Citations

Abstract Glioblastoma (GBM), isocitrate dehydrogenase wild-type (IDHwt) and central nervous system diffuse large B-cell lymphoma (CNS-DLBCL) are aggressive brain tumors with overlapping MRI features, yet distinct treatment approaches. Noninvasive tools are needed to aid in differential diagnosis. Deep learning on T1 postcontrast and T2-weighted MRI sequences were used to differentiate GBM and CNS-DLBCL. A three-stage temporal study design was utilized. Model development was performed on 146 patients with CNS-DLBCL and 146 age-matched, sex-matched, and MRI year–matched patients with GBM diagnosed at Mayo Clinic between 1998 and 2019. Models were tested on independent temporal test cohorts. Initial testing included 240 independent GBM diagnosed at Mayo Clinic between 1998 and 2019. The prospective test cohort included 37 patients with CNS-DLBCL and 256 patients with GBM diagnosed at Mayo Clinic after January 1, 2020, and 36 patients with CNS-DLBCL diagnosed at an external institution. Of the patients diagnosed at Mayo Clinic, 47% had MRIs generated from non-Mayo institutions. Two different model approaches were compared: (i) ensemble approach using area under the receiver operating characteristic curve (AUC) and cross-validation for model selection and (ii) loss approach minimizing cross-entropy loss and cross-validation to evaluate prediction performance. The AUCs on the prospective test cohort were 0.84 [95% confidence interval (CI), 0.78–0.90] and 0.83 (95% CI, 0.77–0.88) for the ensemble and loss approaches, respectively. Stability of ensemble prediction improved with the increasing number of models. Stratified AUC analysis demonstrated consistent performance across sex and age. We utilized a robust temporal study design and applied 2 different analytic approaches to develop a classification model. The findings confirm the feasibility of using MRI-based deep learning models to differentiate GBM from CNS-DLBCL. Significance: GBM, IDHwt and CNS-DLBCL are aggressive brain tumors with overlapping MRI features, yet distinct treatment approaches. Noninvasive tools are needed to aid in differential diagnosis. We developed MRI-based deep learning models to differentiate GBM, IDHwt from CNS-DLBCL using a rigorous three-stage temporal design that included prospective validation. The model AUC on a prospective cohort was 0.84.

Abstract Glioblastoma, IDH-wildtype (GBM), and primary central nervous system lymphoma (PCNSL) are two aggressive brain tumors with distinct treatment strategies but overlapping imaging features on MRI. Accurate preoperative differentiation remains a clinical challenge, due to their overlapping imaging features, even when advanced MRI techniques are employed. This study explored deep learning approaches leveraging multi-sequence MRI to distinguish GBM from PCNSL. Models were developed utilizing 146 PCNSL cases, which were matched to146 GBM by age, sex and year of diagnosis. A 3D-DenseNet121 convolutional neural network was trained using two analytical strategies: (i) loss-based model selection and (ii) ensemble prediction using five-fold cross-validation. Model performance was evaluated on an independent validation set of 256 GBM and 37 PCNSL from the same institution as well as 36 PCNSL cases from an external institution. All MRIs were preprocessed using the FeTS pipeline. Using the independent validation cohort, area under the receiver operating characteristic curve (AUC) and 95% confidence intervals (CI) were computed overall and stratified by demographic characteristics. Using both T1 post-contrast (T1Gd) and T2-weighted images, the loss-based model achieved AUC=0.83 (95% CI: 0.77-0.88) and the ensemble model yielded an AUC=0.82 (95% CI: 0.76-0.89). Model performance was similar across males and females; however, model performance was better for subjects older than 50 years of age versus younger than 50. When using only T1Gd images, the loss-based model dropped to AUC=0.55 (95% CI: 0.48-0.63), whereas the ensemble model had AUC=0.79 (95% CI: 0.72-0.85). Models trained with only T2 images achieved AUCs of 0.67 (95% CI: 0.60-0.73) (loss-based) and 0.75 (95% CI: 0.69-0.82) (ensemble). These results highlight the added value of the multi-sequence input. Our results confirm the feasibility of using automated deep learning models to differentiate GBM from PCNSL. Notably, the model generalized well to external MRI data, emphasizing its potential utility as a diagnostic aid in real-world clinical workflows.

Abstract The 2021 WHO classification redefined glioblastoma IDH-wildtype (GBM), integrating histopathological and tumor molecular features. Diffuse astrocytic glioma lacking histologic features of necrosis and microvascular proliferation are classified as (molecular) GBM if the tumor is IDH- and H3-wildtype and has one or more additional tumor molecular alterations: TERT promoter mutation, EGFR amplification, or +7/-10. Due to the absence of histopathological features of classic GBM and presentation without enhancement on MRI in approximately 50% of cases, molecular GBM frequently resemble low-grade IDH-mutant gliomas. We hypothesized that MRI-based deep learning models could distinguish IDH-mutant glioma from molecular GBM. We retrospectively identified 97 individuals with molecular GBM, who were matched to 97 individuals with IDH-mutant glioma by age, sex, year of diagnosis, and enhancement on MRI. Of the 97 IDH-mutant gliomas, 51% were 1p/19q codeleted, 42% grade 2, 47% grade 3, and 10% grade 4. Preoperative postcontrast T1-weighted (T1C) and T2-weighted (T2) MRI sequences were used to train a 3D-DenseNet121 model. The model was validated on an independent prospective set of 150 IDH-mutant gliomas and 71 molecular GBM. The MRI-based deep learning model achieved an area under the receiver operator curve (AUC) of 0.75 (95% CI: 0.68-0.82) in the prospective validation set. Sex-stratified AUCs were 0.72 (95% CI: 0.59-0.85) for females and 0.76 (95% CI: 0.68-0.85) for males. Age stratified AUCs were 0.71 (95% CI: 0.60-0.83) for patients under 55 years of age and 0.64 (95% CI: 0.50-0.79) for patients older than 55. These results suggest that discriminative MRI features of IDH-mutant glioma and molecular GBM exist and can be identified from deep learning models. However, radiographic ambiguity exists and underscores the need for further optimization of the deep learning architecture or by integrating additional non-invasive modalities. For example, future work entails integrating germline variants into the model, including rs55705857, which is a causal variant for IDH-mutant glioma. With further validation and integration of multi-modal data, a noninvasive tool may support more personalized decision-making.

Abstract Background The CCDC26 germline variant rs55705857 is causal for development of IDH mutant (IDHmut) adult glioma. However, ~60% of IDHmut patients do not carry the rs55705857 risk allele. We aimed to identify variants associated with developing IDHmut glioma among patients that do not have the rs55705857 risk allele and to further understand development of IDHwt glioma. Methods We used three datasets that included 1216 IDHmut and 1442 IDHwt glioma patients and a case–case design to perform genome-wide association (GWAS) analyses comparing IDHmut versus IDHwt glioma. Analyses were performed overall and stratified by rs55705857 genotype and sex. Multivariable logistic regression and regression trees were used to develop models to predict IDH status using germline variants, age, and contrast enhancement on MRI. Results Three regions were identified comparing IDHmut versus IDHwt: rs55705857 (meta P-value [P] = 1.35 × 10−43), PHLDB1 (rs7125115, P = 3.46 × 10−17), and D2HGDH (rs71430382, P = 2.43 × 10−12). When analyzing only patients that do not have the rs55705857 risk allele, PHLDB1 (rs7125115, P = 1.73 × 10−13) and D2HGDH (rs71430382, P = 8.86 × 10−10) were identified. Among patients who have the rs55705857 risk allele, four variants between ROBO1 and ROBO2 (rs4680975, P = 4.65 × 10−7) increased the likelihood of having an IDHwt tumor. Tumor expression of ROBO1 was associated with rs4680975 genotype in IDHwt patients that have the rs55705857 risk allele (P = 0.034). Multivariable logistic analysis demonstrated that rs55705857, rs71430382 (D2HGDH), and age predicted IDH mutation status. Conclusions To understand the development of adult glioma, we demonstrate that D2HGDH and PHLDB1 should be prioritized for functional studies in IDHmut tumors. The ROBO1 region warrants further investigation in IDHwt tumors.

Pediatric central nervous system tumors are the leading cause of cancer-related deaths in children. The five-year survival rate for high-grade glioma in children is less than 20%. The development of new treatments is dependent upon multi-institutional collaborative clinical trials requiring reproducible and accurate centralized response assessment. We present the results of the BraTS-PEDs 2023 challenge, the first Brain Tumor Segmentation (BraTS) challenge focused on pediatric brain tumors. This challenge utilized data acquired from multiple international consortia dedicated to pediatric neuro-oncology and clinical trials. BraTS-PEDs 2023 aimed to evaluate volumetric segmentation algorithms for pediatric brain gliomas from magnetic resonance imaging using standardized quantitative performance evaluation metrics employed across the BraTS 2023 challenges. The top-performing AI approaches for pediatric tumor analysis included ensembles of nnU-Net and Swin UNETR, Auto3DSeg, or nnU-Net with a self-supervised framework. The BraTS-PEDs 2023 challenge fostered collaboration between clinicians (neuro-oncologists, neuroradiologists) and AI/imaging scientists, promoting faster data sharing and the development of automated volumetric analysis techniques. These advancements could significantly benefit clinical trials and improve the care of children with brain tumors.

The translation of AI-generated brain metastases (BM) segmentation into clinical practice relies heavily on diverse, high-quality annotated medical imaging datasets. The BraTS-METS 2023 challenge has gained momentum for testing and benchmarking algorithms using rigorously annotated internationally compiled real-world datasets. This study presents the results of the segmentation challenge and characterizes the challenging cases that impacted the performance of the winning algorithms. Untreated brain metastases on standard anatomic MRI sequences (T1, T2, FLAIR, T1PG) from eight contributed international datasets were annotated in stepwise method: published UNET algorithms, student, neuroradiologist, final approver neuroradiologist. Segmentations were ranked based on lesion-wise Dice and Hausdorff distance (HD95) scores. False positives (FP) and false negatives (FN) were rigorously penalized, receiving a score of 0 for Dice and a fixed penalty of 374 for HD95. The mean scores for the teams were calculated. Eight datasets comprising 1303 studies were annotated, with 402 studies (3076 lesions) released on Synapse as publicly available datasets to challenge competitors. Additionally, 31 studies (139 lesions) were held out for validation, and 59 studies (218 lesions) were used for testing. Segmentation accuracy was measured as rank across subjects, with the winning team achieving a LesionWise mean score of 7.9. The Dice score for the winning team was 0.65 ± 0.25. Common errors among the leading teams included false negatives for small lesions and misregistration of masks in space. The Dice scores and lesion detection rates of all algorithms diminished with decreasing tumor size, particularly for tumors smaller than 100 mm3. In conclusion, algorithms for BM segmentation require further refinement to balance high sensitivity in lesion detection with the minimization of false positives and negatives. The BraTS-METS 2023 challenge successfully curated well- annotated, diverse datasets and identified common errors, facilitating the translation of BM segmentation across varied environments and providing the tools for future development of personalized volumetric reports to patients undergoing BM treatment.

We describe the design and results from the BraTS 2023 Intracranial Meningioma Segmentation Challenge. The BraTS Meningioma Challenge differed from prior BraTS Glioma challenges in that it focused on meningiomas, which are typically benign extra-axial tumors with diverse radiologic and anatomical presentation and a propensity for multiplicity. Nine participating teams each developed deep-learning automated segmentation models using image data from the largest multi-institutional systematically expert annotated multilabel multi-sequence meningioma MRI dataset to date, which included 1000 training set cases, 141 validation set cases, and 283 hidden test set cases. Each case included T2, FLAIR, T1, and T1Gd brain MRI sequences with associated tumor compartment labels delineating enhancing tumor, non-enhancing tumor, and surrounding non-enhancing FLAIR hyperintensity. Participant automated segmentation models were evaluated and ranked based on a scoring system evaluating lesion-wise metrics including dice similarity coefficient (DSC) and 95% Hausdorff Distance. The top ranked team had a lesion-wise median dice similarity coefficient (DSC) of 0.976, 0.976, and 0.964 for enhancing tumor, tumor core, and whole tumor, respectively and a corresponding average DSC of 0.899, 0.904, and 0.871, respectively. These results serve as state-of-the-art benchmarks for future pre-operative meningioma automated segmentation algorithms. Additionally, we found that 1286 of 1424 cases (90.3%) had at least 1 compartment voxel abutting the edge of the skull-stripped image edge, which requires further investigation into optimal pre-processing face anonymization steps.

The Quantitative Imaging Biomarkers Alliance Profile for dynamic susceptibility contrast (DSC) MRI and its longitudinal claims for the precision of a DSC MRI–derived quantitative imaging biomarker to assess the relative cerebral blood volume changes in gliomas are summarized.

Abstract Clinical management of IDH wildtype glioblastoma (GBM) and tumefactive multiple sclerosis (tMS) is drastically different. GBM requires maximal safe resection followed by chemoradiation, while tMS outcome is worsened by surgery and radiotherapy. Noninvasive methods are needed to help with accurate diagnosis of tumor and non-tumor etiologies. To develop an MRI-based classification model, tMS subjects diagnosed prior to January 1, 2020, were matched to tMS by age at diagnosis, sex, index MRI date, and 2D/3D acquisition. Inclusion criteria included one cm minimal lesion size and pre-operative post-contrast T1 and T2 images available for analysis. A 3D-DenseNet121 was used to develop a classification model using prespecified parameters: 650 epochs, batch size 16, learning rate 10-3, cross-entropy loss, and AdamW optimizer. The stopping rule was defined as three sequential differences in epoch cross-entropy loss <0.02. Models were developed using both T1gd and T2, as well as from only T1gd and only T2. Training included 220 subjects (110 GBM, 110 tMS). A 2-stage validation design was used, which included both retrospective and prospective cohorts. Stage 1 consisted of 272 retrospective GBM (diagnosed prior to January 1, 2020). Stage 2 consisted of 69 and 34 prospective (diagnosed after January 1, 2020) GBM and tMS, respectively. External validation on the 272 retrospective GBM demonstrated accuracy of 91%, 84%, and 78% for T1gd+T2, T1gd only, and T2 only, respectively. External validation on the 69 prospective GBM demonstrated an accuracy of 87%, 64%, and 67% for T1gd+T2, T1gd only, and T2 only, respectively. The 34 prospective tMS demonstrated accuracy of 76%, 76%, and 82% for T1gd+T2, T1gd only, and T2 only, respectively. This shows the feasibility of deep learning to aid in differential diagnosis of brain lesions. Future work entails the integration of germline variants into the classification model, including variants associated with the risk of glioma or MS.

The performance of a deep learning model was not different from that of radiologists in the detection of clinically significant prostate cancer at MRI, and gradient-weighted class activation maps localized the tumor.