Citations

Abstract PURPOSE Brain tumor related epilepsy (BTE) is a major co-morbidity in patients with glioma. It is difficult to determine whether the use of anti-epileptic drugs is necessary. We attempted to build a machine-learning model to predict the probability of seizure presentation (SP) with glioma. METHODS We trained a random forest classifier using the following variables: volumetric data of pre-treatment MR images (T1Gd and T2-FLAIR sequences), patient demographics (age; sex), and measurements of tumor proliferation (log(ρ)), invasiveness (log(D)) and their relative ratio (log(ρ/D)). Our cohort consisted of 221 patients total. Using an 80-20 ratio, we used 176 patients (76 SP, 100 nSP) for training and the remaining 45 patients (19 SP, 26 nSP) were used for testing. We also trained on male-only and female-only cohorts to evaluate any sex differences in prediction. For training, 108 males (53 SP, 55 nSP) were used and 28 for testing (14 SP, 14 nSP). We used 72 females (21 SP, 49 nSP) for training and 15 (7 SP, 8 nSP) for testing. We corrected for class imbalance in the female cohort before training. Using 10-fold cross-validation and a separate testing set, we measured performance by ROC curve (AUC), accuracy, sensitivity, and specificity of predictions (average of folds in cross validation). RESULTS The female model achieved the highest AUC (0.853) followed by the mixed model (0.726) and the male model (0.651). In the validation set, the accuracy/sensitivity/specificity of the three cohorts were as follows: mixed (0.726/0.696/0.750), female (0.853/0.830/0.875), and male (0.651/0.577/0.722). The performance of the testing set, in terms of accuracy/sensitivity/specificity were: mixed (0.733/0.74/0.73), female (0.8/0.57/1), and male (0.714/0.64/0.79). CONCLUSION We found a negative correlation between seizure probability and size and invasiveness of tumors. Our model shows promising performance on testing set data. Further cohort studies and training is warranted.

Abstract Recurrent glioblastoma is challenging to distinguish from so called “treatment effect” on routine clinical imaging. Further, within tumor heterogeneity reveals that some regions can be histologically dominated by tumor progression whilst others can be dominated by secondary effects of treatment response. Apparent tumor progression on MRI can be very difficult to manage clinically as it is unclear the degree to which the imaging changes are actually tumor progression vs response to treatment (including inflammatory response and necrosis). In this analysis, we study a unique cohort of patients for whom image localized-biopsies reveal heterogeneity in response vs progression. Our dataset included 70 biopsy samples from 32 patients with GBM each histolopatholgically characterized for tumor abundance vs immune infiltrate. Six multiparametric MRI contrasts were available, including T1, T1gd, T2, FLAIR, SWI, and ADC. Images were co-registered. Radiomic (statistical + texture) features were extracted from the region of six image contrasts locally matched with each biopsy sample. Machine learning models were built to predict each biomarker using radiomic features. Leave-one-out cross validation was used to evaluate the prediction accuracy. Radiomic features were found to be informative to the prediction of biomarkers. ANOVA tests show significant improvement of using radiomic features compared with the null model. The prediction accuracy was higher when considering the biomarkers on a binary scale using the median as the cutoff than on a numerical scale. Spatially-informed radiomics models for tumor progression vs treatment effect are possible and can play an instrumental role in navigating confounding imaging changed common during treatment progression.

Abstract BACKGROUND An important challenge in radiomics research is reproducibility. Images are collected on different image scanners and protocols, which introduces significant variability even for the same type of image across institutions. In the present proof-of-concept study, we address the reproducibility issue by using domain adaptation – an algorithm that transforms the radiomic features of each new patient to align with the distribution of features formed by the patient samples in a training set. METHOD Our dataset included 18 patients in training with a total of 82 biopsy sample. The pathological tumor cell density was available for each sample. Radiomic (statistical + texture) features were extracted from the region of six image contrasts locally matched with each biopsy sample. A Gaussian Process (GP) classifier was built to predict tumor cell density using radiomic features. Another 6 patients were used to test the training model. These patients had a total of 31 biopsy samples. The images of each test patient were purposely normalized using a different approach, i.e., using the CSF instead of the whole brain as the reference. This was to mimic the practical scenario of image source discrepancy between different patients. Domain adaptation was applied to each test patient. RESULTS Among the 18 training patients, the leave-one-patient-out cross validation accuracy is 0.81 AUC, 0.78 sensitivity, and 0.83 specificity. When the trained model was applied to the 6 test patients (purposely normalized using a different approach than that of the training data), the accuracy dramatically reduced to 0.39 AUC, 0.08 sensitivity, and 0.61 specificity. After using domain adaption, the accuracy improved to 0.68 AUC, 0.62 sensitivity, and 0.72 specificity. CONCLUSION We provide candidate enabling tools to address reproducibility in radiomics models by using domain adaption algorithms to account for discrepancy of the images between different patients.

Abstract INTRODUCTION The quantification of intratumoral heterogeneity – through radiomics-based approaches - can help resolve the regionally distinct genetic drug targets that may co-exist within a single Glioblastoma (GBM) tumor. While this offers potential diagnostic value under the paradigm of individualized oncology, clinical decision-making must also consider the degree of uncertainty associated with each model. In this study, we evaluate the performance of a novel machine-learning (ML) algorithm, called Gaussian Process (GP) modeling, that can quantify the impact of multiple sources of uncertainty in ML model development and prediction accuracy, including variabilities in the copy number measurement, radiomics features, training sample characteristics, and training sample size. METHOD We collected 95 image-localized biopsies from 25 primary GBM patients. We coregistered stereotactic locations with preoperative multi-parametric MRI features (conventional MRI, DSC perfusion, Diffusion Tensor Imaging) to generate spatially matched pairs of MRI and copy number variants (CNV) for for each biopsy. We developed a Gaussian Process (GP) model to predict CNV for Epidermal Growth Factor Receptor (EGFR) based on MRI radiomic features in each patient. We used leave-one-patient-out cross validation to quantify prediction accuracy and model uncertainty. Spatial prediction and uncertainty (p-value) maps were overlaid to help visualize regional genetic variation of EGFR and uncertainty of the radiomic predictions. RESULT: The initial GP radiomics model for EGFR amplification (CNV > 3.5) produced a sensitivity of 0.8 and specificity of 0.8. Samples/regions associated with high uncertainty (p-value >0.05) correlated with either 1) extrapolation of radiomic features from the training set-defined feature space or 2) insufficient training samples in the feature space. CONCLUSION We present a ML-based model that quantifies spatial genetic heterogeneity in GBM, while also estimating model uncertainties that result from multi-source data variabilities. This approach lays the groundwork for prospective clinical integration of modeling-based diagnostic approaches in the paradigm of individualized medicine.

Abstract Glioblastomas (GBMs) are known to be complex tumors comprising multiple subpopulations of genetically-distinct cancer cells; it is thought that this genetic variation is a major factor in the lack of observed survival benefit of treatment regimes that target one of these subpopulations. The field of radiogenomics seeks to study correlations between MRI patterns and genetic features of GBM tumors. Spatial radiogenomic maps produced using machine-learning (ML) methods that are trained against information from image-localized patient biopsies identify regions where particular cancer sub-populations are predicted to occur within a GBM, thus non-invasively characterizing the regional genetic variability of these tumors. These tumor subpopulations may also interact with one another, in ways which may be of a competitive or cooperative nature to varying degrees. It is important to ascertain the nature of these interactions, as they may have implications for treatment response to targeted therapies, and characterization of the spatio-temporal dynamics of these co-evolving sub-populations will shed light on why some therapies fail. Here we combine mathematical modeling techniques and spatially-resolved radiogenomic maps to study the nature of these interactions between molecularly-distinct GBM subpopulations. We model the interactions between cell populations using a partial differential equation based formalism. The model is parameterized using radiogenomic ML maps from which we infer the nature of interactions between subpopulations. Furthermore, using maps as inputs, the model turns static maps into dynamic information, thus providing insight into how these subpopulations composing the tumor change over time and the effect this has on observed treatment response for individual patients.

Abstract Glioblastoma (GBM) is the most aggressive primary brain tumor with a median survival of only 15 months with standard of care treatment. Lacunarity, a quantitative morphological measure of how shapes fill space, and fractal dimension, another morphological measure of the complexity of pixel arrangement, of segmented necrotic regions on gadolinium-enhanced T1 weighted (T1gd) MRI have previously been shown to distinguish both overall survival (OS) and progression free survival (PFS) in GBM (n = 95). In our larger patient cohort (n = 389), we sought to validate or refute previously published results connecting morphological metrics and patient survival. We identified pretreatment necrotic regions of our retrospective first-diagnosis GBM patient cohort using segmented T1gd MRI enhancing regions. We calculated lacunarity and fractal dimension across all T1gd MRI slices with enhancing tumor, and used the median lacunarity and fractal dimension values for our analysis. We find that a lacunarity threshold can significantly distinguish OS (14 months vs 19 months median, log-rank p = 0.015, n = 389) and a fractal dimension threshold can significantly distinguish PFS (8 months vs 11 months median, log-rank p = 0.015, n = 123). We believe that morphological metrics such as lacunarity and fractal dimension could play a role in standard-of-care prognostic considerations at tumor presentation. This link between morphological and survival metrics could be driven by underlying biological phenomena or microenvironmental factors that should be further explored.

Abstract INTRODUCTION Glioblastoma (GBM) is the most common malignant primary brain tumor in adults with a median overall survival (OS) of 15months. Despite advancements in treatments, prognosis is dismal and the prognostic significance of tumor location is not entirely understood. METHODOLOGY: In our study, we investigated sex-specific volumetric, tumor growth kinetics, and outcome differences among GBMs in various brain locations. Primary GBM patients with pretreatment magnetic resonance imaging (MRI) data (N=289, 173 males, 116 females) were selected from our brain tumor repository. Tumor abnormality was segmented on T1-weighted post-gadolinium contrast agent (T1Gd) MRIs. We utilized the Harvard-Oxford brain atlases to determine the location of GBMs. RESULTS Overall, our study found smaller tumors in the left hemisphere. This may be expected as left-hemispheric GBM symptoms could present earlier, leading to earlier diagnosis and treatment. However, when the cohort was split by sex, we found this observation significant for females only in the parietal lobe (p < 0.0001). Further, female GBMs demonstrated smaller necrotic volume in the left hemisphere (p = 0.030). Sex-specific differences in incidence were noted in the temporal and occipital lobes (2M:1F). Comparing tumor growth kinetics in different brain locations and hemispheres, females had significantly lower tumor proliferation rates in the left hemisphere (p = 0.009) and lower tumor proliferation rates in the left frontal lobe (p = 0.031). Controlling for treatment, patients with frontal lobe tumors had significantly longer OS compared to those with GBMs in the temporal lobe (p = 0.046, 312 days). Differences in growth velocities were noted between frontal and parietal lobe with frontal GBMs having lower velocities in comparison to parietal lobe GBMs. CONCLUSION Together, our results demonstrate that tumor growth and proliferation rates may vary based on location and sex. Additional research is needed to further explore the clinical significance of tumor location.

Abstract Sex differences in brain tumor related epilepsy (BTE) as an initial symptom of glioma are rarely investigated. Sex differences are being explored with increased interest more recently. Epidemiological studies have generally indicated overall incidence of epilepsy to be slightly higher in males than females, but a higher incidence of idiopathic generalized epilepsy has been seen in females in some studies. Clinicians regularly face the challenge of anti-epileptic drug prescription determination in patients with newly identified brain mass and would greatly benefit from knowing clinical characteristics in BTE. We analyzed clinical characteristics in patients with pathologically-determined glioma with known seizure status at clinical presentation using non-parametric and Fisher’s Exact tests with the entire cohort and male and female separately. From our clinical data repository, we identified 223 patients (females, n=86 (39%); males, n=137 (61%)) divided by those that presented with seizure (SP, n=96, 43%; F, n= 28; M, n= 68) and those that presented without seizure (nSP, n=127, 57%, F n=58, M n=69). In sex-specific analysis males had a significantly (x2=6.28, p=0.01) higher proportion of SP (49.6%) than females (32.6%). There was a dependence between seizure presentation status and tumor grade in males (x2=5.94, p=0.015). Tumor kinetics in the female cohort showed significantly (p< 0.0001) more diffuse tumors in SP compared to nSP. Right-sided tumors were predominantly nSP, while left-sided tumors were split more evenly between SP and nSP. The predominance of nSP among right-sided tumors seems to be driven by females (74%), male right-sided tumors were split more evenly between SP and nSP. Seizure status is dependent on tumor location for the combined cohort, however not for males (x2=5.23, p=0.16) or females (x2=4.34, p=0.23). Clearly there are sex differences and similarities in clinical characteristics at glioma presentation. Further characterization could lead to improved adherence of national standards for AED prescription.

Abstract Glioblastoma (GBM) is the most aggressive primary brain tumor with an average survival of 15 months with standard of care treatment. GBM patients typically present with necrosis surrounded by enhancement on T1-weighted post gadolinium magnetic resonance imaging (T1gd MRI), however some patients present with a significant cystic component. Cysts are caused by different underlying biological mechanisms to necrosis and are important to identify for future clinical investigations. These cystic components can be manually identified through MRI but this process can be time consuming for large patient cohorts. Over the last two decades, our lab has collected serial MRI data of brain tumor patients. With over 70,000 images now in the database and that number increasing daily, it is clear that we have a unique resource for clinical investigation and a need to automate this process. To this end, the aim of this work was to develop and assess the performance of a convolution neural network (CNN) model for automatic detection of cystic GBMs. In this retrospective IRB-approved work, we collected pretreatment MRIs of a patient cohort consisting of 85 patients with a significant cystic component at presentation along with 400 non-cystic GBM, both identified manually through MRI. Image slices with a view of the cystic component were used as positive samples for training. Data were randomly split into training, validation, and test sets using a 70:15:15 ratio. The proportion of positive to negative cases was comparable between sets. Prior to training, we used image augmentation techniques to compensate for the class imbalance in our data. Our results showed that deep learning networks can automatically detect cystic GBMs on MRIs with high accuracy and thus illustrates the potential use of this technique in clinically relevant settings.

Abstract OBJECTIVES Dogma suggests that for brain tumors, regions of enhancement on T1-weighted gadolinium contrast enhanced (T1Gd) magnetic resonance imaging (MRI) correlate with intravenously delivered drug distribution as enhancement indicates a compromised blood-brain barrier (BBB). However, poor response to intravenous therapies highlights the importance of the diffuse disease beyond enhancing regions. This study investigated whether imaging features can provide an accurate prediction of drug distribution. METHODS Eight brain tumor patients (7 gliomas and 1 metastatic adenocarcinoma) were included in this Phase 0 trial. Presurgery T1-weighted, T1Gd, T2-weighted gadolinium contrast enhanced (T2Gd), and T2-weighted Fluid Attenuated Inversion Recovery (T2FLAIR) MRIs were acquired. All images underwent bias correction using the N4 algorithm, standardization of intensities, and registration. Prior to incision, patients received both an antibiotic cefazolin (6% BBB penetrance) and levetiracetam (80% BBB penetrance), an anti-seizure drug. Subsequently, multiple blood samples and image-guided biopsies were taken and analyzed for drug concentration using liquid chromatography mass spectrometry. Biopsy drug levels are reported as Brain-Plasma Ratio (BPR), the ratio of biopsy concentration relative to plasma concentration. Mean image intensity was extracted from an 8x8 mm window surrounding each biopsy location. Regression analysis was performed to determine which combination of image types were linearly predictive of BPR for both drugs. Correlations were also analyzed according to the biopsy location radiographic appearance. RESULTS Regression analysis revealed that T2Gd intensity was linearly predictive of cefazolin BPR and FLAIR intensity was linearly predictive of levetiracetam BPR (p=0.009 and 0.041, respectively). Grouping samples according the the radiographic appearance revealed that levetiracetam BPR had a similar pattern of values to that of FLAIR intensities and cefazolin BPR had a similar pattern to T2, further supporting the regression analysis results. CONCLUSIONS Local concentrations of drug may be related to T2-weighted signals (T2Gd and T2FLAIR) rather than the gadolinium distribution on T1Gd images.