Search

Background Survivors of pediatric central nervous system (CNS) tumors treated with craniospinal irradiation (CSI) exhibit long-term cognitive difficulties. Goals of this study were to evaluate longitudinal effects of candidate and novel genetic variants on cognitive decline following CSI. Methods Intelligence quotient (IQ), working memory (WM), and processing speed (PS) were longitudinally collected from patients treated with CSI (n = 241). Genotype-by-time interactions were evaluated using mixed-effects linear regression to identify common variants (minor allele frequency > 1%) associated with cognitive performance change. Novel variants associated with cognitive decline (P Results Participants were mostly males (65%) diagnosed with embryonal tumors (98%) at a median age of 8.3 years. Overall, 1150 neurocognitive evaluations were obtained (median = 5, range: 2–10 per participant). One of the five loci previously associated with cognitive outcomes in pediatric CNS tumors survivors demonstrated significant time-dependent IQ declines (PPARA rs6008197, P = .004). Two variants associated with IQ in the general population were associated with declines in IQ after Bonferroni correction (rs9348721, P = 1.7 × 10−5; rs31771, P = 7.8 × 10−4). In genome-wide analyses, we identified novel loci associated with accelerated declines in IQ (rs116595313, meta-P = 9.4 × 10−9), WM (rs17774009, meta-P = 4.2 × 10−9), and PS (rs77467524, meta-P = 1.5 × 10−8; rs17630683, meta-P = 2.0 × 10−8; rs73249323, meta-P = 3.1 × 10−8). Conclusions Inherited genetic variants involved in baseline cognitive functioning and novel susceptibility loci jointly influence the degree of treatment-associated cognitive decline in pediatric CNS tumor survivors.

INTRODUCTION Hyperpolarized carbon-13 (HP-13C) MR enables rapid dynamic imaging of metabolic pathways in the human brain using non-toxic, non-radioactive metabolites as tracers. This study presents our unique experience on the benefit of using [1-13C]pyruvate and [2-13C]pyruvate MR for evaluating patients with glioma. METHODS 132 scans (71 using an integrated 13C /1H coil) including steady-state 1H-MRSI (~10 min) and dynamic HP-13C imaging (60 sec) following injection of HP [1-13C]pyruvate (N=125) or [2-13C]pyruvate (N=7) were acquired from 46 patients with glioma (18F/28M; 15 IDH-mutant, 27 IDH-wildtype, 4 IDH-status-unknown). Maps of temporally-summed 13C-metabolite signals, ratios, and kinetic rate constants were calculated for contrast-enhancing, nonenhancing, and normal-appearing-white-matter (NAWM) regions and compared to steady-state metabolic metrics. RESULTS The only adverse event, in a single patient, was a burning sensation after the injection that resolved after saline flush. The mean time-to-injection of HP probes was 58.6±14.0 sec. Signal-to-noise ratios of [1-13C]lactate and [13C]bicarbonate within the NAWM from the HP [1-13C]pyruvate data were 53±39 and 13±6, respectively. The SD/mean of repeated injections (N=3) for lactate/pyruvate and pyruvate-to-lactate conversion rates were 3.8±3.1% and 6.5±3.1% in the NAWM, respectively. Patients with progressive GBM had significantly higher lactate/pyruvate and lower bicarbonate/lactate (p< 0.05) in contrast- and non-enhancing lesions compared to NAWM. Significantly elevated lactate/pyruvate and reduced bicarbonate/lactate (p< 0.01) were found in contrast-enhancing compared to nonenhancing regions, whereas choline/NAA and steady-state 1H-lactate levels were similar. HP [2-13C]pyruvate data showed reduced glutamate/pyruvate and pyruvate-to-glutamate conversion rates in T2 lesions compared to contralateral normal-appearing brain in IDH-mutant gliomas, consistent with known metabolic reprogramming. CONCLUSION This study demonstrates the potential benefit of dynamic HP-13C MRI for evaluating patients with glioma, which provides unique and spatially distinct contrast compared to steady-state metabolic imaging. Ongoing studies will further characterize dynamic metabolism in specific glioma subtypes and provide biomarkers for evaluating responses to treatment.

Although physiologic (diffusion-weighted and perfusion-weighted) MRI has shown promise in identifying regions of recurrent tumor (rTumor) in patients with glioblastoma suspected of progression, distinguishing treatment-induced effects (TxE) from rTumor on anatomical MRI remains a challenge. Whereas prior larger-scale machine learning (ML)-based studies mostly utilize anatomical imaging alone and/or perform lesion-level predictions, this study aimed to develop a non-invasive, radiopathomic tool for regional probabilistic mapping of TxE using 208 tissue-samples (55 pathologically-confirmed TxE, 153 recurrent glioblastoma) acquired from 107 patients with known spatial coordinates on pre-surgical MRI. We tested the hypothesis that applying a deep-learning (DL) model that included physiological MRI can: 1) more accurately identify areas of TxE that mimic rTumor on anatomical MRI and 2) better generalize to an independent test set than ML-models or a DL-model that uses anatomical MRI alone. An 80/20 split for training/validation was used after 1/3 of the patients were withheld for testing. Oversampling of TxE samples was employed to address class imbalance and an equal proportion of TxE samples was maintained across all datasets. Three ML-models, their ensemble, and a deep 4D-convolutional-neural-network were trained based on normalized anatomical (post-contrast T1, T2-FLAIR), diffusion-weighted (ADC, FA), and DSC perfusion-weighted (PeakHeight, %recovery) images cropped to 10mm-cubic patches centered on the coordinates from where tissue was obtained. Although Random Forest and voting-ensembled ML-models using all imaging and the anatomical DL-model had the best validation performance (AUC=0.81-0.82), these models did not generalize (test AUC=0.58-0.59). The DL-model including physiologic images had slightly lower validation AUC (0.78) but the best overall test AUC (0.795), indicating superior generalizability. Elevated blood volume (nPeakHeight) was the most important feature. Our DL-model’s interpretability was also demonstrated by disrupting class separation after shuffling voxels within each input patch. These results suggest that using deep-learning with physiologic MRI can improve intratumoral classification of TxE from rTumor.

INTRODUCTION Pathologically aggressive tumor biology can extend beyond the contrast-enhancing or non-enhancing anatomical lesions in patients with glioma. Identification of malignant regions can help guide diagnosis and subsequent treatment planning. This study leverages a unique multi-parametric MRI dataset with tissue samples of known spatial coordinates to noninvasively predict cellular proliferation (KI-67) and a novel index of tumor aggressiveness (TAI), that combines proliferation, cellularity, and tumor-score. METHODS 420 tissue samples were collected from 162 patients with newly-diagnosed glioma (47% IDH-wildtype). Clinical imaging consisted of T2-weighted, T2-FLAIR, T1-weighted pre- and post-contrast images, and apparent diffusion coefficient (ADC) and fractional anisotropy (FA) from diffusion-weighted imaging. Mean normalized imaging metrics were quantified from 5mm spheres centered at the location of the tissue sample. A single spectrum was reconstructed at the location of each tissue sample from 3D 1H-MR Spectroscopic Imaging (MRSI) before quantifying normalized metabolite peak-heights for choline, creatine, NAA, lactate/lipid, and relative indices. Univariate mixed-effects linear regression models were employed and features with p< 0.2 were selected for subsequent model building. Support vector machine (SVM), random forest, and gradient boosting machine-learning algorithms were trained and tested on a ⅔-⅓ train-test split with 4-fold cross-validation in training to predict a high/low KI-67 and TAI. RESULTS Although none of the individual imaging metrics were significantly associated with KI-67 in the univariate analysis, all diffusion and several MRSI metrics (ncholine, nNAA, CNI, excess choline and creatine) were significantly associated with cellularity. Preliminary multivariate analyses to date suggest that the best radiopathomic model performance is achieved when an SVM was used along with T1-precontrast, nADC, and all metabolite levels (mean cross-validation AUC=0.73 and accuracy=.77). CONCLUSION Our results suggest that multi-parametric physiologic and metabolic MRI are useful for radiopathomic-mapping of tumor aggressiveness and are currently being optimized in a larger cohort.

BackgroundDiagnostic classification of diffuse gliomas now requires an assessment of molecular features, often including IDH-mutation and 1p19q-codeletion status. Because genetic testing requires an invasive process, an alternative noninvasive approach is attractive, particularly if resection is not recommended. The goal of this study was to evaluate the effects of training strategy and incorporation of biologically relevant images on predicting genetic subtypes with deep learning.MethodsOur dataset consisted of 384 patients with newly diagnosed gliomas who underwent preoperative MRI with standard anatomical and diffusion-weighted imaging, and 147 patients from an external cohort with anatomical imaging. Using tissue samples acquired during surgery, each glioma was classified into IDH-wildtype (IDHwt), IDH-mutant/1p19q-noncodeleted (IDHmut-intact), and IDH-mutant/1p19q-codeleted (IDHmut-codel) subgroups. After optimizing training parameters, top performing convolutional neural network (CNN) classifiers were trained, validated, and tested using combinations of anatomical and diffusion MRI with either a 3-class or tiered structure. Generalization to an external cohort was assessed using anatomical imaging models.ResultsThe best model used a 3-class CNN containing diffusion-weighted imaging as an input, achieving 85.7% (95% CI: [77.1, 100]) overall test accuracy and correctly classifying 95.2%, 88.9%, 60.0% of the IDHwt, IDHmut-intact, and IDHmut-codel tumors. In general, 3-class models outperformed tiered approaches by 13.5%-17.5%, and models that included diffusion-weighted imaging were 5%-8.8% more accurate than those that used only anatomical imaging.ConclusionTraining a classifier to predict both IDH-mutation and 1p19q-codeletion status outperformed a tiered structure that first predicted IDH-mutation, then 1p19q-codeletion. Including apparent diffusion coefficient (ADC), a surrogate marker of cellularity, more accurately captured differences between subgroups.

INTRODUCTION Noninvasive, radiopathomic mapping of tumor aggressiveness can benefit patients with glioma by guiding the selection of tissue samples for diagnosis, increasing extent of resection, and non-invasively characterizing residual tumor burden for subsequent treatment. Although prior studies have demonstrated the utility of metabolic metrics quantified from 1H-MR Spectroscopy (MRS) in probing tumor pathology, this study evaluated the benefit of using the entire 1D-spectrum and deep learning for radiopathomic mapping of intratumoral cellularity, proliferation (ki-67), and a new tumor aggressiveness index (TAI) defined as log((n(ki−67)+n(cellularity))*tumor-score). METHODS Multi-voxel 1H-MRS was acquired on 281 patients newly diagnosed with a glioma (47% IDH-wildtype) immediately before surgical resection. After reconstructing individual spectra at the locations where tissue samples were obtained during surgery and normalizing by NAA in contralateral normal-appearing-white-matter, 607 spectra with corresponding histopathology were deemed of sufficient quality for analysis. A 1D convolutional-neural-network with bidirectional long- and short-term memory deep-learning model using the entire spectrum (0.6-3.6ppm) was compared to mixed-effects regression (with choline-to-NAA index[CNI]) and Random Forest (with CNI+normalized peak heights) models for predicting ki-67, cellularity, and TAI. Results & DISCUSSION Using deep-learning on the entire spectrum resulted in 10.3%-22.1% lower mean absolute error (MAE) and 0.32-0.37 higher R2 values compared to using CNI alone or a random forest model with multiple metabolic metrics. MAE values for all 3 deep-learning models were 26-44% < 1 standard deviation of the ground truth, demonstrating reasonable prediction accuracy within the test data set. Although the lowest MAE (0.16) and highest R2 (0.41) was attained when predicting TAI with deep-learning, the prediction of cellularity resulted in the lowest %MAE. Colormaps of predicted pathology identified regions of heightened aggressiveness surrounding tissue samples with most abnormal pathological features that sometimes extended beyond the non-enhancing lesion. Current work is evaluating the clinical utility of our deep-learning model and predicted maps of aggressiveness.

BACKGROUND Monitoring lower-grade gliomas (LrGGs) for disease progression is made difficult by the limits of anatomical MRI to distinguish treatment-related tissue changes from tumor progression. MR spectroscopic imaging (MRSI) offers additional metabolic information that can help address these challenges. The goal of this study was to compare longitudinal changes in multiparametric MRI, including diffusion-weighted imaging, perfusion imaging, and 3D MRSI, for LrGG patients who progressed at the final time point and those who remained clinically stable. METHODS Forty-one patients with LrGG who were clinically stable were longitudinally assessed for progression. Changes in anatomical, diffusion, perfusion, and MRSI data were acquired and compared between patients who remained clinically stable and those who progressed. RESULTS Thirty-one patients remained stable, and 10 patients progressed. Over the study period, progressed patients had a significantly greater increase in normalized choline, choline-to-N-acetylaspartic acid index (CNI), normalized creatine, and creatine-to-N-acetylaspartic acid index (CRNI), than stable patients. CRNI was significantly associated with progression status and WHO type. Progressed astrocytoma patients had greater increases in CRNI than stable astrocytoma patients. CONCLUSIONS LrGG patients in surveillance with tumors that progressed had significantly increasing choline and creatine metabolite signals on MRSI, with a trend of increasing T2 FLAIR volumes, compared to LrGG patients who remained stable. These data show that MRSI can be used in conjunction with anatomical imaging studies to gain a clearer picture of LrGG progression, especially in the setting of clinical ambiguity.

BACKGROUND: Children with medulloblastoma are at risk of hearing loss (HL) following cranial radiotherapy and platinum-based chemotherapy. Keilty et al. 2021. J Clin Oncol. (PMID: 34570616) recently published a computer-based HL risk calculator based on children with central nervous system tumors treated with photon radiotherapy. The objective of this study was to evaluate the performance of the risk calculator in an independent cohort of children with medulloblastoma treated with photon radiotherapy. METHODS: We identified children with medulloblastoma treated with cisplatin and photon radiotherapy at Texas Children’s Hospital between 1997-2007. Audiograms were graded using the International Society of Pediatric Oncology-Boston scale (SIOP-Boston). The risk calculator considered age at treatment, time since radiotherapy, cumulative cisplatin dose (mg/m2), and cochlea radiation dose (Gy). The calculated risk was compared between ears with and without clinical evidence of grades 3/4 HL using t-test. RESULTS: We identified 45 eligible patients (male: 31, female: 14) diagnosed at a median age of nine years (range: 3-18 years). The mean cochlea dose was 37.2 Gy (range: 25.2-54.9 Gy) and mean cumulative cisplatin dose was 352 mg/m2 (range: 55-860 mg/m2). Audiograms of 89 total ears were available with a median follow-up of 65 months post-diagnosis (range: 13-163 months). The median number of post-radiotherapy audiograms per patient was 6 (range: 2-15). A total of 22 ears (25%) belonging to 13 individual patients (29%) met the SIOP-Boston criteria for grades 3/4 HL. The mean of predicted probability of severe HL based on the calculator for ears with SIOP-Boston grades 3/4 was 24% (range: 10% - 48%) compared to 32% (range: 10%-70%) among ears with grades < 3 HL. CONCLUSION: The published HL risk calculator did not clearly differentiate between patients with and without clinically significant HL in our cohort. Future multicenter studies are required to validate this tool and improve the prognostic potential.³

BackgroundDifferentiating treatment-induced injury from recurrent high-grade glioma is an ongoing challenge in neuro-oncology, in part due to lesion heterogeneity. This study aimed to determine whether different MR features were relevant for distinguishing recurrent tumor from the effects of treatment in contrast-enhancing lesions (CEL) and non-enhancing lesions (NEL).MethodsThis prospective study analyzed 291 tissue samples (222 recurrent tumor, 69 treatment-effect) with known coordinates on imaging from 139 patients who underwent preoperative 3T MRI and surgery for a suspected recurrence. 8 MR parameter values were tested from perfusion-weighted, diffusion-weighted, and MR spectroscopic imaging at each tissue sample location for association with histopathological outcome using generalized estimating equation models for CEL and NEL tissue samples. Individual cutoff values were evaluated using receiver operating characteristic curve analysis with 5-fold cross-validation.ResultsIn tissue samples obtained from CEL, elevated relative cerebral blood volume (rCBV) was associated with the presence of recurrent tumor pathology (P < 0.03), while increases in normalized choline (nCho) and choline-to-NAA index (CNI) were associated with the presence of recurrent tumor pathology in NEL tissue samples (P < 0.008). A mean CNI cutoff value of 2.7 had the highest performance, resulting in mean sensitivity and specificity of 0.61 and 0.81 for distinguishing treatment-effect from recurrent tumor within the NEL.ConclusionAlthough our results support prior work that underscores the utility of rCBV in distinguishing the effects of treatment from recurrent tumor within the contrast enhancing lesion, we found that metabolic parameters may be better at differentiating recurrent tumor from treatment-related changes in the NEL of high-grade gliomas.

Although current advances for automated glioma lesion segmentation and volumetric measurements using deep learning have yielded high performance on newly-diagnosed patients, response assessment in neuro-oncology still relies on manually-drawn, cross-sectional areas of the tumor because these models do not generalize to patients in the post-treatment setting, where they are most needed in the clinic. Surgical resections, adjuvant treatment, or disease progression can alter the characteristics of these lesions on T2-weighted imaging, causing measures of segmentation accuracy, typically measured by Dice coefficients of overlap (DCs), to drop by ~15%. To improve the generalizability of T2-lesion segmentation to patients with glioma post-treatment, we evaluated the effects of: 1) training with different proportions of newly-diagnosed and treated gliomas, 2) applying transfer learning from pre- to post-treatment domains, and 3) incorporating a loss term that spatially weights the lesion boundaries with greater emphasis in training. Using 425 patients (208 newly-diagnosed, 217 post-Tx, with 25 treated patients withheld as a test set) and a top-performing model previously trained on newly-diagnosed gliomas, we found that DCs increased by 10% (to 0.84) then plateaued after including ~25% of post-treatment patients in training. Transfer learning (pre-training on newly-diagnosed and finetuning with post-treatment data) significantly improved Hausdorf distances (HDs), a measure more sensitive to changes at the lesion boundaries, by 17% after including 26% post-treatment images in training, while DCs remained similar. Although modifying our loss functions with boundary-weighted penalizations resulted in comparable DCs to using standard DC loss, HD measures were further reduced by 26%, suggesting that HDs may be a more sensitive metric to subtle changes in segmentation accuracy than DCs. Current work is evaluating their utility in providing accurate volumes for real-time response assessment in the clinic using workflows that have recently been deployed on our clinical PACs system.

INTRODUCTION The goal of this study was to characterize progressive and pseudoprogressive GBM using multi-parametric hyperpolarized (HP)-13C / 1H MRI. METHODS Dynamic HP-13C MRI was acquired from 13 patients with progressive GBM [patients (scans): 2(3) IDH-mutant; 11(13) IDH-wildtype] and 2 IDH-wildtype patients (3 scans) demonstrating pseudo-progression following intravenous injection of HP [1-13C]pyruvate. Frequency-selective echo-planar imaging (3s temporal resolution, 3.38 cm3 spatial resolution) captured [1-13C]pyruvate metabolism to [1-13C]lactate and 13C-bicarbonate in the brain. Dynamic 13C data were kinetically modeled to obtain the pyruvate-to-lactate conversion rate constant k PL and temporally summed to calculate 13C-metabolite percentiles and ratios (linearly interpolated 2x in-plane). 1H imaging included T2, post-Gd T1, perfusion (nCBV, %recovery), diffusion (ADC), and lactate-edited spectroscopy (CNI, choline-to-NAA index; 1H-lactate). The normal-appearing white matter (NAWM), non-enhancing lesion (NEL), and contrast-enhancing lesion (CEL) were segmented from 1H images. 13C-resolution masks were iteratively applied on a voxel-wise basis to evaluate 1H imaging parameters within each ROI and multi-parametric data were collectively evaluated using a mixed effects model in R. RESULTS Progressive IDH-mutant GBM compared to wildtype counterparts displayed increased perfusion %recovery (p < 0.001) and k PL (p < 0.01), together with reduced 1H-lactate (p < 0.001) and pyruvate percentile (p < 0.01), in the T2 lesion. Among IDH-wildtype progressive GBM, the CEL was distinguished from NEL/NAWM by increased nCBV (p < 0.05/0.001), 1H-lactate (p < 0.05/0.001); and decreased bicarbonate / lactate (p < 0.05/0.001). The CEL and NEL were collectively distinguished from NAWM by elevated CNI (p < 0.001/0.001), ADC (p < 0.05/0.001), pyruvate percentile (p < 0.001/0.001), lactate percentile (p < 0.001/0.001), and relative lactate / pyruvate (p < 0.001/0.05). Psuedo-progressive IDH-wildtype GBM displayed lower k PL (T2 Lesion; p < 0.01) and nCBV (CEL; p < 0.01) compared to progressive GBM. CONCLUSION HP-13C parameters can potentially augment proton imaging and demonstrated Warburg-associated metabolic alterations.

Physiologic changes quantified by diffusion and perfusion MRI have shown utility in predicting treatment response in glioblastoma (GBM) patients treated with cytotoxic therapies. We aimed to investigate whether quantitative changes in diffusion and perfusion after treatment by immune checkpoint inhibitors (ICIs) would determine 6-month progression-free survival (PFS6) in patients with recurrent GBM.Inclusion criteria for this retrospective study were: (i) diagnosis of recurrent GBM treated with ICIs and (ii) availability of diffusion and perfusion in pre and post ICI MRI (iii) at ≥6 months follow-up from treatment. After co-registration, mean values of the relative apparent diffusion coefficient (rADC), Ktrans (volume transfer constant), Ve (extravascular extracellular space volume) and Vp (plasma volume), and relative cerebral blood volume (rCBV) were calculated from a volume-of-interest of the enhancing tumor. Final assignment of stable/improved versus progressive disease was determined on 6-month follow-up using modified Response Assessment in Neuro-Oncology criteria.Out of 19 patients who met inclusion criteria and follow-up (mean ± SD: 7.8 ± 1.4 mo), 12 were determined to have tumor progression, while 7 had treatment response after 6 months of ICI treatment. Only interval change of rADC was suggestive of treatment response. Patients with treatment response (6/7: 86%) had interval increased rADC, while 11/12 (92%) with tumor progression had decreased rADC (P = 0.001). Interval change in rCBV, Ktrans, Vp, and Ve were not indicative of treatment response within 6 months.In patients with recurrent GBM, interval change in rADC is promising in assessing treatment response versus progression within the first 6 months following ICI treatment.• In recurrent GBM treated with ICIs, interval change in rADC suggests early treatment response.• Interval change in rADC can be used as an imaging biomarker to determine PFS6.• Interval change in MR perfusion and permeability measures do not suggest ICI treatment response.

INTRODUCTION Radiation therapy (RT) remains an integral role in the treatment of adolescent brain tumors, despite evidence of its long-term effects including cognitive impairment, vascular injury and reduced white matter integrity. While prior studies have related vascular injury to cognitive decline and associated more severe cognitive impairment with a whole-brain versus a focal RT approach, the relationship between underlying imaging, clinical, and treatment parameters has yet to be explored in this population. In this study we used multimodal 7 Tesla MR imaging to probe RT-induced changes in the brain and identified risk factors for clinical outcome. METHODS Twenty-three patients (age 6–25 years) with non-supratentorial tumors treated with RT as children and 4 nonirradiated control patients (ages 13–16 years) were scanned on a 7T MRI system; eight patients underwent serial imaging 0.9–3.7 years following the first scan. Simultaneous MR-veniography and angiography, and 90-direction, dual-shell multi-band diffusion MRI were used to assess the relationships among cerebral microbleed (CMB) development, changes in arterial radii, and whole-brain white matter connectivity. A computerized cognitive battery (Cogstate) evaluated multiple domains of cognitive function. Multiple univariate and multivariate regression models with multiple comparison corrections identified risk factors. RESULTS Cognitive status measured via executive function and working memory tasks revealed the strongest associations with type of RT and imaging parameters. Specific risk factors for worse outcome included whole-brain RT, RT at a younger age, and time since RT. On imaging this corresponded to increased CMB burden, decreased arterial volume, and reduced global structural connectivity; intrasubject serial imaging followed these trends. CONCLUSION 7T-MRI was highly sensitive to CMBs with cumulative incidence rates greatly exceeding prior 3T studies in this population. This work demonstrates the value of multimodal 7T-MRI in providing metrics that reflect cognitive deficits arising from RT and identifying patients who would benefit the most from cognitive rehabilitation.