Current emerging MRI tools for radionecrosis and pseudoprogression diagnosis.

PURPOSE OF REVIEW: This review aims to cover current MRI techniques for assessing treatment response in brain tumors, with a focus on radio-induced lesions. RECENT
FINDINGS: Pseudoprogression and radionecrosis are common radiological entities after brain tumor irradiation and are difficult to distinguish from real progression, with major consequences on daily patient care. To date, shortcomings of conventional MRI have been largely recognized but morphological sequences are still used in official response assessment criteria. Several complementary advanced techniques have been proposed but none of them have been validated, hampering their clinical use. Among advanced MRI, brain perfusion measures increase diagnostic accuracy, especially when added with spectroscopy and susceptibility-weighted imaging. However, lack of reproducibility, because of several hard-to-control variables, is still a major limitation for their standardization in routine protocols. Amide Proton Transfer is an emerging molecular imaging technique that promises to offer new metrics by indirectly quantifying intracellular mobile proteins and peptide concentration. Preliminary studies suggest that this noncontrast sequence may add key biomarkers in tumor evaluation, especially in posttherapeutic settings.
SUMMARY: Benefits and pitfalls of conventional and advanced imaging on posttreatment assessment are discussed and the potential added value of APT in this clinicoradiological evolving scenario is introduced.

INTRODUCTION

MRI plays a key-role in brain tumor follow-up, allowing to monitor response to treatment or the detection of progression, and therefore, driving critical clinical decisions. From year to year, as the therapeutic arsenal increases, assessing treatment efficacy, especially after radiotherapy or during immunotherapy, becomes more difficult. The distinction between pseudoprogression and radionecrosis from true progression or stable disease is often not possible with conventional MRI sequences and requires advanced imaging. However, no advanced MRI protocol has yet been validated. The aim of this article is to review the current knowledge on the posttherapeutic evaluation of brain tumors with a focus on the potential added value of Amide Proton Transfer (APT), a new promising noncontrast MRI technique belonging to Chemical Exchange Saturation Transfer (CEST) imaging domain.

no caption available

THE FUNDAMENTALS OF RADIONECROSIS AND PSEUDOPROGRESSION

Radiotherapy is a well established treatment option against brain primary tumors or metastases, which together account for the majority of brain tumors [1]. Stereotactic radiosurgery (SRS) is currently the most relevant therapeutic option for selected patients with brain metastasis and its indications are continuously expanding [2▪▪]. Maximum well tolerated surgical resection followed by a 6-week course of radiotherapy concurrently with temozolomide chemotherapy is the cornerstone of glioblastoma treatment [3]. Salvage re-irradiation may also be considered in some case of recurrent glioblastoma [4], although an optimal dosing regimen has not been established [5].

Unfortunately, brain radiotherapy is likely to induce chronic inflammatory reactions [6,7] and possibly result in necrotic and edematous lesions that can be extremely difficult to distinguish from tumor recurrence with both conventional and advanced MRI sequences [8,9,10▪▪,11].

Several risk factors can predispose to radio-induced brain complications. Among them, the type of systemic concurrent antitumor therapy plays an important role in enhancing radiation toxicity [12]. Notably, Immune-Checkpoint Inhibitors (ICIs), a recent target strategy against metastatic disease [2▪▪,13], are often combined with SRS [14] as focal irradiation can improve systemic antitumor immunity, a phenomenon typically referred as the abscopal effect [15,16]. Nevertheless, this combinatorial approach increases the risk of radiation necrosis [17], with largely unknown pathophysiological mechanisms [18]. Additionally, even in the absence of irradiation, immunotherapy agents can provoke unconventional transient immune-related phenomena that lead to misleading pseudoprogressing contrast-enhancing lesions [19].

Due to widespread use of these advanced therapeutic approaches, knowledge of these complex cutting-edge treatments is essential to properly assess treatment response and guide subsequent clinical decisions.

Definitions

There is no clear evidence nor consensus on the distinction between pseudoprogression and radionecrosis, probably because of the paucity of histopathological data, difficulties in assessing the correct histological diagnosis, given the frequent ‘mixed’ pattern with tumor remnants and radiation-induced tissue changes in biopsy samples [20], and the absence of large-scale, harmonized, multicenter, prospective researches.

Pseudoprogression and radionecrosis are radiologically defined by a new or enlarging area(s) in the radiation field that resolves without treatment modification. When images of radiation-induced lesions appear shortly after the end of radiotherapy (within the first 6 months for most of the studies [8,21,22], whereas others only consider the first 2 [23] or 3 months [24,25]), the term pseudoprogression is preferred, especially in the context of diffuse glioma. On the contrary, radionecrosis emerges at a later stage (from around 6 months to several years) [8,26].

Mechanism

Pathologically, no clear boundaries separate these two entities as the physiopathology of the radio-induced lesion is dynamic [27,28]. Pseudoprogression is probably an expression of early delayed brain injury and is dominated by vascular damage (i.e. vasodilation and increased capillary permeability), resulting in vasogenic edema that normally resolves spontaneously, and is often associated with transient demyelination [29]. Conversely, radionecrosis is part of late delayed brain injury and is characterized by a mixture of vascular endothelial damage and demyelination lesions, followed by neuronal death, and often does not recede [24].

Even though our understanding of the biomolecular pathways following radiotherapy is still very limited, blood vessel damage has been repeatedly recognized as one of the core step in the development of radiation toxicity [29-31], leading to hypoxia and upregulation of Hypoxia Inducible Factor-1 alpha (HIF-1α) in microglia and subsequent Vascular Endothelial Growth Factor (VEGF) induction, and several pro-inflammatory cytokines release [32]. VEGF over-expression results in leaky angiogenesis and ultimately facilitates radio-induced necrotic lesion expansion [33-35].

Incidence

Pseudoprogression and radionecrosis are common entities in both gliomas and in brain metastases. A recent meta-analysis showed that pseudoprogression occurred in 36% (95% confidence interval, 33– 40%) high-grade glioma patients [36]. At the same time, nearly a third of SRS-treated metastases show a transient, moderate volume increase at around 6 weeks after treatments, which sometimes lasts beyond 15 months [26,37,38]. The incidence of later (after 6 months) radionecrotic lesions is less understood for both primary and secondary brain tumors, as reported incidence ranges widely from 5 to 50%, and is probably underestimated [24,34,39-41].

Risk factors

Radiotherapy toxicity is complex and depends on several parameters. Total irradiation dose [42,43], treated volume [44,45] and the dose per fraction [46] are important and obvious risk factors. In the context of diffuse gliomas, it has been suggested that MGMT promoter methylated tumors are more prone to develop pseudoprogression [47] and that temozolamide increases radiation-induced lesions in both high grade [23] and low-grade glioma [48]. Immunotherapy may raise the incidence of radionecrosis in brain metastases treated with SRS [17,19] but further studies are needed to quantify its impact [14,18,49]. All in all, the intertangled network of these heterogeneous predisposing factors is largely unknown, and it is impossible to predict individual sensitivity to radiation toxicity. Large, multivariate analysis are certainly needed, and they should include an extensive panel on molecular tumor characteristics and all different types of concurrent or adjuvant antitumor agents.

Clinical implications

Radiation-induced toxicity is usually asymptomatic or paucisymptomatic [44]. When radiation necrosis is clinically meaningful, it is commonly treated with high-dose steroids or with surgical debulking [50,51]. Given the role of VEGF on the radiation-induced progressing lesion [35,52], it is not surprising that bevacizumab, an anti-VEGF monoclonal antibody, is also an effective treatment for symptomatic steroid-resistant radiation necrosis of both primary [53] and secondary tumors [54,55,56] and that it may help reducing steroid dosage [57]. Hyperbaric oxygen [58], laser interstitial thermal therapy [50] and anti-TNF antibodies [59,60] are other treatment options but have only been investigated in preliminary reports.

In the most frequent scenario, the clinical challenge is not simply to treat radiation-induced lesions but firstly to diagnose it correctly. Resolving this clinical problem is crucial as it permits to avoid premature cessation of effective treatments or delays in withdrawal of ineffective treatments. To date, no validated single MRI-based imaging metric can differentiate between treatment response and tumor recurrence after radiation therapy. Therefore, a multimodality approach is almost always required, although sometimes still insufficient because of intrinsic limitations of the available MRI sequences, as furtherly discussed.

CURRENT MRI IMAGING IN POSTTREATMENT TUMOR EVALUATION

The Response Assessment in Neuro-Oncology (RANO) working group has provided consensus response criteria for high-grade [61] and low-grade [62] gliomas, and for brain metastases with the commendable purposes of accuracy and reproducibility between different institutions. At present, however, it is extremely difficult to combine these two characteristics in posttherapeutic assessment of brain tumors as accuracy requires advanced multimodal imaging protocols that are extremely difficult to standardize and validate in multicenter studies. On the other hand, reproducibility can be achieved with conventional imaging but it is often far from being accurate. At present, on the urge of interpretation of clinical trials, reproducibility has been preferred and RANO criteria are based on conventional MRI sequences [63]. To overcome inherent limitation of RANO criteria on the evaluation of pseudoprogressing lesions during ICIs trials, immunotherapy response assessment for neuro-oncology (iRANO) criteria [64] have been developed but they are still based on morphological imaging features, and therefore, lead to delayed diagnosis.

Conventional sequences

Visual inspection of T1-weighted (T1w), T2-weighted (T2w), and fluid-attenuated inversion recovery (FLAIR) sequences is essential for detecting fine anatomical details but it is useless in the metabolic discrimination of space-occupying lesion. Morphological imaging after contrast agent injection detects blood–brain barrier leakage, which is present in both radiation-induced inflammation and in neoplastic lesion [8]. Early studies explored the value of T2w and T1w postcontrast imaging in differentiating tumor recurrence from radiation necrosis in brain metastases treated with SRS. Evaluation of the spatial concordance between the boundaries of the lesion on T2w and T1w postcontrast imaging show a T1/T2 match (in favor of tumor recurrence) or T1/T2 mismatch (in favor of radionecrosis) [65]. Quantitative determination of lesion quotient, defined as the area of a hypointense nodule on T2w divided by its area on contrast-enhanced T1w, suggested to show different results in radionecrosis (lesion quotient less than 0.3) and in recurrent metastasis (lesion quotient greater than 0.6) [66]. These qualitative and quantitative signs were not confirmed in further studies [67,68].

Type of contrast uptake has also been investigated, and a ‘cut-green pepper’, a ‘soapbubbles’ or a ‘gruyere’ cheese contrast enhancement appearance has been related to radionecrosis [69,70] but the subjective evaluation of these signs limits their reproducibility [9].

Overall, conventional sequence evaluation does not distinguish between tumor and posttherapeutic lesion at an early stage [71], whereas in the later stage, the tumor progresses while the posttherapeutic lesion remains stable, shrink or disappear.

Diffusion-weighted imaging

Diffusion-weighted imaging and its various extensions (diffusion tensor imaging, diffusion kurtosis imaging, neurite orientation dispersion and density imaging, diffusional and constrained diffusional variance decomposition) can provide important complementary information on tissue microstructure in treatment-naive patients [72-76]. Conversely, after radiotherapy, the evaluation of diffusion-derived changes must be interpreted with caution as studies on this topic are contradictory [77,78] and similar diffusion water molecules metric can reflect opposite phenomena (i.e. diffusion restriction because of hypercellularity or postradiotherapy coagulative necrosis) [72,79]. An example of diffusion restriction in a pathologically proved radionecrosis is shown in Fig. 1. Detection of restricted diffusion foci in irradiated tumor successively treated with Bevacizumab is a well known form of therapy-induced necrosis [80,81]. Radiologist and clinicians should be aware of this entity and avoid misdiagnosis with stroke or tumor progression.

FIGURE 1

Pathologically proven radionecrosis in a 45 years old woman with anaplastic pleomorphic xanthoastrocytoma treated with Stupp protocol (radiotherapy and concurrent chemotherapy with temozolomide). Axial 3D T1w spin echo before (a) and after (b) gadolinium injection showed a contrast-enhancing intra-axial frontal lesion, with some hemorrhagic remnants that were spontaneously bright on T1w. This lesion was surrounded by a large nonenhancing FLAIR hyperintensity, visible on axial 3D FLAIR (c) and it was characterized by a strong diffusion restriction (d: axial diffusion with b100 value, e: ADC cartography). No hyperperfusion was seen [f: axial arterial spin labelling (ASL) perfusion]. This radio-induced lesion developed 2 months after the end of radiotherapy. ADC, apparent diffusion coefficient.

Intravoxel Coherent Motion Imaging, a novel diffusion technique that produces simultaneous diffusion and perfusion maps, can have a potential role in identifying radiation-induced changes in gliomas and brain metastases treated with SRS, as a pilot works suggest [82].

Susceptibility-weighted imaging

Several paramagnetic and diamagnetic sources of signal, such as deoxygenated hemoglobin, tissue calcifications or iron deposits can alter magnetic susceptibility and thus lead to susceptibility-weighted imaging (SWI) changes. They can be present in brain metastases, especially of melanoma [83], and in glioma [84]. In this latter group, the degree of intratumoral susceptibility signal intensity (ITSS) was shown to be positively correlated with glioma grade and higher perfusion values [85]. After radiotherapy, a marked increase of SWI signal changes within the radiation field is often observed, traducing blood vessel injury and radiotherapy-related remnants [79,86]. Measuring the proportion of hemorrhage shown in SWI lowers false-positive rate in the differentiation of recurrence from radionecrosis-based simply on perfusion measurements [87] and increases overall survival prediction [87].

coefficient is also sensitive to hemorrhage and calcifications, and it can be measured within a specific region of interest by fitting signal decay through multiechoes gradient echo sequences. A pilot study in glioblastoma patients showed that this coefficient, also referred as apparent transverse relaxation, shows lower value in pseudoprogressing compared with progressing contrast –enhancing lesions [88]. Preclinical data suggest that coefficient might even predict radionecrosis 10 weeks before morphological changes [89].

Brain perfusion

Brain MRI tumor perfusion measures can be acquired through dynamic susceptibility contrast (DSC), dynamic contrast-enhanced (DCE), or arterial spin labelling (ASL) techniques. Relative Cerebral Blood Volume (rCBV) value, a semi-quantitative vascularity measure derived from DSC perfusion, is the most used advanced MRI indicator in posttreatment tumor assessment [90-93], with a cutoff of the contralateral white matter Region of Interest (ROI) usually ranging above 1.5–2 for tumor lesions [94-96]. Elevated rCBV is a marker of increased microvascular density and often reflects tumor aggressiveness, correlating with glioma grade [97] and survival [98] (except for oligodendrogliomas [99]). The increase in rCBV compared with baseline after antitumor treatment predicts a worse outcome [100,101] and seems more accurate in survival prediction than histopathologic grade in glioma [102].

In SRS-treated metastasis, the percentage signal-intensity recovery (PSR) towards baseline, an indicator of capillary permeability, appears to be a better prognostic indicator of metastatic tumor progression than rCBV as radiation-induced lesions show a higher PSR than do recurrent tumors [103].

Unfortunately, because of the difficulty in controlling several technical acquisition and postprocessing variables, there are considerable interinstitutional and even intrainstitutional variations in DSC-derived measurements [104]. This variability severely limits reproducibility and hampers the implementation of DSC perfusion in the routine assessment of treatment response (e.g. current RANO criteria) [105].

DCE-MRI data enable the evaluation of several pharmacokinetics parameters. The most investigated is the volume transfer coefficient (Ktrans) that reflects the tumor vascular permeability and is higher in radiation necrosis than in tumor progression [106-109]. In brain metastatic disease, DCE technique appears to have better diagnostic accuracy than DCS perfusion in the differential diagnosis between radiation necrosis and tumor recurrence [110]. In contrast, high-grade glioma DSC-derived blood volume measures seem more accurate than DCE-derived permeability measures in the same distinction [111,112].

ASL is a contrast-free perfusion technique obtained through magnetically labeled blood protons. ASL imaging is starting to be studied in high-grade gliomas [113], especially in posttherapeutic settings, where it showed no need of leakage-correction algorithms, fewer susceptibility artifacts than DSC perfusion [114] but contrasting accuracy results. To date, no studies have been published on the ASL performance in brain metastasis posttreatment evaluation.

Spectroscopy

Magnetic resonance spectroscopy (MRS) provides insight into metabolic tissue features by noninvasively detecting solute protons in water. Concentration of choline-containing compound metabolites, markers of cell membrane turnover, is highly increased in tumor progression [115], whereas lipids and lactates have been shown to dominate in posttreatment disease, suggesting cellular necrosis [116]. Several metabolite ratios have been proposed to diagnose radionecrotic lesion (summarized in previous reviews [117-119]), and Cho/NAA and Cho/Cr ratios seem to be the best discriminators [120] but none of them have been validated in multicenter studies. When compared with perfusion, MRS showed inferior discriminating abilities [121]. However, the combination of MRS and perfusion imaging increases diagnostic accuracy [111,122], hence with sensitivity and specificity values that still prevent replacing invasive biopsy sampling or serial imaging confirmation.

D-2-hydroxyglutarate (2HG) MRS [123] is an umbrella term that refers to MRS techniques that can measure 2HG oncometabolite [124] in IDH-mutant diffuse gliomas. Preliminary longitudinal 2HG MRS evaluations have shown that this biomarker decreases after antitumor treatments [125] and possibly increases in tumor recurrence [126], thus this novel noninvasive technique could also aid in IDH-mutant glioma posttreatment assessment.

Texture analysis and radiomics

The term ‘texture analysis’ encompasses different computational methods that are used to quantify the spatial arrangement of image signal intensities. After ROI definition and preprocessing, several features, imperceptibles to the human eye, are extracted, selected and finally classified [127]. The complementary information provided by this noninvasive, objective and possibly fully automatic approach is clearly attractive but lack of standardization and overfitting issues are still major constraints [128,129]. When these multiple parameters are used to predict clinical and biological variables, this multistep approach is referred to as ‘radiomics’ [130].

MR texture analysis and radiomics have been conducted mainly with conventional sequences, with already interesting results, especially in the evaluation of tumor shape features and surface irregularities in the diagnosis of glioblastoma versus pseudoprogression [131]. Adding complementary advanced sequences to this complex evaluation indeed promises to achieve higher diagnostic performances [132,133▪▪].

CHEMICAL EXCHANGE SATURATION TRANSFER IMAGING: A NEW TOOL ON MRI ARSENAL

Chemical exchange saturation transfer (CEST) is a molecular imaging technique recently available on 3 Tesla MRI scanners [134,135] that detects low-concentration solute molecules with exchangeable hydrogen protons. By applying a radiofrequency pulse at their resonance frequencies, the chemical species of interest – such as amide (NH), amine (NH2) or hydroxyl group (OH) – reach a saturation state and their labile excited hydrogen protons are exchanged with the nonexcited hydrogen protons of solvent water. If this process is repeated continuously for a few seconds of RF irradiation, saturation builds up, which decreases water signal, thus indirectly reflecting the concentration of the targeted species with amplified detection. Their detectability is, for example, amplified by a factor of 100, if this exchange takes places 100 times [136].

To extract the CEST signal of interest, multiple samples are acquired around the frequency of water and molecules of interest to correct field inhomogeneities [137,138], for denoising purposes [139], and for averaging of the sampled values to increase the signal-to-noise ratio [140]. This set of sampled volumes (normalized for an unsaturated volume) is called the Z-Spectrum [136].

The most marked clinical distinction for this molecular technique is whether the CEST agent is endogenous (and therefore, is already found in the human body, such as the amide and amine groups from peptides and proteins [141]) or exogenous (and therefore, needs to be administered, such as glucose-based agents [142] or Iopamidol [143]).

Amide CEST imaging, also known as APT, has been shown to provide more stable and sensitive detection compared with other CEST agent on clinical 3 Tesla scanners, and therefore is, to date, the most used CEST technique, in and out of the neurological field [144]. In neuro-oncology, an increase in APT-weighted (APTw) signal intensity is observed in tumor tissues, because of elevated concentration of intracellular mobile proteins and peptides, and consequent increase of protein backbone amides.

APTw imaging has demonstrated promising results in glioma grading [141,145,146], in the identification of higher cellularity and proliferation area in heterogeneous diffuse glioma [147], as in IDH status and 1p/19q co-deletion prediction [148-151].

Another major potential application of APTw imaging is the assessment of response to treatment, with the hypothesis that tumor hypercellularity leads to an increase in APTw signal intensity compared with lower cellular density of therapeutic related changes. In 2011, a first preclinical study showed a significant decrease in APTw signal intensity in the radionecrotic lesions compared with gliomas [152]. Shortly thereafter, another preclinical work showed that APTw values decrease more rapidly than diffusion and ASL perfusion values, suggesting that Amide CEST could prompt early response information [153]. The former group then conducted a first clinical radiohistopathological validation in glioma patients and found positive significant correlations between APTw signal intensity and both cellularity and proliferation index [154], therefore, suggesting that APTw values are a marker of active glioma in posttreatment settings. These results are in line with other studies that have been conducted in diffuse gliomas, especially in context of the antiangiogenic therapy [155] and in the distinction between progressing and pseudoprogressing lesion [149]. Interestingly, the changes in APTw signal intensity was also seen to spatially overlap FET-PET data for both contrast-enhancing and noncontrast-enhancing glioma lesions [156], which is easily explained as biological processes behind this different imaging modalities are similar.

Concerning brain metastases, previous study reported that not only the CEST signal from amide groups but also from the Nuclear Overhauser Enhancement (NOE) effects, a protein un-folding biomarker [157] that can be derived through CEST imaging fitting [138,158], was able to distinguish radiation necrosis from tumor progression, and NOE signal intensity provided the best separation of these two conditions [158].

Advances in CEST imaging are increasing the ability to extract both individual APT and NOE biomarkers in clinical routine [134,159] without the contamination of magnetization transfer effect coming from semisolid macromolecules [136].

Figure 2 illustrates an example of biopsy-proven tumor progression that was early detected by APTw imaging and neither by conventional nor perfusion MRI, whereas Fig. 3 displays a radiation-induced lesion that regressed at 6-month follow-up.

FIGURE 2

Biopsy-proven tumor progression early detected by Amide Proton Transfer and not by conventional nor perfusion MRI. A 54-year-old woman with a single brain metastases of a breast cancer was treated with surgical resection and stereotactic radiosurgery (SRS) to the surgical cavity. Nine months after SRS, a postrolandic lesion appeared, visible on noncontrast 3D T1w spin echo (a) and showing a linear contrast-enhancement (b), and a FLAIR hyperintensity on axial 3D FLAIR sequence (c). Axial SWI (d) showed several hemosiderin remnants, axial diffusion with b100 value (e) and ADC cartography (f) showed no diffusion restriction, dynamic susceptibility contrast (DSC) perfusion (g) no neoangiogenesis. Therefore, conventional and advanced imaging suggested a radio-induced lesion. In addition, amide signal increased intensity on fluid-suppressed [150] APTw imaging (h) suspected tumor progression, and this was confirmed by a surgical resection done shorty after a 3-month MRI follow-up (w: 3D T1w spin echo without contrast injection, x: axial diffusion with b1000 value, y: axial 3D FLAIR, z: axial 3D T1 spin echo with contrast injection). Olea Sphere 3.0 software (Olea Medical, La Ciotat, France) was used to compute APTw and perfusion maps. APTw, Amide Proton Transfer-weighted; ADC, apparent diffusion coefficient; SWI, susceptibility-weighted imaging.

FIGURE 3

Radiation-induced contrast-enhancing lesion that regressed at 6-month follow-up. A 46-year-old patient with a single brain metastases of a lung cancer was treated with surgical resection and stereotactic radiosurgery (SRS) to the surgical cavity. Four months after SRS, an irregular, nonnodular contrast enhancing lesion appeared (a: noncontrast 3D T1w spin echo, b: postcontrast 3D T1w spin echo, c: axial 3D FLAIR) with several hemosiderin remnants visible on axial SWI (d). The lesion showed neither diffusion restriction (e: axial diffusion with b100 value, f: ADC map) nor neoangiogenesis on axial DSC perfusion (g) nor amide signal increased intensity on fluid-suppressed APT-weighted imaging (h). Therefore, a radio-induced lesion was suspected. This diagnosis was confirmed at a 6-month MRI follow-up, where the contrast enhancement disappeared (w: 3D T1w spin echo without contrast injection, x: axial diffusion with b1000 value, y: axial 3D FLAIR, z: axial 3D T1 spin echo with contrast injection). Olea Sphere 3.0 software (Olea Medical, La Ciotat, France) was used to compute the APTw map. APTw, Amide Proton Transfer-weighted; ADC, apparent diffusion coefficient; SWI, susceptibility-weighted imaging.

CONCLUSION

Radionecrosis and pseudoprogression are common phenomena that urge precise imaging diagnosis to provide optimal early patient care. Shortcomings of conventional MRI are well known, whereas the added value of complementary advanced imaging sequences, especially perfusion, needs to be accurately established, as inherent limitations have been reported for each sequence. In this evolving scenario, APT-CEST metrics offer new problem-solving tools that expand MRI armamentarium for assessing treatment response. A multimodal machine-learning analysis that includes the best performing perfusion technique with validated molecular information provided by CEST imaging promises highly accurate personalized patient care of each individual brain lesion.

Acknowledgements

We acknowledge the help of Stéphane Lehéricy (Department of Neuroradiology, La Pitié Salpêtrière - Charles Foix Hospital, Paris, France), Julian Jacob (Department of Radiation Oncology, La Pitié Salpêtrière - Charles Foix Hospital, Paris, France) and Malgorzata Marjanska (Center for Magnetic Resonance Research, Department of Radiology, Minneapolis, Minnesota, United States) for their critical revision of the article. We thank Christos Papageorgakis (Department of Research & Innovation, Olea Medical, La Ciotat, France) for MR acquisition processing support.

Financial support and sponsorship

None.

Conflicts of interest

L.N. has no financial disclosure. S.C. is currently an employee at Olea Medical (Research & Innovation Team Leader – Molecular Imaging Project Manager).