Post-Surgical Depositions of Blood Products Are No Major Confounder for the Diagnostic and Prognostic Performance of CEST MRI in Patients with Glioma

Amide proton transfer (APT) and semi-solid magnetization transfer (ssMT) imaging can predict clinical outcomes in patients with glioma. However, the treatment of brain tumors is accompanied by the deposition of blood products within the tumor area in most cases. For this reason, the objective was to assess whether the diagnostic interpretation of the APT and ssMT is affected by methemoglobin (mHb) and hemosiderin (Hs) depositions at the first follow-up MRI 4 to 6 weeks after the completion of radiotherapy. A total of 34 participants underwent APT and ssMT imaging by applying reconstruction methods described by Zhou et al. (APTwasym), Goerke et al. (MTRRexAPT and MTRRexMT) and Mehrabian et al. (MTconst). Contrast-enhancing tumor (CE), whole tumor (WT), mHb and Hs were segmented on contrast-enhanced T1wCE, T2w-FLAIR, T1w and T2*w images. ROC-analysis, Kaplan–Meier analysis and the log rank test were used to test for the association of mean contrast values with therapy response and overall survival (OS) before (WT and CE) and after correcting tumor volumes for mHb and Hs (CEC and WTC). CEC showed higher associations of the MTRRexMT with therapy response (CE: AUC = 0.677, p = 0.081; CEC: AUC = 0.705, p = 0.044) and of the APTwasym with OS (CE: HR = 2.634, p = 0.040; CEC: HR = 2.240, p = 0.095). In contrast, WTC showed a lower association of the APTwasym with survival (WT: HR = 2.304, p = 0.0849; WTC: HR = 2.990, p = 0.020). Overall, a sophisticated correction for blood products did not substantially influence the clinical performance of APT and ssMT imaging in patients with glioma early after radiotherapy.


Introduction
Standard of care for diffuse glioma includes maximum safe resection, with subsequent radio-and chemotherapy [1].Yet, since treatment-related changes, such as pseudoprogression and radionecrosis have similar morphological imaging features compared to progressing glioma tissue, therapy response assessment and survival prediction are wellknown challenges in clinical neuro-oncology, with possibly harmful consequences for the patient [2].In this regard, the promising clinical potential of functional MR imaging methods, including perfusion-weighted imaging, MR spectroscopy and chemical exchange saturation transfer (CEST) imaging has been shown in recent years [3,4].Imaging of the amide-proton transfer (APT) and of the semi-solid magnetization transfer (ssMT) relies on the selective radio frequency (RF) saturation of protons bound in peptide bindings or sub-cellular macromolecules, with subsequent magnetization transfer to bulk water through chemical exchange or spin-spin couplings, respectively [5] (collectively referred to as CEST imaging in the following).Several groups have demonstrated that APT and ssMT imaging could predict the therapy response and survival of patients with glioma before and after radiochemotherapy [6][7][8][9][10].Others have demonstrated that imaging of the APT and ssMT could also be used to differentiate radiation-induced changes from tumor progression as early as the first follow-up after the completion of radiotherapy [11][12][13][14][15][16][17].However, treatment of brain tumors is frequently accompanied by perioperative and radiation-induced disruptions of the blood-brain-barrier with subsequent depositions of blood products, such as methemoglobin (mHb) and hemosiderin (Hs), in the tumor area [18].mHb and Hs are both proteinaceous, contain paramagnetic iron (Fe 3+ ) and exist to varying degrees as conglomerates of insoluble macromolecules.For this reason, mHb and Hs not only have a strong influential impact on T 1 and T 2 , but should also contribute to the APT and ssMT pool.However, whilst several groups have demonstrated that CEST imaging of the APT can differentiate between acute and subacute stages of intracranial hemorrhage [19][20][21], little is known about the influence of mHb and Hs on the clinical performance of APT and ssMT imaging in patients with glioma in the post-radiotherapy interval.Furthermore, CEST contrasts are heavily dependent on the metrics used for their reconstruction form the Z-spectrum [7,22,23].Therefore, the purpose of this study was to assess the relevance of advanced correction methods for mHb and Hs depositions in the tumor area for the clinical performance of APT and ssMT contrasts under the application of different reconstructions methods first described by Zhou et al. (APTw asym ) [24], Goerke et al. (MTR Rex APT and MTR Rex MT) [25] and Mehrabian et al. (MT const ) [7].

Materials and Methods
Eligible for this prospective clinical study were all patients who received radiotherapy for diffuse glioma at the Department of Radiation-Oncology of the University Hospital Heidelberg between September 2018 and December 2021, and who were 18 years of age or older, had a Karnofsky Performance Score of at least 50 and had the legal capacity to consent.Eventually, 72 study participants (61 with initial disease and 11 with relapsing/progressive disease) were enrolled and received CEST imaging at the first follow-up MRI 4 to 6 weeks after the completion of radiotherapy.Two participants had to be excluded from the analysis due to heavy motion artifacts, seven due to incomplete datasets and one for excessive perioperative ischemia.The data cut-off was 3 May 2022.The association of CEST imaging with therapy response and progression-free survival (PFS), as well as with overall survival (OS), was previously investigated in two studies involving 61 and 49 participants of the same cohort, respectively [5,10].Therapy response was assessed based on longitudinal clinical data and MRI according to the revised response assessment in neuro oncology (RANO) criteria by two radiologists with 6 (N.v.K.D.) and 11 (D.P.) years of experience in neuroimaging at the time of data acquisition [26].The results of the assessment were reconciled with institutional multidisciplinary tumor board decisions to account for potential changes in relevant medications, such as antiangiogenic or cytotoxic drugs, and changes in clinical status.Overall survival (OS) was assessed by written request to the relevant public registries and was available for 54 of 62 evaluable participants.Due to differences in the tumor biology between midline gliomas and hemispherical gliomas, the data of five study participants with midline gliomas were additionally excluded from the analysis [27,28].Finally, given that the associations of the investigated contrasts with therapy response and survival are influenced by the presence of residual contrast enhancement on MRI [6,11], the data of 34 participants with available survival data, hemispherical gliomas and residual contrast enhancement were analyzed (Figure 1).multidisciplinary tumor board decisions to account for potential changes in relevant medications, such as antiangiogenic or cytotoxic drugs, and changes in clinical status.Overall survival (OS) was assessed by written request to the relevant public registries and was available for 54 of 62 evaluable participants.Due to differences in the tumor biology between midline gliomas and hemispherical gliomas, the data of five study participants with midline gliomas were additionally excluded from the analysis [27,28].Finally, given that the associations of the investigated contrasts with therapy response and survival are influenced by the presence of residual contrast enhancement on MRI [6,11], the data of 34 participants with available survival data, hemispherical gliomas and residual contrast enhancement were analyzed (Figure 1).

Figure 1.
Flowchart.Displayed is a description of the whole study cohort participants and the 34 participants that were eventually included in the analysis.Between September 2018 and December 2021, 72 participants with diffuse glioma received CEST imaging 4 to 6 weeks after the completion of radiotherapy.In total, 11 participants had to be excluded from the analysis due to the indicated reasons.Given that diffuse midline gliomas are biologically distinct from hemispherical gliomas [27,28] and previous studies observed a dependency of CEST contrasts on the presence of residual tumor-associated contrast enhancement on MRI [6,11], the datasets of 34 participants with hemispherical gliomas, presence of residual contrast enhancement on MRI and available survival data were included in the analysis.The associations of mean CEST contrast values with therapy response and overall survival were tested by Kaplan-Meier analyses and log rank tests.KPI = Karnofsky Performance Score; w. = with; w/o.= without.
Histology: For all of the 34 study participants evaluated, tumor tissue was available for histopathological analysis after biopsy or surgical resection.Routinely, IDH-, ATRX-, LOH1p19q-and MGMT-status were assessed, and histopathological classification was performed in accordance with the 2016 version of the World Health Organization (WHO)'s criteria for the classification of primary central nervous system (CNS) tumors.Flowchart.Displayed is a description of the whole study cohort participants and the 34 participants that were eventually included in the analysis.Between September 2018 and December 2021, 72 participants with diffuse glioma received CEST imaging 4 to 6 weeks after the completion of radiotherapy.In total, 11 participants had to be excluded from the analysis due to the indicated reasons.Given that diffuse midline gliomas are biologically distinct from hemispherical gliomas [27,28] and previous studies observed a dependency of CEST contrasts on the presence of residual tumorassociated contrast enhancement on MRI [6,11], the datasets of 34 participants with hemispherical gliomas, presence of residual contrast enhancement on MRI and available survival data were included in the analysis.The associations of mean CEST contrast values with therapy response and overall survival were tested by Kaplan-Meier analyses and log rank tests.KPI = Karnofsky Performance Score; w. = with; w/o.= without.
Histology: For all of the 34 study participants evaluated, tumor tissue was available for histopathological analysis after biopsy or surgical resection.Routinely, IDH-, ATRX-, LOH1p19q-and MGMT-status were assessed, and histopathological classification was performed in accordance with the 2016 version of the World Health Organization (WHO)'s criteria for the classification of primary central nervous system (CNS) tumors.Please see Table 1 for a detailed description of the histopathological tumor characteristics for all evaluated participants. 1FU = follow-up MRI; 2 GBM = glioblastoma, 3 WHO II-IV = World Health Organization classification system for primary brain tumors grade II-IV; 4 IDH = isocitrate-dehydrogenase isotype 1/2; 5 wt = wildtype; 6 mut = mutation.
Image Acquisition and Postprocessing: Image acquisition was performed on a 3T whole-body MR scanner (MAGNETOM Prisma; Siemens Healthineers, Erlangen, Germany) with an integrated transmit body coil and a 64-channel head/neck receiving coil.The CEST data were processed in Matlab ® (Mathworks, version 2019b, Natick, MA, USA) using customized scripts.
Imaging of the APT and ssMT according to Goerke et al. (MTR Rex APT and MTR Rex MT): A 3D spiral-centrally reordered gradient-echo acquisition sequence (snapshot CEST [29,30]) was applied with the same image readout parameters (matrix = 128 × 104 × 16, resolution = 1.7 × 1.7 × 3 mm 3 ) and presaturation as previously described by Goerke et al. [25].For presaturation, trains of 148 Gaussian-shaped radio frequency (RF) pulses (echo time (TE) = 2.75 ms, repetition time (TR) = 5.5.ms, flip angle = 7 • , puls length (tp) = 0.02 s and duty cycle = 80%) with two amplitudes (B 1 = flip angle/(γ•tp)) of 0.6 µT and 0.9 µT were acquired at 57 unequally distributed offsets in the range between ±250 ppm and −300 ppm for normalization at two M 0 , resulting in a saturation time of 3.7 s and a total measurement time of 7:34 min.The WASABI [31] (3:41 min) approach was applied to yield B 0 and B 1 maps, using the same image readout and similar presaturation parameters as described above.In this case, presaturation was performed by sampling 31 equally distributed frequency offsets around ±2 ppm.For post-processing, the CEST and WASABI data were first co-registered with a rigid registration algorithm in MITK (version v2022.10).Then, the CEST data were processed in Matlab ® (Mathworks, version 2019b, Natick, MA, USA).A correction of B 0 inhomogeneities was achieved by shifting the Z-spectra along ∆ω [32] and denoising was achieved under the application of a principle component-based algorithm [32].The reconstruction of the MTR Rex APT and MTR Rex MT from the Z-spectrum was performed as described in [25] under the application of a four-pool Lorentzian-fit ±250 ppm up-and down-field of the water offset (0 ppm: direct water saturation, 3.5 ppm: APT, −3.5 ppm: rNOE, and −2. pools, Z ref = fitted pools-pool of interest).A two-point "contrast-correction" method first proposed by Windschuh et al. [33] was applied for additional B 1 -correction.
Imaging of the ssMT according to Mehrabian et al. (MT const ): Image readout parameters, presaturation, co-registration of CEST data and B 0 correction were the same as described above.The MT const was reconstructed from the Z-spectrum of the CEST data with B 1 = 0.6 µT with a Lorentzian fit around ±6 ppm with [7]: Imaging of the APT according to Zhou et al. (APTw asym ): Again, pulse sequence and image readout parameters were the same as described above [24].In accordance with recent consensus guidelines [23], four rectangular RF pulses with a B 1 of 2 µT (t sat = 0.2 s and 95% duty cycles) at 16 frequency offsets at ±4 (1), ±3.75 (2), ±3.5 (2), ±3.25 (2), and ±3 (1) ppm and an additional M 0 at −300 ppm were obtained, resulting in a scan time of 2:00 min.Co-registration of the CEST data and B 0 -correction of the Z-spectra was performed under the application of similar post-processing methods as described above with APTw = Z(−3.5 ppm) − Z(3.5 ppm).
Quantitative T 1 mapping: The longitudinal relaxation time of water (T 1 ) was measured via quantitative mapping with the same image readout parameters as above.A saturation recovery sequence with recovery times (t rec ) of 0.1, 0.25, 0.5, 1.0, 1.5, 2.0, 2.5, 3.5, 5.0, 7.5 and 10.0 s and M z (t rec [25] was applied, resulting in a scan time of 1:15 min. Segmentation of tumor volumes and blood products: Three-dimensional segmentations of tumor volumes and blood products were also performed in Matlab ® (Mathworks, version 2019b, Natick, MA, USA).mHb (detectable in 14 cases) was defined as hyperintense material on T 1 w, and Hs (detectable in 33 cases) as dark tissue on T 2 *w susceptibilityweighted imaging (SWI).Contrast-enhancing (CE) and whole tumor volumes (WT) were segmented on contrast-enhanced T 1 w (T1w CE ) and T 2 w fluid-attenuated inversion recovery (T 2 w-FLAIR) images.WT encompassed CE plus adjacent T 2 w-FLAIR-hyperintense signal alterations.Larger mHB depositions that were identifiable as such on T 1 w CE and T 2 w-FLAIR (e.g., due to localization in the resection cavity or associated T 2 w-FLAIR signal drops that indicated surrounding Hs (Figure 2b)) were grossly excluded (Figure S1).Contrast-enhancing and whole tumor volumes corrected for mHb and Hs (CE C and WT C ) were calculated from CE and WT in Matlab ® by subtracting the overlapping mHb and Hs volumes (Figure A1).
Statistical analyses: Mann-Whitney-U-test was applied to test for differences between the mean CEST contrast values of blood products and corrected tumor volumes, as well as between uncorrected and corrected tumor volumes.Receiver operating characteristic (ROC) analyses were performed to test for the association of mean CEST contrast values of uncorrected and corrected tumor volumes with therapy response, as assessed according to the RANO criteria.Kaplan-Meier analyses and log rank tests were used to test for the association of mean CEST contrast values of uncorrected and corrected tumor volumes with OS.In-house software in Matlab ® (Mathworks, version 2019b, Natick, MA, USA) was used for all statistical analyses.p ≤ 0.05 was considered as being statistical significant.Statistical analyses: Mann-Whitney-U-test was applied to test for differences between the mean CEST contrast values of blood products and corrected tumor volumes, as well as between uncorrected and corrected tumor volumes.Receiver operating characteristic (ROC) analyses were performed to test for the association of mean CEST contrast values of uncorrected and corrected tumor volumes with therapy response, as assessed according to the RANO criteria.Kaplan-Meier analyses and log rank tests were used to test for the association of mean CEST contrast values of uncorrected and corrected tumor volumes with OS.In-house software in Matlab ® (Mathworks, version 2019b, Natick, MA, USA) was used for all statistical analyses.p ≤ 0.05 was considered as being statistical significant.

Results
In total, 72 study participants (mean age 59 ± 16 years; 43 male) underwent CEST MRI at the first follow-up 4 to 6 weeks after the completion of radiotherapy.The data of 34 participants (19 male, 15 female; mean age of 59.2 ± 15.6 years) with diffuse hemispherical glioma who had available data on therapy response and survival, and showed residual contrast enhancement on MRI were included in the analysis.A total of 16 participants were assessed as having stable disease (SD) and 18 were assessed as having progressive disease (PD).Median OS was 287 days (min.63 and max.1271 days), with 24/34 participants having reached an endpoint by the data cut-off on 3 May 2022.Detailed clinical characteristics of the analyzed study cohort are provided in Table 1.

CEST Contrast Maps of Participants with Larger Depositions of mHb and Hs
Exemplary contrast maps of four study participants with larger depositions of mHb (2) and Hs (2) in the tumor area are displayed in Figures 2 and 3, respectively.Associated quantitative T 1 maps are depicted in Figure A2.Visually, larger depositions of Hs showed remarkably higher values on the MTR Rex APT contrast maps sharply confined to the drawn ROIs (Figure 3).This was not observed on the other contrast maps.Furthermore, larger mHb depositions visually showed remarkably dropped values on MTR Rex APT, MTR Rex MT, MT const and T 1 (Figures 2 and A2) contrast maps, which was recapitulated by lower MTR Rex APT, MTR Rex MT and MT const mean values of mHb in comparison to CE C and WT C (Table A1).Exemplary fitted Z-spectra for representative voxels of contrastenhancing tumor tissue, peritumoral T 2 w-FLAIR-hyperintense signal alterations, and mHb and Hs in an exemplary participant with larger depositions of mHb and Hs are shown in Figures S1-S4.

Differences between the Mean CEST Contrast Values of Uncorrected and Corrected Tumor Volumes
Even though there were differences between the mean MTR Rex APT, MTR Rex MT and MT const contrast values of mHb and CE C /WT C , as well as the mean MTR Rex MT and APTw asym contrast values of Hs and CE C /WT C (Table A1), there were no differences between the mean values of any contrast for uncorrected and corrected tumor volumes (Figure 4).

Results
In total, 72 study participants (mean age 59 ± 16 years; 43 male) underwent CEST MRI at the first follow-up 4 to 6 weeks after the completion of radiotherapy.The data of 34 participants (19 male, 15 female; mean age of 59.2 ± 15.6 years) with diffuse hemispherical glioma who had available data on therapy response and survival, and showed residual contrast enhancement on MRI were included in the analysis.A total of 16 participants were assessed as having stable disease (SD) and 18 were assessed as having progressive disease (PD).Median OS was 287 days (min.63 and max.1271 days), with 24/34 participants having reached an endpoint by the data cut-off on 3 May 2022.Detailed clinical characteristics of the analyzed study cohort are provided in Table 1.

CEST Contrast Maps of Participants with Larger Depositions of mHb and Hs
Exemplary contrast maps of four study participants with larger depositions of mHb (2) and Hs (2) in the tumor area are displayed in Figures 2 and 3, respectively.Associated quantitative T1 maps are depicted in Figure A2.Visually, larger depositions of Hs showed remarkably higher values on the MTRRexAPT contrast maps sharply confined to the drawn ROIs (Figure 3).This was not observed on the other contrast maps.Furthermore, larger mHb depositions visually showed remarkably dropped values on MTRRexAPT, MTR-RexMT, MTconst and T1 (Figures 2 and A2) contrast maps, which was recapitulated by lower MTRRexAPT, MTRRexMT and MTconst mean values of mHb in comparison to CEC and WTC (Table A1).Exemplary fitted Z-spectra for representative voxels of contrast-enhancing tumor tissue, peritumoral T2w-FLAIR-hyperintense signal alterations, and mHb and Hs in an exemplary participant with larger depositions of mHb and Hs are shown in Figures S1-S4.

Association of CEST Contrast Values of Uncorrected and Corrected Tumor Volumes with Therapy Response
In the ROC analyses, the MTR Rex MT was the only contrast that showed a noticeable improvement regarding the association of the mean contrast values of tumor tissue with therapy response at the first follow-up.The area under the curve (AUC) for differentiating participants with PD and SD according to the mean contrast values (with PD > SD) was 0.677 (p = 0.081) for CE and 0.705 (p = 0.044) for CE C .However, there were no differences in AUCs for WT (AUC = 0.635, p = 0.184) and WT C (AUC = 0.628, p = 0.184).The association of MT const mean contrast values with therapy response of uncorrected and corrected tumor volumes did not show relevant differences (CE: AUC = 0.826, p = 0.001; CE C : AUC = 0.816, p = 0.002; WT: AUC = 0.868, p < 0.001; WT C : AUC = 0.861, p < 0.001).The MTR Rex APT (CE: AUC = 0.438, p = 0.546; CE C : AUC = 0.424, p = 0.458; WT: AUC = 0.566, p = 0.523; WT C : 0.569, p = 0.501) and APTw asym (CE: AUC = 0.514, p = 0.904; CE C : AUC = 0.504, p = 0.986; WT: AUC = 0.538, p = 0.717; WT C : AUC = 0.552, p = 0.617) mean contrast values did not show any association with therapy response before and after the correction.The ROC curves for the investigated CEST contrasts are displayed in Figure 5.The results of the ROC analysis are summarized in Table A2.

Association of Mean CEST Contrast Values of Uncorrected and Corrected Tumor Volumes with Overall Survival
In the Kaplan-Meier analyses, the APTw asym mean contrast values showed a slightly lower association with survival for CE C (CE: HR = 2.634, p = 0.040; CE C : HR = 2.634, p = 0.040) and a slightly higher association with survival for WT C (WT: HR = 2.304, p = 0.084; WT C : HR = 2.990, p = 0.020), compared to the respective uncorrected tumor volumes, with a shorter OS of participants with higher mean values compared to the cohort median (CE:  A3.

Association of CEST Contrast Values of Uncorrected and Corrected Tumor Volumes with Therapy Response
In the ROC analyses, the MTRRexMT was the only contrast that showed a noticeable improvement regarding the association of the mean contrast values of tumor tissue with therapy response at the first follow-up.The area under the curve (AUC) for differentiating participants with PD and SD according to the mean contrast values (with PD > SD) was 0.677 (p = 0.081) for CE and 0.705 (p = 0.044) for CEC.However, there were no differences in AUCs for WT (AUC = 0.635, p = 0.184) and WTC (AUC = 0.628, p = 0.184).The association of MTconst mean contrast values with therapy response of uncorrected and corrected tumor volumes did not show relevant differences (CE: AUC = 0.826, p = 0.001; CEC: AUC = 0.816, p = 0.002; WT: AUC = 0.868, p < 0.001; WTC: AUC = 0.861, p < 0.001).The MTRRexAPT (CE:

Supermedian Analysis of the Mean CEST Contrast Values of Uncorrected and Corrected Tumor Volumes
To understand the impact of the blood product correction on the tumor-associated mean CEST contrast values in greater detail, we also assessed how many participants switched from the respective groups with higher mean CEST contrast values compared to the respective cohort medians (supermedian) to the groups with lower mean values (sub-

Supermedian Analysis of the Mean CEST Contrast Values of Uncorrected and Corrected Tumor Volumes
To understand the impact of the blood product correction on the tumor-associated mean CEST contrast values in greater detail, we also assessed how many participants switched from the respective groups with higher mean CEST contrast values compared to the respective cohort medians (supermedian) to the groups with lower mean values (submedian) and vice versa upon correcting tumor volumes for mHb and Hs.We observed that for CE C in comparison to CE, one participant switched from super-to submedian and one participant from sub-to supermedian for MTR Rex APT, MTR Rex MT and APTw asym .Additionally, for WT C in comparison to WT one participant switched from super-to submedian and one participant from sub-to supermedian for MTR Rex MT.However, for the MT const , no participants switched from super-to sub-or sub-to supermedian upon correcting any tumor volume for blood products.The results of this analysis with corresponding tumor mean CEST contrast values for the assessed participants and respective cohort medians are summarized in Table S1.

Discussion
Whilst several groups demonstrated the potential of CEST imaging in the diagnostic follow-up after hemorrhagic stroke, little is known about the influence of post-therapeutic depositions of blood breakdown products on the clinical performance of APT and ssMT imaging in the early post-radiotherapy interval.For this reason, the purpose of this study was to assess the impact of advanced correction methods for mHb and Hs on the association of most commonly employed APT and ssMT contrasts with therapy response and OS at the first follow-up 4 to 6 weeks after the completion of radiotherapy at 3T.Even though, the MTR Rex APT contrast maps showed markedly elevated values in correspondence with Hs, and the MTR Rex APT, MTR Rex MT and MT const contrast maps showed noticeably dropped values in correspondence with mHb, no relevant differences in mean contrast values between uncorrected and corrected tumor volumes could be detected.However, for corrected contrast-enhancing tumor volumes, a slightly stronger association of the MTR Rex MT with therapy response was observed, whilst the MTR Rex APT showed a moderately weaker association with survival.Interestingly, the APTw asym showed contradicting trends, with a somewhat weaker association with survival for corrected contrast-enhancing tumor volumes, and a slightly stronger association with survival for corrected whole tumor volumes.Concurrently, the association of MT const mean values with therapy response and survival were unaffected by the correction.
In previous studies, Sawaya et al., Wang et al. and Ma et al. observed that asymmetrybased APTw imaging showed markedly elevated contrast values in rat models and in patients with acute and subacute cerebral bleeding, which very likely corresponded to accumulations of deoxygenized hemoglobin and mHb [13,19,20,[34][35][36].Lai et al., on the other hand, observed significantly reduced contrast values in the subacute stage of cerebral hemorrhage in a preclinical study, using an apparent exchange-dependent relaxation compensated metric of the APT (APT AREX ) [21].Contrary to these findings, to our knowledge, there are no available published results on the CEST contrast behaviors of Hs.
The findings of Lai et al. are mirrored by decreased MTR Rex APT, MTR Rex MT and MT const values for mHb in comparison to corrected tumor volumes, which were observed on this study (Figure 2).Given that mHb contains paramagnetic Fe 3+ , the observed contrast patterns might at least in part be explained by residual T 1 contributions especially to the MT const , but to a lesser extent also to the other investigated CEST contrasts [5][6][7]22,34].Since magnetization transfer between protons and free water through chemical exchange is base-triggered, pH might be another factor that could influence particularly APT-weighted CEST contrasts [5,22].The visibly increased values on MTR Rex APT contrast maps that corresponded to Hs, on the other hand, are harder to explain (Figure 2).Hs consists of intracellular accumulations of insoluble and partially digested ferritin, which should be associated with a rather acidotic intralysosomal milieu, T 1 contributions from paramagnetic Fe 3+ and fewer mobile amide protons due to the insoluble state of the proteinaceous compounds [34].Taken together, further research is needed to understand the physicochemistry behind the observed CEST contrast patterns of mHb and Hs.
Despite these observations and considerations, no relevant differences in the investigated CEST contrast mean values could be observed between the uncorrected and corrected tumor volumes.Concurrently, the correction only had a very minor impact on the association of the MTR Rex MT with therapy response and of the MT Rex APT and APTw asym with survival in this relatively small clinical cohort of 34 participants.Even though the contribution of mHb and Hs depositions to the contrast behavior of uncorrected and corrected tumor volumes could not be quantified, it seems reasonable to speculate that their amount was simply too small over the whole cohort to produce relevant effects.This might implicate that whilst larger mHb and Hs depositions are very visible, especially on MTR Rex APT contrast maps, advanced correction methods for the evaluation of CEST contrasts in the post therapeutic setting in representative clinical cohorts of patients with glioma could be of secondary relevance.
The relatively small cohort size, the subjective determination of mHb and Hs on T 1 w and T 2 *w imaging, and the lack of corresponding histopathological data for the assessed blood products are the major limitations of this study.Even though histopathological confirmation of remaining blood products in the tumor area is impossible to obtain, future studies assessing the impact of mHb and Hs on the clinical performance of CEST imaging in the early post-therapeutic interval might benefit from larger sample sizes (e.g., in multicenter trials) and support from AI-based automated segmentation tools for the definition of specific blood products.

Conclusions
A sophisticated correction for methemoglobin and hemosiderin did not substantially alter the clinical performance of APT and ssMT imaging at the first follow-up 4 to 6 weeks after the completion of radiotherapy in 34 participants with glioma.Larger blood product depositions were visible on APT and ssMT contrast maps and had minor effects on the clinical performance of the MTR Rex MT regarding therapy response assessment, and on that of the MTR Rex APT and APTw asym regarding patient outcome prediction.

Figure 1 .
Figure 1.Flowchart.Displayed is a description of the whole study cohort participants and the 34 participants that were eventually included in the analysis.Between September 2018 and December 2021, 72 participants with diffuse glioma received CEST imaging 4 to 6 weeks after the completion of radiotherapy.In total, 11 participants had to be excluded from the analysis due to the indicated reasons.Given that diffuse midline gliomas are biologically distinct from hemispherical gliomas[27,28]   and previous studies observed a dependency of CEST contrasts on the presence of residual tumorassociated contrast enhancement on MRI[6,11], the datasets of 34 participants with hemispherical gliomas, presence of residual contrast enhancement on MRI and available survival data were included in the analysis.The associations of mean CEST contrast values with therapy response and overall survival were tested by Kaplan-Meier analyses and log rank tests.KPI = Karnofsky Performance Score; w. = with; w/o.= without.

Figure 2 .
Figure 2. Exemplary contrast maps of two participants with larger methemoglobin depositions (mHb) in the tumor area (a,b).Given are T2* susceptibility-weighted (SWI), T1w, contrast-enhanced T1w (T1wCE) and T2w-FLAIR images, as well as CEST contrast maps of MTRRexAPT, MTRRexMT, APTwasym and MTconst imaging.The ROIs indicate the T1w-hyperintese mHb on all contrast maps.mHb visually showed markedly decreased values pronounced on MTRRexAPT, MTRRexMT and MTconst contrast maps.The figure highlights the visible depression of the investigated CEST contrasts in correspondence to larger mHb depositions.

Figure 2 .
Figure 2. Exemplary contrast maps of two participants with larger methemoglobin depositions (mHb) in the tumor area (a,b).Given are T 2 * susceptibility-weighted (SWI), T 1 w, contrast-enhanced T 1 w (T 1 w CE ) and T 2 w-FLAIR images, as well as CEST contrast maps of MTR Rex APT, MTR Rex MT, APTw asym and MT const imaging.The ROIs indicate the T 1 w-hyperintese mHb on all contrast maps.mHb visually showed markedly decreased values pronounced on MTR Rex APT, MTR Rex MT and MT const contrast maps.The figure highlights the visible depression of the investigated CEST contrasts in correspondence to larger mHb depositions.

Figure 3 .
Figure 3. Exemplary contrast maps of two participants with larger hemosiderin depositions (Hs) in the tumor area (a,b).Given are T 2 * susceptibility-weighted (SWI), T 1 w, contrast-enhanced T 1 w (T 1 w CE ) and T 2 w-FLAIR images, as well as CEST contrast maps of MTR Rex APT, MTR Rex MT, APTw asym and MT const imaging.The ROIs indicate Hs (dark on SWI) on all contrast maps.The MTR Rex APT visually showed markedly elevated contrast values corresponding sharply to Hs in these participants.The figure highlights the visible increase, especially of the MTR Rex APT in correspondence to larger Hs depositions.

Figure 4 .
Figure 4. Mean CEST contrast values of uncorrected and corrected tumor volumes.Given are violin plots for MTRRexAPT, MTRRexMT, APTwasym and MTconst mean contrast values for contrast-enhancing and whole tumor volumes without (CE and WT) and with (CEC and WTC) correction for mHb and Hs.The figure highlights that there were no relevant differences between the CEST contrast values of tumor volumes that were uncorrected and corrected for methemoglobin (mHb) and hemosiderin (Hs) depositions.

Figure 4 .Figure 5 .Figure 5 .Figure 6 .
Figure 4. Mean CEST contrast values of uncorrected and corrected tumor volumes.Given are violin plots for MTR Rex APT, MTR Rex MT, APTw asym and MT const mean contrast values for contrastenhancing and whole tumor volumes without (CE and WT) and with (CE C and WT C ) correction for mHb and Hs.The figure highlights that there were no relevant differences between the CEST contrast values of tumor volumes that were uncorrected and corrected for methemoglobin (mHb) and hemosiderin (Hs) depositions.

Figure 6 .
Figure 6.Association of the mean CEST contrast values for uncorrected and corrected tumor volumes with survival.The figure shows Kaplan-Meier plots displaying the association of MTR Rex APT, MTR Rex MT, APTw asym and MT const mean contrast values of uncorrected (CE and WT) and corrected (CE C and WT C ) tumor volumes with overall survival.(a) shows the plots for contrast-enhancing tumor volumes (CE and CE C ).(b) shows the plots for whole tumor volumes (WT and WT C ).The survival of participants with mean values below the group medians is indicated by dark blue graphs for uncorrected tumor volumes (CE and WT) and by turquoise graphs for corrected tumor volumes (CE C and WT C ).The survival of participants with mean values above the group medians is indicated by brown graphs for uncorrected tumor volumes (CE and WT) and by orange graphs for corrected tumor volumes (CE C and WT C ). Statistically significant results (p ≤ 0.05) are indicated with an asterisk (*).The figure highlights that there were only marginal differences in the association of CEST contrast values with overall survival that mainly affected the MTR Rex APT and the APTw asym .

Figure A1 .Figure A2 .
Figure A1.Segmentation of uncorrected and corrected tumor volumes.Displayed are representative ROIs of methemoglobin (mHb), hemosiderin (Hs), contrast-enhancing tumor (CE), corrected contrastenhancing tumor (CE C ), whole tumor and corrected whole tumor (WT C ) in two representative participants with larger depositions of mHb (a) and Hs (b).Depicted are mHb on T 1 w imaging, Hs on susceptibility-weighted imaging (SWI), contrast-enhancing tumor volumes without (CE) and with correction for mHb and Hs (CE C ) on contrast-enhanced T 1 w (T 1 w CE ) images, as well as whole tumor volumes (WT) without and with correction for mHb and Hs (WT C ) on T 2 w-FLAIR images.WT encompassed contrast-enhancing tumor tissue and T 2 w-FLAIR-hyperintense tissue changes.Larger mHb depositions that were identifiable as such due to localization in the resection cavity or larger surrounding Hs deposition with marked T 2 w-FLAIR signal drop were grossly excluded for CE and WT.CE C and WT C were calculated from CE and WT in MATLAB ® (Mathworks, version 2019b, MA, USA) by subtracting overlapping mHb and Hs.The figure illustrates how the investigated blood products and uncorrected tumor volumes were defined on T 1 w, SWI, T 1 w CE and T 2 w-FLAIR imaging in an exemplary participant with coexisting depositions of mHb and Hs and highlights the differences between uncorrected and corrected tumor volumes.

Figure A2 .
Figure A2.T1 maps of exemplary participants with larger methemoglobin and hemosiderin depositions.Displayed are the T1 maps for the four participants with larger methemoglobin (mHb; Subfigures (a,b)) and hemosiderin (Hs; (c,d)) depositions in the tumor area displayed in Figures 2 and 3. Given are T 2 * susceptibility-weighted (SWI), T 1 w, contrast-enhanced T 1 w (T 1 wCE) and T 2 w-FLAIR images, as well as quantitative T 1 contrast maps.The ROIs indicate the T 1 w-hyperintese mHb (a,b) and Hs (dark on SWI; (c,d)) on all contrast maps.mHb visually showed markedly decreased T 1 values.Hs visually showed diffusely elevated T 1 values (d).The figure highlights the T 1 contribution of mHb and Hs depositions, which especially affect MT const , as well as MTR Rex APT and MTR Rex MT imaging.

Table 1 .
Clinical characteristics.Displays a summary of the most relevant clinical characteristics for all 34 evaluated study participants with glioma.

Table A3 .
Association of mean CEST contrast values with survival for uncorrected and corrected tumor volumes.