Imaging meningioma biology: Machine learning predicts integrated risk score in WHO grade 2/3 meningioma.

Background: Meningiomas are the most common primary brain tumors. While most are benign (WHO grade 1) and have a favorable prognosis, up to one-fourth are classified as higher-grade, falling into WHO grade 2 or 3 categories. Recently, an integrated risk score (IRS) pertaining to tumor biology was developed and its prognostic relevance was validated in a large, multicenter study. We hypothesized imaging data to be reflective of the IRS. Thus, we assessed the potential of a machine learning classifier for its noninvasive prediction using preoperative magnetic resonance imaging (MRI). Methods: In total, 160 WHO grade 2 and 3 meningioma patients from 2 university centers were included in this study. All patients underwent surgery with histopathological workup including methylation analysis. Preoperative MRI scans were automatically segmented, and radiomic parameters were extracted. Using a random forest classifier, 3 machine learning classifiers (1 multiclass classifier for IRS and 2 binary classifiers for low-risk and high-risk prediction, respectively) were developed in a training set (120 patients) and independently tested in a hold-out test set (40 patients). Results: Multiclass IRS classification had a test set area under the curve (AUC) of 0.7, mostly driven by the difficulties in clearly separating medium-risk from high-risk patients. Consequently, a classifier predicting low-risk IRS versus medium-/high-risk showed a very high test accuracy of 90% (AUC 0.88). In particular, "sphericity" was associated with low-risk IRS classification. Conclusion: The IRS, in particular molecular low-risk, can be predicted from imaging data with high accuracy, making this important prognostic classification accessible by imaging.

Resolving spatial response heterogeneity in glioblastoma.

PURPOSE: Spatial intratumoral heterogeneity poses a significant challenge for accurate response assessment in glioblastoma. Multimodal imaging coupled with advanced image analysis has the potential to unravel this response heterogeneity. METHODS: Based on automated tumor segmentation and longitudinal registration with follow-up imaging, we categorized contrast-enhancing voxels of 61 patients with suspected recurrence of glioblastoma into either true tumor progression (TP) or pseudoprogression (PsP). To allow the unbiased analysis of semantically related image regions, adjacent voxels with similar values of cerebral blood volume (CBV), FET-PET, and contrast-enhanced T1w were automatically grouped into supervoxels. We then extracted first-order statistics as well as texture features from each supervoxel. With these features, a Random Forest classifier was trained and validated employing a 10-fold cross-validation scheme. For model evaluation, the area under the receiver operating curve, as well as classification performance metrics were calculated. RESULTS: Our image analysis pipeline enabled reliable spatial assessment of tumor response. The predictive model reached an accuracy of 80.0% and a macro-weighted AUC of 0.875, which takes class imbalance into account, in the hold-out samples from cross-validation on supervoxel level. Analysis of feature importances confirmed the significant role of FET-PET-derived features. Accordingly, TP- and PsP-labeled supervoxels differed significantly in their 10th and 90th percentile, as well as the median of tumor-to-background normalized FET-PET. However, CBV- and T1c-related features also relevantly contributed to the model's performance. CONCLUSION: Disentangling the intratumoral heterogeneity in glioblastoma holds immense promise for advancing precise local response evaluation and thereby also informing more personalized and localized treatment strategies in the future.

LST-AI: A deep learning ensemble for accurate MS lesion segmentation.

Automated segmentation of brain white matter lesions is crucial for both clinical assessment and scientific research in multiple sclerosis (MS). Over a decade ago, we introduced an engineered lesion segmentation tool, LST. While recent lesion segmentation approaches have leveraged artificial intelligence (AI), they often remain proprietary and difficult to adopt. As an open-source tool, we present LST-AI, an advanced deep learning-based extension of LST that consists of an ensemble of three 3D U-Nets. LST-AI explicitly addresses the imbalance between white matter (WM) lesions and non-lesioned WM. It employs a composite loss function incorporating binary cross-entropy and Tversky loss to improve segmentation of the highly heterogeneous MS lesions. We train the network ensemble on 491 MS pairs of T1-weighted and FLAIR images, collected in-house from a 3T MRI scanner, and expert neuroradiologists manually segmented the utilized lesion maps for training. LST-AI also includes a lesion location annotation tool, labeling lesions as periventricular, infratentorial, and juxtacortical according to the 2017 McDonald criteria, and, additionally, as subcortical. We conduct evaluations on 103 test cases consisting of publicly available data using the Anima segmentation validation tools and compare LST-AI with several publicly available lesion segmentation models. Our empirical analysis shows that LST-AI achieves superior performance compared to existing methods. Its Dice and F1 scores exceeded 0.62, outperforming LST, SAMSEG (Sequence Adaptive Multimodal SEGmentation), and the popular nnUNet framework, which all scored below 0.56. Notably, LST-AI demonstrated exceptional performance on the MSSEG-1 challenge dataset, an international WM lesion segmentation challenge, with a Dice score of 0.65 and an F1 score of 0.63-surpassing all other competing models at the time of the challenge. With increasing lesion volume, the lesion detection rate rapidly increased with a detection rate of >75% for lesions with a volume between 10 mm3 and 100 mm3. Given its higher segmentation performance, we recommend that research groups currently using LST transition to LST-AI. To facilitate broad adoption, we are releasing LST-AI as an open-source model, available as a command-line tool, dockerized container, or Python script, enabling diverse applications across multiple platforms.

Quantitative Assessment of Tumor Contact with Neurogenic Zones and Its Effects on Survival: Insights beyond Traditional Predictors.

So far, the cellular origin of glioblastoma (GBM) needs to be determined, with prevalent theories suggesting emergence from transformed endogenous stem cells. Adult neurogenesis primarily occurs in two brain regions: the subventricular zone (SVZ) and the subgranular zone (SGZ) of the hippocampal dentate gyrus. Whether the proximity of GBM to these neurogenic niches affects patient outcome remains uncertain. Previous studies often rely on subjective assessments, limiting the reliability of those results. In this study, we assessed the impact of GBM's relationship with the cortex, SVZ and SGZ on clinical variables using fully automated segmentation methods. In 177 glioblastoma patients, we calculated optimal cutpoints of minimal distances to the SVZ and SGZ to distinguish poor from favorable survival. The impact of tumor contact with neurogenic zones on clinical parameters, such as overall survival, multifocality, MGMT promotor methylation, Ki-67 and KPS score was also examined by multivariable regression analysis, chi-square test and Mann-Whitney-U. The analysis confirmed shorter survival in tumors contacting the SVZ with an optimal cutpoint of 14 mm distance to the SVZ, separating poor from more favorable survival. In contrast, tumor contact with the SGZ did not negatively affect survival. We did not find significant correlations with multifocality or MGMT promotor methylation in tumors contacting the SVZ, as previous studies discussed. These findings suggest that the spatial relationship between GBM and neurogenic niches needs to be assessed differently. Objective measurements disprove prior assumptions, warranting further research on this topic.

Human-Level Differentiation of Medulloblastoma from Pilocytic Astrocytoma: A Real-World Multicenter Pilot Study.

Medulloblastoma and pilocytic astrocytoma are the two most common pediatric brain tumors with overlapping imaging features. In this proof-of-concept study, we investigated using a deep learning classifier trained on a multicenter data set to differentiate these tumor types. We developed a patch-based 3D-DenseNet classifier, utilizing automated tumor segmentation. Given the heterogeneity of imaging data (and available sequences), we used all individually available preoperative imaging sequences to make the model robust to varying input. We compared the classifier to diagnostic assessments by five readers with varying experience in pediatric brain tumors. Overall, we included 195 preoperative MRIs from children with medulloblastoma (n = 69) or pilocytic astrocytoma (n = 126) across six university hospitals. In the 64-patient test set, the DenseNet classifier achieved a high AUC of 0.986, correctly predicting 62/64 (97%) diagnoses. It misclassified one case of each tumor type. Human reader accuracy ranged from 100% (expert neuroradiologist) to 80% (resident). The classifier performed significantly better than relatively inexperienced readers (p < 0.05) and was on par with pediatric neuro-oncology experts. Our proof-of-concept study demonstrates a deep learning model based on automated tumor segmentation that can reliably preoperatively differentiate between medulloblastoma and pilocytic astrocytoma, even in heterogeneous data.

Toward image-based personalization of glioblastoma therapy: A clinical and biological validation study of a novel, deep learning-driven tumor growth model.

Background: The diffuse growth pattern of glioblastoma is one of the main challenges for accurate treatment. Computational tumor growth modeling has emerged as a promising tool to guide personalized therapy. Here, we performed clinical and biological validation of a novel growth model, aiming to close the gap between the experimental state and clinical implementation. Methods: One hundred and twenty-four patients from The Cancer Genome Archive (TCGA) and 397 patients from the UCSF Glioma Dataset were assessed for significant correlations between clinical data, genetic pathway activation maps (generated with PARADIGM; TCGA only), and infiltration (Dw) as well as proliferation (ρ) parameters stemming from a Fisher-Kolmogorov growth model. To further evaluate clinical potential, we performed the same growth modeling on preoperative magnetic resonance imaging data from 30 patients of our institution and compared model-derived tumor volume and recurrence coverage with standard radiotherapy plans. Results: The parameter ratio Dw/ρ (P < .05 in TCGA) as well as the simulated tumor volume (P < .05 in TCGA/UCSF) were significantly inversely correlated with overall survival. Interestingly, we found a significant correlation between 11 proliferation pathways and the estimated proliferation parameter. Depending on the cutoff value for tumor cell density, we observed a significant improvement in recurrence coverage without significantly increased radiation volume utilizing model-derived target volumes instead of standard radiation plans. Conclusions: Identifying a significant correlation between computed growth parameters and clinical and biological data, we highlight the potential of tumor growth modeling for individualized therapy of glioblastoma. This might improve the accuracy of radiation planning in the near future.

A Learnable Prior Improves Inverse Tumor Growth Modeling.

Biophysical modeling, particularly involving partial differential equations (PDEs), offers significant potential for tailoring disease treatment protocols to individual patients. However, the inverse problem-solving aspect of these models presents a substantial challenge, either due to the high computational requirements of model-based approaches or the limited robustness of deep learning (DL) methods. We propose a novel framework that leverages the unique strengths of both approaches in a synergistic manner. Our method incorporates a DL ensemble for initial parameter estimation, facilitating efficient downstream evolutionary sampling initialized with this DL-based prior. We showcase the effectiveness of integrating a rapid deep-learning algorithm with a high-precision evolution strategy in estimating brain tumor cell concentrations from magnetic resonance images. The DL-Prior plays a pivotal role, significantly constraining the effective sampling-parameter space. This reduction results in a fivefold convergence acceleration and a Dice-score of 95.

Deep Learning to Differentiate Benign and Malignant Vertebral Fractures at Multidetector CT.

Background Differentiating between benign and malignant vertebral fractures poses diagnostic challenges. Purpose To investigate the reliability of CT-based deep learning models to differentiate between benign and malignant vertebral fractures. Materials and Methods CT scans acquired in patients with benign or malignant vertebral fractures from June 2005 to December 2022 at two university hospitals were retrospectively identified based on a composite reference standard that included histopathologic and radiologic information. An internal test set was randomly selected, and an external test set was obtained from an additional hospital. Models used a three-dimensional U-Net encoder-classifier architecture and applied data augmentation during training. Performance was evaluated using the area under the receiver operating characteristic curve (AUC) and compared with that of two residents and one fellowship-trained radiologist using the DeLong test. Results The training set included 381 patients (mean age, 69.9 years ± 11.4 [SD]; 193 male) with 1307 vertebrae (378 benign fractures, 447 malignant fractures, 482 malignant lesions). Internal and external test sets included 86 (mean age, 66.9 years ± 12; 45 male) and 65 (mean age, 68.8 years ± 12.5; 39 female) patients, respectively. The better-performing model of two training approaches achieved AUCs of 0.85 (95% CI: 0.77, 0.92) in the internal and 0.75 (95% CI: 0.64, 0.85) in the external test sets. Including an uncertainty category further improved performance to AUCs of 0.91 (95% CI: 0.83, 0.97) in the internal test set and 0.76 (95% CI: 0.64, 0.88) in the external test set. The AUC values of residents were lower than that of the best-performing model in the internal test set (AUC, 0.69 [95% CI: 0.59, 0.78] and 0.71 [95% CI: 0.61, 0.80]) and external test set (AUC, 0.70 [95% CI: 0.58, 0.80] and 0.71 [95% CI: 0.60, 0.82]), with significant differences only for the internal test set (P < .001). The AUCs of the fellowship-trained radiologist were similar to those of the best-performing model (internal test set, 0.86 [95% CI: 0.78, 0.93; P = .39]; external test set, 0.71 [95% CI: 0.60, 0.82; P = .46]). Conclusion Developed models showed a high discriminatory power to differentiate between benign and malignant vertebral fractures, surpassing or matching the performance of radiology residents and matching that of a fellowship-trained radiologist. © RSNA, 2024 See also the editorial by Booz and D'Angelo in this issue.

LST-AI: a Deep Learning Ensemble for Accurate MS Lesion Segmentation.

Automated segmentation of brain white matter lesions is crucial for both clinical assessment and scientific research in multiple sclerosis (MS). Over a decade ago, we introduced an engineered lesion segmentation tool, LST. While recent lesion segmentation approaches have leveraged artificial intelligence (AI), they often remain proprietary and difficult to adopt. As an open-source tool, we present LST-AI, an advanced deep learning-based extension of LST that consists of an ensemble of three 3D-UNets. LST-AI explicitly addresses the imbalance between white matter (WM) lesions and non-lesioned WM. It employs a composite loss function incorporating binary cross-entropy and Tversky loss to improve segmentation of the highly heterogeneous MS lesions. We train the network ensemble on 491 MS pairs of T1w and FLAIR images, collected in-house from a 3T MRI scanner, and expert neuroradiologists manually segmented the utilized lesion maps for training. LST-AI additionally includes a lesion location annotation tool, labeling lesion location according to the 2017 McDonald criteria (periventricular, infratentorial, juxtacortical, subcortical). We conduct evaluations on 103 test cases consisting of publicly available data using the Anima segmentation validation tools and compare LST-AI with several publicly available lesion segmentation models. Our empirical analysis shows that LST-AI achieves superior performance compared to existing methods. Its Dice and F1 scores exceeded 0.62, outperforming LST, SAMSEG (Sequence Adaptive Multimodal SEGmentation), and the popular nnUNet framework, which all scored below 0.56. Notably, LST-AI demonstrated exceptional performance on the MSSEG-1 challenge dataset, an international WM lesion segmentation challenge, with a Dice score of 0.65 and an F1 score of 0.63-surpassing all other competing models at the time of the challenge. With increasing lesion volume, the lesion detection rate rapidly increased with a detection rate of >75% for lesions with a volume between 10mm3 and 100mm3. Given its higher segmentation performance, we recommend that research groups currently using LST transition to LST-AI. To facilitate broad adoption, we are releasing LST-AI as an open-source model, available as a command-line tool, dockerized container, or Python script, enabling diverse applications across multiple platforms.

Imaging the WHO 2021 Brain Tumor Classification: Fully Automated Analysis of Imaging Features of Newly Diagnosed Gliomas.

BACKGROUND: The fifth version of the World Health Organization (WHO) classification of tumors of the central nervous system (CNS) in 2021 brought substantial changes. Driven by the enhanced implementation of molecular characterization, some diagnoses were adapted while others were newly introduced. How these changes are reflected in imaging features remains scarcely investigated. MATERIALS AND METHODS: We retrospectively analyzed 226 treatment-naive primary brain tumor patients from our institution who received extensive molecular characterization by epigenome-wide methylation microarray and were diagnosed according to the 2021 WHO brain tumor classification. From multimodal preoperative 3T MRI scans, we extracted imaging metrics via a fully automated, AI-based image segmentation and processing pipeline. Subsequently, we examined differences in imaging features between the three main glioma entities (glioblastoma, astrocytoma, and oligodendroglioma) and particularly investigated new entities such as astrocytoma, WHO grade 4. RESULTS: Our results confirm prior studies that found significantly higher median CBV (p = 0.00003, ANOVA) and lower median ADC in contrast-enhancing areas of glioblastomas, compared to astrocytomas and oligodendrogliomas (p = 0.41333, ANOVA). Interestingly, molecularly defined glioblastoma, which usually does not contain contrast-enhancing areas, also shows significantly higher CBV values in the non-enhancing tumor than common glioblastoma and astrocytoma grade 4 (p = 0.01309, ANOVA). CONCLUSIONS: This work provides extensive insights into the imaging features of gliomas in light of the new 2021 WHO CNS tumor classification. Advanced imaging shows promise in visualizing tumor biology and improving the diagnosis of brain tumor patients.

Molecular imaging of gliomas.

Molecular characterization has become a key diagnostic tool for the classification and grading of primary brain tumors. Molecular markers, such as isocitrate dehydrogenase (IDH) mutation status, 1p/19q codeletion, methylation of the O(6)-methylguanine-DNA methyltransferase (MGMT) promoter, or CDKN2A/B homozygous deletion discriminate different tumor entities and grades, and play a crucial role for treatment response and prognosis. In recent years, magnetic resonance imaging (MRI), whose main functions has been to detect a tumor, to provide spatial information for neurosurgical and radiotherapy planning, and to monitor treatment response, has shown potential in assessing molecular features of gliomas from image-based biomarkers. As an outstanding example, numerous studies have proven that the T2/FLAIR mismatch sign can identify IDH-mutant, 1p/19q non-codeleted astrocytomas with a specificity of up to 100%. For other purposes, multiparametric MRI, often coupled with machine learning methods, seems to achieve the highest accuracy in predicting molecular markers. Relevant future applications might be anticipating changes in the molecular composition of gliomas and providing useful information about the cellular and genetic heterogeneity of gliomas, especially in the non-resected tumor parts.

Learn-Morph-Infer: A new way of solving the inverse problem for brain tumor modeling.

Current treatment planning of patients diagnosed with a brain tumor, such as glioma, could significantly benefit by accessing the spatial distribution of tumor cell concentration. Existing diagnostic modalities, e.g. magnetic resonance imaging (MRI), contrast sufficiently well areas of high cell density. In gliomas, however, they do not portray areas of low cell concentration, which can often serve as a source for the secondary appearance of the tumor after treatment. To estimate tumor cell densities beyond the visible boundaries of the lesion, numerical simulations of tumor growth could complement imaging information by providing estimates of full spatial distributions of tumor cells. Over recent years a corpus of literature on medical image-based tumor modeling was published. It includes different mathematical formalisms describing the forward tumor growth model. Alongside, various parametric inference schemes were developed to perform an efficient tumor model personalization, i.e. solving the inverse problem. However, the unifying drawback of all existing approaches is the time complexity of the model personalization which prohibits a potential integration of the modeling into clinical settings. In this work, we introduce a deep learning based methodology for inferring the patient-specific spatial distribution of brain tumors from T1Gd and FLAIR MRI medical scans. Coined as Learn-Morph-Infer, the method achieves real-time performance in the order of minutes on widely available hardware and the compute time is stable across tumor models of different complexity, such as reaction-diffusion and reaction-advection-diffusion models. We believe the proposed inverse solution approach not only bridges the way for clinical translation of brain tumor personalization but can also be adopted to other scientific and engineering domains.

Geometry-Aware Neural Solver for Fast Bayesian Calibration of Brain Tumor Models.

Modeling of brain tumor dynamics has the potential to advance therapeutic planning. Current modeling approaches resort to numerical solvers that simulate the tumor progression according to a given differential equation. Using highly-efficient numerical solvers, a single forward simulation takes up to a few minutes of compute. At the same time, clinical applications of tumor modeling often imply solving an inverse problem, requiring up to tens of thousands of forward model evaluations when used for a Bayesian model personalization via sampling. This results in a total inference time prohibitively expensive for clinical translation. While recent data-driven approaches become capable of emulating physics simulation, they tend to fail in generalizing over the variability of the boundary conditions imposed by the patient-specific anatomy. In this paper, we propose a learnable surrogate for simulating tumor growth which maps the biophysical model parameters directly to simulation outputs, i.e. the local tumor cell densities, whilst respecting patient geometry. We test the neural solver in a Bayesian model personalization task for a cohort of glioma patients. Bayesian inference using the proposed surrogate yields estimates analogous to those obtained by solving the forward model with a regular numerical solver. The near real-time computation cost renders the proposed method suitable for clinical settings. The code is available at https://github.com/IvanEz/tumor-surrogate.

Sequential and Hybrid PET/MRI Acquisition in Follow-Up Examination of Glioblastoma Show Similar Diagnostic Performance.

Both positron emission tomography (PET) and magnetic resonance imaging (MRI), including dynamic susceptibility contrast perfusion (DSC-PWI), are crucial for treatment monitoring of patients with high-grade gliomas. In clinical practice, they are usually conducted at separate time points. Whether this affects their diagnostic performance is presently unclear. To this end, we retrospectively reviewed 38 patients with pathologically confirmed glioblastoma (IDH wild-type) and suspected tumor recurrence after radiotherapy. Only patients who received both a PET−MRI (where DSC perfusion was acquired simultaneously with a FET-PET) and a separate MRI exam (including DSC perfusion) were included. Tumors were automatically segmented into contrast-enhancing tumor (CET), necrosis, and edema. To compare the simultaneous as well as the sequential DSC perfusion to the FET-PET, we calculated Dice overlap, global mutual information as well as voxel-wise Spearman correlation of hotspot areas. For the joint assessment of PET and MRI, we computed logistic regression models for the differentiation between true progression (PD) and treatment-related changes (TRC) using simultaneously or sequentially acquired images as input data. We observed no significant differences between Dice overlap (p = 0.17; paired t-test), mutual information (p = 0.18; paired t-test) and Spearman correlation (p = 0.90; paired t-test) when comparing simultaneous PET−MRI and sequential PET/MRI acquisition. This also held true for the subgroup of patients with >14 days between exams. Importantly, for the diagnostic performance, ROC analysis showed similar AUCs for differentiation of PD and TRC (AUC simultaneous PET: 0.77; AUC sequential PET: 0.78; p = 0.83, DeLong's test). We found no relevant differences between simultaneous and sequential acquisition of FET-PET and DSC perfusion, also regarding their diagnostic performance. Given the increasing attention to multi-parametric assessment of glioma treatment response, our results reassuringly suggest that sequential acquisition is clinically and scientifically acceptable.

Predicting Glioblastoma Recurrence from Preoperative MR Scans Using Fractional-Anisotropy Maps with Free-Water Suppression.

Diffusion tensor imaging (DTI), and fractional-anisotropy (FA) maps in particular, have shown promise in predicting areas of tumor recurrence in glioblastoma. However, analysis of peritumoral edema, where most recurrences occur, is impeded by free-water contamination. In this study, we evaluated the benefits of a novel, deep-learning-based approach for the free-water correction (FWC) of DTI data for prediction of later recurrence. We investigated 35 glioblastoma cases from our prospective glioma cohort. A preoperative MR image and the first MR scan showing tumor recurrence were semiautomatically segmented into areas of contrast-enhancing tumor, edema, or recurrence of the tumor. The 10th, 50th and 90th percentiles and mean of FA and mean-diffusivity (MD) values (both for the original and FWC-DTI data) were collected for areas with and without recurrence in the peritumoral edema. We found significant differences in the FWC-FA maps between areas of recurrence-free edema and areas with later tumor recurrence, where differences in noncorrected FA maps were less pronounced. Consequently, a generalized mixed-effect model had a significantly higher area under the curve when using FWC-FA maps (AUC = 0.9) compared to noncorrected maps (AUC = 0.77, p < 0.001). This may reflect tumor infiltration that is not visible in conventional imaging, and may therefore reveal important information for personalized treatment decisions.