AI and Cancer Research ── Brain ── 2026-11-16
Brain tumours: treatment, trials and AI
- Abstract
To assess the feasibility of personalized tumor-masked structural connectomics combined with interpretable machine learning (ML) for differentiating low-grade gliomas (LGG) from high-grade gliomas (HGG), and to identify candidate tract-specific connectomic biomarkers of glioma aggressiveness. Thirty-five patients with histopathologically confirmed gliomas (20 LGG, 15 HGG) underwent preoperative diffusion tensor imaging and T1-weighted MRI. Tumor-masked probabilistic tractography was used to reconstruct individualized structural connectomes based on 50 white matter (WM) tract labels. We evaluated 296 graph-theoretical features using sequential feature selection and Random Forest classification within a strictly isolated nested cross-validation framework. We used SHAP analysis for model interpretation. Sensitivity analyses compared clinical-radiological, connectomic, and combined feature sets. The leak-free pure connectomics model achieved a mean ROC-AUC of 0.814 ± 0.327 (95% CI, 0.500-1.000), with a macro-F1 score of 0.786 ± 0.265. Restricting the analysis to non-zero connectomic features yielded a mean AUC of 0.833 ± 0.167. The clinical-radiological model achieved a mean AUC of 0.817 ± 0.149, while the combined clinical-radiological and connectomic model showed the highest point estimate (AUC, 0.933 ± 0.109; 95% CI, 0.798-1.068). SHAP analysis highlighted the sagittal stratum, medial lemniscus, and superior fronto-occipital fasciculus among the most informative WM pathways. Tumor-masked structural connectomics combined with interpretable ML demonstrated preliminary feasibility for identifying candidate imaging biomarkers associated with glioma aggressiveness in neuroradiological practice. These findings are exploratory and represent internal estimates from a single cohort; external validation in independent, multi-center cohorts is required before any potential clinical application.
Journal IF-equivalent: 2.8 (OpenAlex 2-year mean citedness, value as of 2026-11-12, retrieved 2026-11-13; not the official Clarivate IF)Reference: Franco P, Montalba C, Espinoza I, Cornejo MD, Torres F, Bennett C, et al. Tumor-masked structural connectomics for assessing tract-specific white matter disruption associated with glioma aggressiveness: an interpretable machine learning pilot study. Neuroradiology. 2026 Nov 14 [Epub ahead of print]. doi:10.1007/s00234-026-04232-0. PMID: 43033259.Checked: Abstract only - Abstract
OBJECTIVE: Preoperative knowledge of intracranial meningioma grade could significantly enhance surgical planning, but reliable noninvasive prediction remains an unmet clinical need. The aim of this study was to develop and externally validate a machine learning (ML) model incorporating traditional radiomic features, along with novel fractal and topological features, to predict WHO meningioma grade from preoperative MRI using a large multicenter cohort. METHODS: This retrospective diagnostic study analyzed a cohort of patients in the BraTS Meningioma Challenge. Patient data (n = 909) from 6 academic medical centers were used for model training and internal validation (n = 699) and one-time final external validation (n = 210). Tumors were segmented into three regions and radiomic features extracted across four MRI sequences. Eight different ML algorithms (e.g., based on neural networks, decision trees, clustering, and linear relationships) were tested across a total of 2889 candidate features. The primary outcome was the area under the receiver operating characteristic curve (AUROC) for prediction of binarized WHO grade (1 vs 2 or 3). RESULTS: The final model used a support vector machine algorithm comprising 20 features and achieved an AUROC of 0.824 (95% CI 0.755-0.883) on unseen test data, with favorable calibration. Feature importance analysis revealed that topological complexity added complementary predictive value to conventional radiomics features. The final model is publicly available for anyone to use. CONCLUSIONS: This externally validated ML model accurately predicted meningioma grade using routine preoperative MRI, with the potential to guide surgical decision-making and serve as a basis for future imaging-based brain tumor classifications.
Journal IF-equivalent: 3.9 (OpenAlex 2-year mean citedness, value as of 2026-11-13, retrieved 2026-11-15; not the official Clarivate IF)Reference: Adil SM, Zachem TJ, Warman P, Abdelgadir J, Patel N, Wissel BD, et al. Machine learning prediction of meningioma grade using preoperative MRI: a large multicenter study. J Neurosurg. 2026 Nov 1 [Epub ahead of print]. doi:10.3171/2026.5.jns252856. PMID: 43024876.Checked: Abstract only - Abstract
Background: Despite advances in surgical technique, postoperative tumor recurrence or progression (R/P) remains a major challenge among patients with clinically non-functioning pituitary adenomas (NFPA). Reliable tools for individualized R/P risk stratification are limited.
Objective: To develop and evaluate machine learning (ML) models that integrate clinical, immunohistochemical, and radiomic features based on preoperative MRI towards predicting tumor R/P in patients with NFPA who underwent pituitary surgery.
Methods: We retrospectively analyzed data of patients with NFPA who underwent their first pituitary operation at Massachusetts General Hospital between 2005–2019, had preoperative postcontrast MRI images available, and ≥3 months of follow-up and conducted an exploratory multimodal modelling analysis. Supervised ML models combining clinical and radiomic features were then trained to predict R/P and were internally evaluated using cross-validation. SHAP analysis was used to identify predictors of R/P.
Results: 537 patients met inclusion criteria (median age: 58 yr, range: 18–92 yr; 59% male). Overall, 120 patients (22%) experienced R/P during a median follow-up of 74 months. Median R/P-free survival was 170 months. Six ML models were evaluated, wherein TabPFN-based models outperformed others (AUROC: 0.687). Younger age at diagnosis, larger tumor size, higher Ki 67 index, cavernous sinus invasion, male sex, and positive immunostaining for pituitary hormones were predictors of R/P. First order radiomic features also contributed to the prediction of R/P in several models.
Conclusions: The TabPFN model best predicted tumor R/P in the presence cohort. Prospective studies are needed to confirm and extend the present findings.
Journal IF-equivalent: 0.7 (OpenAlex 2-year mean citedness, value as of 2026-11-15, retrieved 2026-11-16; not the official Clarivate IF)Reference: Habib U, Bridge CP, Mikamoto M, Rapalino O, Kim AE, Miller KK, et al. Machine Learning-Based Prediction of Tumor Recurrence or Progression in Patients with Non-Functioning Pituitary Adenomas. Journal of the Endocrine Society. 2026 Nov 14 [Epub ahead of print]. doi:10.1210/jendso/bvag268.Checked: Abstract only - Abstract
BACKGROUND: Gliomas are the most common primary malignant brain tumors and exhibit highly heterogeneous morphology and infiltrative growth patterns. Accurate delineation of tumor subregions in magnetic resonance imaging (MRI) is therefore critical for diagnosis, treatment planning, and longitudinal monitoring, yet manual segmentation remains time-consuming and subject to inter-observer variability. PURPOSE: Accurate glioma segmentation in magnetic resonance imaging (MRI) is essential for diagnosis, treatment planning, and reproducible tumor delineation. However, for clinical adoption, automated segmentation systems must provide not only high accuracy but also robustness and interpretability. This study aimed to develop an explainable and reliable deep learning framework for glioma segmentation and to evaluate its performance on both internal and external datasets. METHOD: We used the BraTS 2021 training dataset, which contains 1251 multi-institutional pre-operative glioma MRI cases. After comparing multiple sequence combinations, T1ce, T2, and Fluid Attenuated Inversion Recovery (FLAIR) were selected as the final three-channel input. Preprocessing included tumor-oriented slice selection, cropping, Min-Max normalization, and augmentation. We developed AXUNet, a UNet-based architecture that integrates memory-efficient Xception blocks with a self-attention module combining kernel-based pixel attention and dot-product channel attention. Internal ablation analysis and Gradient-weighted Class Activation Mapping (Grad-CAM)-based visualization were also performed. RESULTS: On the BraTS 2021 test partition, AXUNet achieved the highest mean Dice score (93.73), outperforming UNet (93.26), Inception-UNet (90.88), Attention Gate UNet (AG-UNet) (90.38), Attention ResUNet (AResUNet) (92.80), and Xception-UNet (93.24). Region-wise Dice scores were 92.59 for Whole Tumor (WT), 86.81 for Tumor Core (TC), and 84.89 for Enhancing tumor (ET). Although AResUNet achieved a slightly higher ET score (84.92), AXUNet demonstrated the strongest performance in mean Dice, WT, and TC. To further assess generalizability, AXUNet was independently evaluated on 219 previously unseen BraTS 2023 cases through the Synapse platform, achieving Dice scores of 86.29% for WT, 78.53% for TC, and 69.00% for ET. CONCLUSION: AXUNet demonstrates that combining Xception-based encoding with self-attention can improve glioma segmentation while also supporting explainability and external robustness, positioning it as a promising framework for MRI-based glioma segmentation. Its clinical relevance, and the practical value of its interpretability outputs, remain to be established through further, dedicated validation beyond Dice-based benchmarking.
Journal IF-equivalent: 2.4 (OpenAlex 2-year mean citedness, value as of 2026-11-04, retrieved 2026-11-05; not the official Clarivate IF)Reference: Moodi F, Khodadadi Shoushtari F, Valizadeh G, Mazinani D, Mobarak Salari H, Sina S, et al. Attention Xception U‐Net (AXUNet): An explainable combination of CNN and self‐attention for reliable glioma segmentation in MRI. J Applied Clin Med Phys. 2026 Nov 13;27(11):e70902. doi:10.1002/acm2.70902.Checked: Abstract only