturation and other symmetric components not associated with the APT effect. Reduced MTR_asym(3.5ppm) (red arrow in Fig. 2d) is attributable to a pH reduction, which causes a slower exchange. Although these APT effects are small (a few percent on the water signal), they correspond to a detection sensitivity of molar concentration. The MTR_asym at 3.5 ppm has multiple contributions:

$$MT{R}_{\mathit{\text{asym}}}\left(3.5\text{ppm}\right)=\mathit{\text{APTR}}+MT{R^{\prime }}_{\mathit{\text{asym}}}\left(3.5ppm\right)$$ [2]

$$\mathit{\text{APTR}}= k_{sw}\cdot \left(\frac{\left[\mathit{\text{amide proton}}\right]}{\left[\mathit{\text{water proton}}\right]}\right)\cdot T_1\cdot \left(1-e^{-t_{sat}/T_1}\right)$$ [3]

where APTR is the proton transfer ratio associated with amide protons, and an additional term, MTR’_asym(3.5ppm), accounts for asymmetric effects on the MTR spectrum not associated with APT. These effects can include intramolecular and intermolecular nuclear Overhauser enhancement (NOE) effects of aliphatic protons of mobile cellular macromolecules in tissue, including semisolid MT asymmetry. Due to these contributions, APT signals quantified by the MTR asymmetry analysis are called APT-weighted (APTw) signals⁴⁰. Eq. [3] is based on the simplified assumption of a two-pool exchange model (free bulk water and amide proton pools), in which APTR is proportional to the ratio of amide proton and water proton concentration (square brackets, [.]), and the exchange rate (k_sw) from the amide proton pool to the water proton pool, which is related to tissue pH. Zhou et al. derived the calibration curve of pH using the equation, APTR = 5.73 × 10^pH−9.4, by comparing the in vivo and postmortem MTR_asym(3.5ppm) signals and using ³¹P MRS for intracellular pH assessment⁴⁶. With the calibration formula, an absolute pH map could be generated, as shown in Fig. 2a, which correctly outlined the ischemic area in the caudate nucleus, a region commonly affected by infarction following MCAO, where ischemia was confirmed by histology acquired eight hours later (Fig. 2b)⁴⁶. However, the APTR-pH equation was derived with several simplified assumptions, e.g., negligible changes in the amide proton content, water content, water relaxation time, and the effect of temperature on the exchange rate.

Figure. 2.

(a) A pH map of a rat MCAO model at two hours after occlusion. The average ischemic pH was 6.52 ± 0.32. (b) Triphenyl tetrazolium chloride (TTC) histology image. (c) Z-spectrum and (d) MTR asymmetry spectrum of the rat brain (n=7). The Z-spectrum is a measurement of normalized water saturation (S_sat/S₀) as a function of RF saturation frequency. Magnetization transfer ratio (MTR=1- S_sat/S₀) asymmetric analysis is used to measure the APT signal by subtracting the MTR at −3.5 ppm, with respect to water, from that at +3.5 ppm (Reproduced, in part, with permission from Zhou et al. Nature Med 2003; 9: 1085–1090).

Following the initial study, many APT-based pH imaging studies have demonstrated the possibility of using APT-MRI to detect an acidosis-based ischemic penumbra in rat models of permanent MCAO. Importantly, Sun et al.⁵¹ showed that a pH deficit based on APTw MRI in the first 3.5-hrs correctly predicted the infarct 24-hrs later, as shown in Fig. 3. Although two rats (Fig. 3c) showed a different degree of mismatch between diffusion, pH, and perfusion deficits, only pH deficits predicted well the evolution of infarction. However, the diffusion deficit underestimated the final lesion and the perfusion deficit overestimated it. Interestingly, several animals showed negligible apparent diffusion coefficient (ADC) effects in the hyperacute state, despite the presence of perfusion and pH effects, suggesting that pH changes can occur before diffusion changes⁵¹. Many preclinical studies have consistently shown the salvageable pH-defined penumbra, supporting the hypothesis that changes in pH are a key target with which to identify the ischemic penumbra^46,51–56. These promising results in the preclinical setting stimulated the clinical implementation of APT-based pH imaging for stroke imaging.

Figure. 3.

(a) Group analysis of ischemic volume evolution (n=18) with a PWI/DWI mismatch, comparing areas of pH change and diffusion changes as a fraction of the perfusion deficit region. (b) Parcellation of the ischemic area in terms of three zones: a DWI deficit that will most likely proceed to infarction; a pH/DWI mismatch region at risk for infarction; and a PWI/pH mismatch not at risk. (c) Evolution of a pH deficit (orange) and a diffusion deficit (black), with regard to the perfusion deficit (purple), as a function of time, post-MCAO, in two rats. The T₂ image at 24-hrs (the right-most column in c) shows final infarction area predicted well by diffusion and pH deficit regions. (Reproduced, in part, with permission from Sun et al. J Cereb Blood Flow Metab 2007; 27: 1129–1136).

4. APT clinical research

In recent years, many groups have developed pH-sensitive APT imaging sequences on clinical scanners to improve the translation of the technique to clinical practice. The first human study was reported by Zhao et al.⁵⁷, revealing that the APTw effect was hypointense compared to contralateral normal-appearing tissue, consistent with the findings in the preclinical studies. The APTw contrast of approximately 1% was observed when a radio frequency (RF) saturation power of 2 μT was applied. However, the observed signal might be still contaminated by other confounding factors, such as changes in amide proton concentrations, water relaxation times, upfield NOE signals, and conventional MT asymmetry effects during the acute and subacute stages (average 4.3 ± 2.5 days after the onset of the stroke), even though the MTR asymmetry analysis was performed to remove the contribution of symmetric MT and direct water saturation effects. Later, Tee et al.⁵⁸ compared different APT imaging analysis methods (conventional MTR asymmetry vs. Bayesian model-based fitting) and evaluated their pH sensitivity using hyperacute stroke patients (median, 2 hrs 59 min after the onset of the stroke). Acidic pH is generally assumed to be the dominant contributor to APT signal contrast at the hyperacute stage because the changes in mobile protein and peptide content, water relaxation times, and MTR asymmetry are more likely to negligible early in the evolution of ischemic stroke, as described in a previous animal study⁴⁶. Although the APT effect measured by the both analysis methods was consistently lower in the infarct core (diffusion deficit) than in contralateral normal-appearing tissue, the model-based fitting approach provided a higher APT stroke contrast-to-noise ratio and a lower variation in gray and white matter compared to the conventional asymmetry analysis, probably due to the inherent robustness of model fitting to noisy data (with outliers)⁵⁹. These early stroke patient studies showed that APT imaging in stroke patients was feasible on a 3T clinical-field-strength scanner.

The ability to identify potential areas of reversible and salvageable brain tissue after ischemic stroke helps to identify stoke patients for intravenous, intra-arterial, or mechanical reperfusion treatment at various therapeutic windows. Harston et al.⁶⁰ demonstrated the potential clinical use of pH-sensitive APT-MRI in hyperacute ischemic stroke by assessing the relationship between intracellular pH and final tissue outcome in data from a prospective cohort of 12 patients with hyperacute ischemic stroke (median, 2 hrs 59 min after the onset of the stroke). Importantly, within the perfusion deficit lesion, the region identified as the ischemic core (tissue present in both diffusion deficit and final FLAIR infarct) was found to have a lower APT signal, consistent with a lower pH, than the infarct growth area (ischemic penumbra, tissue present in the final FLAIR infarct, but not in the diffusion deficit). In addition, the tissue identified as benign oligemia (tissue present in the perfusion deficit, but not the final FLAIR infarct) had a significantly higher APT signal, consistent with a higher pH, than either the ischemic core or infarct growth, as shown in Fig. 4. These findings suggest that the addition of pH-sensitive APT imaging to the diffusion and perfusion MRI protocol could better visualize an ischemic penumbra, thus improving predictions of final infarct size and outcome.

Figure. 4.

(a) Quantified APT effect in the ischemic core, infarct growth, and oligemia region of a representative hyperacute stroke patient. (b) The mean relative APTR* (relative to the contralateral normal brain tissue) shows significant differences between the different regions of interest, where the error bars represent 95% confidence intervals and **** P < 0.0001. (Reproduced, in part, with permission from Harston et al. Brain 2015; 138: 36–42).

The APTw image contrast of the ischemic acidosis penumbra is small (~0.5%) compared to the inherent tissue signal, which limits detection sensitivity to pH. APT signals calculated by MTR asymmetry are further reduced due to the presence of NOE signals at roughly −3.5 ppm upfield from water. To separate the pH-sensitive APT effect from NOE signals, Heo et al.⁶¹ performed a study in 30 ischemic stroke patients (< 7 hrs after the onset of the stroke) using the so-called extrapolated semisolid magnetization transfer reference (EMR) approach^62,63. The NOE-free APT (APT^#) signal contrast between ischemic lesions and normal tissue was substantially increased—nearly three times larger than that based on conventional MTR asymmetry analysis—as shown in Fig. 5, thus allowing the more reliable delineation of an acidotic ischemic penumbra. The patient study⁶¹ showed that pH deficits were smaller than perfusion deficits and equal to or larger than diffusion deficits (Fig. 6). Similar to the result from Harston et al.⁶⁰, the APT deficit may reflect the area of the expected final infarction without reperfusion, and therefore, the acidosis-based penumbra, in line with suggestions by Heiss, Hossmann, and Baron^2,18,19,30. A further finding was that hyperintense APT signals were observed in hemorrhagic transformation (HT) regions as a result of abundant mobile protein and peptide contents in the blood (Fig. 7), which is in agreement with the results of a previous animal study⁵⁴, while the hypointense APT signals in ischemic lesions were primarily attributable to local tissue acidosis. These initial clinical data suggest that APT imaging is feasible for acute stroke patients and may provide extra information about the infarct in ischemic strokes, by virtue of pH changes.

Figure. 5.

Conventional MR angiography, DWI, APTw, APT^#, and NOE^# images from an acute stroke patient (<7 h from symptom onset) with a right MCA occlusion (magenta arrows). Two regions of interest, an ischemic lesion that was DWI-hyperintense (red dotted line on DWI) and a contralateral normal tissue (white dotted line on DWI) were drawn. While APTw-MRI for pH analysis was confounded by upfield NOE effects of mobile proteins and peptides, the NOE-free APT^# contrast between normal and ischemic lesions was substantially increased, nearly three times more than that based on MTR_asym analysis. The APT^# signal contrast (−1.45 ± 0.4%) was significantly larger than the APTw signal contrast (−0.39 ± 0.5%, p<0.001) (Reproduced, in part, with permission from Heo et al. Magn Reson Med 2017; 78: 871–880).

Figure. 6.

Comparison of diffusion, pH, and perfusion deficits and quantitative pH/diffusion and perfusion/pH scatterplots in an acute stroke patient. Of the 13 patients scanned at the acute phase (< 7-hr), three patients showed an acidosis-based ischemic penumbra. The distributions of the diffusion deficit area (infarct core (red)), pH-diffusion mismatch (acidosis-based penumbra (green)), and perfusion-pH mismatch (acidosis-based benign oligemia (blue)) were markedly different from those of the contralateral normal tissue (black) (Reproduced, in part, with permission from Heo et al. Magn Reson Med 2017; 78: 871–880).

Figure. 7.

Serial multimodality MR images of a representative acute stroke patient with a left MCA occlusion (yellow arrows) at three time points. The DWI and ADC showed large, acute ischemic areas caused by cytotoxic edema. In addition, perfusion-based relative bolus transit time (rBTT) and time to peak (TTP) showed obvious hypoperfusion in a slightly larger region than the diffusion abnormality. The high APT signal intensities observed at one week can be attributed to a hemorrhage (white arrows) caused by abundant mobile proteins and peptides in the blood. The patient presented outside of a treatment time window, did not received an acute intervention, and subsequently suffered a spontaneous symptomatic hemorrhagic transformation of the ischemic infarct (Reproduced, in part, with permission from Heo et al. Magn Reson Med 2017; 78: 871–880).

After the initial successes of pH-sensitive APT imaging in stroke patients, further studies have followed, such as those investigating dynamic APT signal changes in ischemic tissue at various clinical stages of stroke and correlating APT signal intensities with clinical outcome. Song et al.⁶⁴ evaluated the dynamic pH changes at the hyperacute, acute, early subacute, and late subacute stages using APTw MRI. Compared to the contralateral normal-appearing white matter, APTw signals were significantly lower in ischemic tissue for all four stages. In that study, the APTw signals of the ischemic lesion were lowest at the hyperacute stage (onset time < 6 hrs), but gradually increased from the onset time, suggesting that the tissue acidification in the ischemic tissue may decrease with time from stoke onset. However, the pH contribution on the APTw signals might also vary with time and with changes in upfield NOE signals and water relaxation times due to the development of cytotoxic and vasogenic edema. All these different contributions are highly dependent on the RF saturation parameters^65,66 used in the CEST acquisition, and thus, it has not been possible to resolve these different contributions across the studies performed in patients, to date.

Lin et al.⁶⁷ and Momosaka et al.⁶⁸ investigated the correlation between APTw signals and clinical stroke severity using the National Institutes of Health Stroke Scale (NIHSS), and the prognosis with the 90-day modified Rankin Scale (mRS) score. Both studies showed that the APTw signal was significantly inversely correlated with the NIHSS score. More importantly, the APTw signal within an infarct lesion was significantly lower in the poor prognosis group (mRS score > 2) than in the good prognosis group (mRS score < 2). Furthermore, Yu et al.⁶⁹ assessed the therapeutic efficacy of supportive treatments in ischemic stroke patients using APTw MRI. Forty-three ischemic patients (median, three days after the onset of the stroke) who were not suitable for thrombolysis, and underwent only supportive treatment that included antiplatelet and anticoagulation therapy and free radical scavenging, were imaged by APTw MRI prior to treatment and underwent follow-up scans after treatment. The APTw signal contrast between ischemic lesions and contralateral normal-appearing white matter tissues had significant correlations with the NIHSS at arrival. Most of the ischemic stroke patients (24 of 26 patients) showed progressively increased APTw signal intensities in the ischemic lesion over all time points post-treatment, with improvements in clinical symptoms, and thus, the NIHSS. Although the clinical results should be interpreted with caution due to the small sample size of the studies and the inclusion of heterogenous populations, the findings described above raise the intriguing possibility that pH-sensitive APT imaging could be used to assess stroke severity, the therapeutic effect of ischemic stroke treatments, and in the prediction of long-term clinical outcome for ischemic stroke patients.

5. Discussion

To date, clinical studies of ischemic stroke appear to agree with preclinical work, which demonstrated that the APT signal intensity in ischemic tissue is lower compared to normal, healthy tissue in early stroke imaging^41,42. The reduced APT signal has been widely hypothesized to be the result of tissue acidosis due to lactic acid accumulation by enhancing anaerobic glycolysis. This acidosis hypothesis seems most likely to be valid in the early hyperacute phase where there has been less time for additional tissue changes to occur that might affect other contributors to the APT signal. Various clinical studies have included ischemic stroke patients who were scanned days after the onset of the stroke^67,69. These studies found that the APT signal within the ischemic lesion gradually increased as time progressed from symptom onset, and yet, the ischemic region still showed a significant difference when compared to the normal tissue area. One of these studies even reported that the APT signal in the ischemic lesion became hyperintense when imaged one week after the onset of the stroke. It is currently premature to ascribe a pH change as the main contributing factor to the decreased or increased APT signal observed in the days immediately after stroke. A cascade of physiological changes occur after stroke; thus, relating the APT signal change to a pH drop or to normalization may not be warranted, especially in the late stage of stroke^70–75. Many physiological changes, such as protein disintegration, edema, and inflammation, have been observed after stroke, which likely influence APT signals. As a result, more work is required to investigate the exact source(s) or pathway(s) that led to the current observations.

One of the main limitations faced by APT clinical studies is the recruitment of stroke patients. This is especially challenging in the hyperacute setting where patients can find it difficult to tolerate research MRI scans and early evaluation and treatment are time-critical. In addition to that, follow-up imaging is also difficult to acquire because one in eight strokes are fatal within the first 30 days and almost two-thirds of strokes lead to disability. As a result, the follow up recruitment to confirm APT findings is often incomplete or leads to the use of inconsistent final infarct definitions. This particularly affects studies that target the investigation of APT effect across different stroke time points.

In addition, there are several technical hurdles in the clinical translation of APT-MRI. These include a relatively long scan time because of the use of a long RF saturation pulse per frequency offset, and a series of saturation frequencies that must be acquired to form a Z-spectrum. The American Heart Association (AHA) recommends that imaging occur within 25 mins of emergency department (ED) arrival and that interpretation of imaging should occur within 45 mins of ED arrival. There is, thus, only a 20-min time window in which to acquire and interpret the images^76–78. A long APT imaging can lead to more motion artifacts in restless patients, which can produce unusable images, wasting precious minutes. For improvement in acquisition speed, APT imaging could be combined with advanced fast imaging technologies, including reduced k-space sampling with parallel imaging^79,80, compressed sensing^81–83 and MR fingerprinting^84–89. Importantly, quick image reconstruction, in addition to imaging acquisition, is essential for clinical translation. The development of real-time, pH-diffusion mismatch analysis software for unsupervised and fully automated processing on scanners is, therefore, highly desired. A faster APT scan would not only help to improve patient comfort and compliance, but also could be used to improve image quality. Furthermore, higher spatial resolution and volumetric 3D imaging would be valuable for the more precise depiction of an acidosis-based penumbra.

There remains some uncertainty regarding pH specificity of APT MRI in ischemic lesions. APT contrast is believed to be predominantly driven by ischemic acidosis during the very early stages of ischemia where water content and relaxation weighted images may not show contrast. However, as previous studies indicate^58,90, vasogenic edema begins to develop during the first 24 to 48 hrs after the onset of a stroke and alters the magnetic properties of water and causes a change in the diffusivity in the ischemic area, which affects water content, T₁, and T₂ relaxation times, all of which can also influence APT contrast. Another possible issue is the possible contribution of other CEST signals to APT contrast, such as signals from faster exchanging protons, such as on amine and guanidinium. However, CEST effects are dominated by the slowly exchanging amide protons at lower RF saturation powers, with faster exchanging protons contributing at higher RF saturation powers. Therefore, employing a low-power RF saturation may be sufficient to avoid contributions from the rapidly exchangeable protons that occur at the APT frequency. CEST specificity is expected to increase at higher magnetic field strength (7T) due to high spectral resolution. However, ultrahigh field 7T machines are not accessible for most acute clinical applications and MRI as a first-line imaging modality is currently still limited to standard (1.5T) and high field (3T) strength MRI. New APT imaging techniques have been developed to correct water relaxation and multiple overlapping CEST effects, improving pH specificity, but have not yet been demonstrated in stroke patient studies^91–94. Other potential factors affecting APT signals are static magnetic field (B₀) and RF field (B₁) inhomogeneities, which cause a shift in the water resonance frequency and spatial variation in labeling efficiency, respectively. Several correction methods have been developed to mitigate the effect of the field inhomogeneities on APT images^95,96.

A standardized APT imaging protocol is necessary for the proper design of prospective, population-based studies of acute stroke, which includes RF saturation parameters, imaging acquisition schemes, and image analysis methods. Although the first clinical study by Zhao et al.⁵⁷ aimed to propose an optimal RF saturation power parameter for APT imaging of stroke with limited hardware (e.g., RF amplifiers and transmitter/receiver RF coils) performance, further investigations of the standardization of scan parameters and imaging sequences to enhance the pH specificity and sensitivity are needed. In addition, the post-processing methods used for APT-MRI are not standardized. Several previous studies demonstrated that various analysis methods can lead to variation in APT signal intensities or contrast between ischemic and normal tissues^58,61,97, as shown in Fig. 8⁹⁸. Therefore, standardization of APT image analysis is necessary for the incorporation of this analysis into clinical studies and the promotion of its use in clinical trials and clinical practice.

Figure. 8.

Representative results from an animal (top two rows) and an ischemic stroke patient (bottom two rows) quantified using some of the common analysis methods (MTR_aysm, T₁-norm MTR_asym, AREX, LDA, multi-Lorentzian fitting, APTR*, MTR_{asym_comp}). It is obvious from the results that different analysis methods produce different ischemic areas, as highlighted by the green rectangular box and red arrows, but they all show significant differences when compared to the contralateral normal-appearing tissue. The animal results (top two rows) are reproduced, in part, with permission from Foo et al. Magn Reson Med 2021; 85: 2188–2200 and the patient results (bottom two rows) are reproduced, in part, with permission from Tee et al., NMR Biomed 2014;27(9):1019–1029.

Despite the challenges of using MRI to triage acute stroke patients, there is a growing movement to expand its use. It is possible to meet the time metrics specified in the AHA guidelines^99,100 because MRI can exclude stroke mimics, thereby avoiding unnecessary and inappropriate administration of thrombolysis, which is costly and can be harmful. MRI is currently used to triage patients who present in an unknown window^37,101. Although it has been shown that CT triage does not benefit stroke patients who present with non-disabling deficits, it appears that MRI may be able to identify a population that would benefit^102,103. Thus, adding APT imaging to current clinical sequences could further enhance its ability to expand treatment options. For instance, the APT-based penumbra would be the first and only way to identify at-risk tissue without exogenous contrast, the administration of which is not always possible or successful. Another potential key application is the differentiation of atheromatous-based vessel occlusion from embolic-based vessel occlusion, with the former more likely to result in a chronic perfusion deficit (without a change in pH) indicating a target that may be less amenable to mechanical thrombectomy. Thus, as MRI becomes a more commonly used modality in the acute stroke field, it is important that APT methodologies are ready to be integrated into clinical practice. In addition, recent preclinical and clinical studies have shown that APT-MRI is capable of detecting and differentiating hemorrhage stroke (increased concentration of mobile protein in blood, showing hyperintensity with respect to normal tissue) from ischemic stroke (decreased pH due to tissue acidosis, showing hypointensity)^{54,61,104,105}. Although it may be premature to declare that APT-MRI is as good as CT or conventional T₁w/T₂w MRI for the detection of hemorrhage in acute stroke patients, APT-MRI alone would provide the simultaneous visualization and separation of hemorrhagic and ischemic stroke at hyperacute stage, substantially simplifying stroke patient management.

6. Conclusions

The majority of the reported findings in preclinical and clinical studies have highlighted the potential of APT imaging to revolutionize clinical stroke imaging. This can potentially lead to better diagnosis and stratification of ischemic stroke patients for thrombolysis and thrombectomy, although larger clinical trials and further developments are still required to validate the clinical opportunity offered by translating APT imaging into clinical stroke practice. The addition of APT-based imaging to the standard MRI protocol could better visualize an ischemic penumbra, thereby identifying patients for whom thrombolytic therapy is appropriate, even if their clinical course lies outside the conventionally defined time window.

Abbreviations:

ADC

Apparent Diffusion Coefficient

APT

Amide Proton Transfer

APTw

Amide Proton Transfer-Weighted

CBF

Cerebral Blood Flow

CEST

Chemical Exchange Saturation Transfer

CT

Computed Tomography

DWI

Diffusion-Weighted Imaging

EMR

Extrapolated semisolid Magnetization transfer Reference

MCAO

Middle Cerebral Artery Occlusion

mRS

modified Rankin Scale

MTR

Magnetization Transfer Ratio

NIHSS

National Institutes of Health Stroke Scale

NOE

Nuclear Overhauser Enhancement

PET

Positron Emission Tomography

PWI

Perfusion-Weighted Imaging

tMCAO

Transient Middle Cerebral Artery Occlusion