Ex vivo platforms can rapidly and cost-effectively screen patient-derived tumor cells or tissue. Artificial intelligence (AI) algorithms can search and identify patterns in large datasets and provide predictions. This review focuses on integrating microphysiological platforms with AI to inform physician and patient decision-making. Combining efficacy, safety, and pharmacology results from drug screens with the output of extensive AI searches can yield insights to guide physician and patient decision-making and potentially improve a patient’s prognosis. We detail ex vivo platforms at different stages of development that represent the diversity of approaches: a microphysiological system and a high-throughput screen that assesses drug cytotoxicity in both bulk and drug-tolerant tumor cells. We review AI approaches that can enhance the utility of microphysiological platforms. Integrating emerging microphysiological platforms with AI is expected to significantly impact physician and patient choice of treatment.
The clinical benefit of bevacizumab (Bev) in glioblastoma (GBM) is typically transient, as tumors develop adaptive resistance characterized by an infiltrative shift from angiogenesis dependent growth. This study investigated Annexin A2 (ANXA2), a regulator of both angiogenesis and invasion, as a potential biomarker of Bev resistance. We analyzed 66 tissue-specimens from 33 GBM patients using quantitative real-time PCR and immunohistochemistry/immunofluorescence. The cohort included 15 Bev-naïve cases and 18 neoadjuvant Bev (neoBev) cases. 33 refractory specimens-including unique paired samples from the same patients at different treatment phases (initial resection vs. refractory stage) were evaluated to correlate ANXA2 expression levels with progression-free survival (PFS), overall survival (OS), and MRI recurrence patterns. Multivariate analysis identified high ANXA2 expression as a significant independent poor prognostic factor for both PFS and OS. In the neoBev group, patients with low ANXA2 mRNA expression levels demonstrated significantly superior survival outcome compared to those with high expression. While ANXA2 mRNA levels tended to increase at the refractory stage, a significant negative correlation was observed between the number of Bev cycles and ANXA2 expression levels. Histopathological analyses revealed intense ANXA2 expression in the tumor vasculature and stroma at the GBM leading edge, co-localization with the hypoxia marker hypoxia-inducible factor-1α. ANXA2 may serve as a predictive biomarker of the adaptive transition from an angiogenic to infiltrative phenotype with Bev therapy. These findings suggest that ANXA2 serves as a predictive biomarker for Bev clinical benefit and may represent a candidate therapeutic target requiring further validation in GBM.
Brain metastases from gastrointestinal (GI) malignancies are heterogeneous, while stereotactic radiosurgery dose selection remains guided by lesion size, location, prior treatment, and physician judgment rather than individualized estimates of local failure and toxicity. We developed THINKERS-GI, a three-expert mixture-of-experts framework for outcome prediction and dose-policy evaluation after Gamma Knife radiosurgery (GKRS). We performed a retrospective single-center study of 300 patients with 1,233 GI brain metastases. Pretreatment variables were used to train a three-expert mixture-of-experts neural network with discrete-time competing-risk survival modeling. Margin dose was incorporated as a queryable input for candidate-dose evaluation. Internal validation used grouped 5-fold cross-validation and a closed grouped holdout set. Performance was assessed using time-dependent AUC, C-index, Brier score, calibration, and MAE. In grouped cross-validation, time-dependent AUC ranged from 0.808 to 0.894, with a C-index of 0.889 ± 0.034. The calibrated 12-month Brier score was 0.019 ± 0.013, and MAE for time to local failure was 1.8 ± 0.7 months. In the grouped holdout set, 12-month AUC was 0.806 (95
A previous double-blind, randomized, placebo-controlled, phase II clinical trial reported beneficial effects of a short-term treatment (10 days) with murine nerve growth factor (mNGF) eye drops on visual function in children with optic pathway gliomas (OPG). The present study aimed to evaluate long-term changes in clinical and neuroradiological parameters in the cohort of OPG patients who had previously participated in the phase II mNGF trial. Fifteen of the 18 patients originally enrolled in the phase II mNGF trial agreed to undergo clinical and neuroradiological monitoring over a 48-month follow-up period. Of these, 9 had originally been randomized to mNGF and 6 to placebo; no additional treatment (mNGF, chemotherapy, or radiotherapy) was administered during the extended follow-up. Every 6 months, patients underwent general clinical and neuro-ophthalmological examination, visual evoked potentials (VEP), and photopic negative response of the electroretinogram (PhNR). Brain MRI was performed every 12 months. Comparison of initial and final follow-up median values revealed no statistically significant changes in visual acuity, VEP amplitude, PhNR amplitude, or visual field radius. No significant differences were observed in any parameter relative to baseline values of the mNGF trial. Brain MRI demonstrated stable disease in all patients throughout the observation period. These findings, obtained in an observational extension of the original randomized cohort, indicate favorable long-term safety and tolerability of a short-term course of topical mNGF in children with OPG, with sustained visual and neuroradiological stability over four years, rather than evidence of persistent treatment efficacy. Further prospective, adequately powered and randomized clinical studies are needed to confirm both the short- and long-term clinical efficacy of NGF treatment in preventing OPG-induced visual loss.
Gliomas are highly invasive primary brain tumors, and additional biomarkers are needed to improve prognostic stratification and clarify the mechanisms underlying tumor invasion. Although α-synuclein (α-syn), encoded by SNCA, is a central protein in neurodegenerative disease, its role in glioma remains poorly defined. SNCA/α-syn expression was evaluated in glioma specimens from Beijing Tiantan Hospital and in the Chinese Glioma Genome Atlas and The Cancer Genome Atlas cohorts. Associations with clinicopathological characteristics, molecular features, and overall survival were analyzed. Gain- and loss-of-function experiments were performed in U87 cells and patient-derived glioma stem-like cells (GSLCs), followed by proliferation, migration, invasion, and pathway-rescue assays. An orthotopic xenograft model was used to assess tumor growth and survival in vivo. SNCA/α-syn expression was lower in high-grade gliomas and was associated with favorable molecular features and prolonged survival. α-Syn overexpression inhibited glioma cell proliferation, migration, and invasion, whereas α-syn knockdown promoted proliferation and invasion in GSLCs. Mechanistically, α-syn increased CDH13 expression, activated JNK/c-Jun signaling, and reduced MMP3 and MMP9 expression. CDH13 knockdown or JNK inhibition attenuated these effects and restored glioma cell migration and invasion. In vivo, α-syn overexpression suppressed orthotopic tumor growth and prolonged survival. High SNCA/α-syn expression is associated with favorable prognosis, whereas α-syn functions as a suppressor of glioma progression, at least partly through the CDH13–JNK/c-Jun axis.
Primary Large-B-Cell Lymphoma of the CNS (PCNS-LBCL) is a rare, aggressive tumor often sensitive to corticosteroid therapy (CST). This multicenter study investigated the spectrum of radiological responses to routine CST and evaluated its impact on patient prognosis. The researchers utilized a prospective cohort of 18 patients for volumetric MRI analysis and a combined prospective-retrospective cohort of 31 patients for 2D tumor analysis. Patients received CST post-biopsy, with follow-up MRIs performed on the seventh day (median) after surgery. The study correlated radiological responses with CST dosage, LDH levels, and overall survival (OS) using Kaplan-Meier and Firth-corrected Cox regression. In the prospective cohort, 83.3
Leptomeningeal disease (LMD) from solid tumors carries poor prognosis and substantial neurologic morbidity. Radiotherapy is used for palliation and disease control, but outcomes across whole-brain radiotherapy (WBRT), photon craniospinal irradiation (CSI), and proton CSI remain incompletely defined. We conducted a PRISMA-based systematic review of PubMed-indexed studies reporting radiotherapy outcomes for adult solid-tumor LMD. Studies were categorized as photon WBRT/non-CSI, photon CSI, or proton CSI. Primary outcomes were median overall survival (OS) and progression-free survival (PFS), including CNS-PFS or intracranial PFS when reported. Grade ≥ 3 hematologic and non-hematologic toxicities were extracted when available. Thirty-two unique studies/reports were included: 21 photon WBRT/non-CSI, 6 photon CSI, and 6 proton CSI studies. The median of reported median OS values were 4.5 months for photon WBRT/non-CSI, 4.1 months for photon CSI, and 10.7 months for proton CSI. Median PFS was 3.9 months for photon WBRT/non-CSI and 4.6 months for proton CSI; no extractable median PFS was available for photon CSI. Among studies with extractable denominators, crude pooled grade ≥ 3 hematologic toxicity was 2.8
The optimal extent of resection (EOR) required to minimize recurrence risk in craniopharyngioma remains unclear. The objective of this study was to quantitatively evaluate postoperative tumor volume and EOR as determinants of progression-free survival (PFS) in surgically treated adamantinomatous craniopharyngiomas. We retrospectively measured pre- and post-surgical tumor volumes in 60 patients using manual MRI-based segmentation. None received planned adjuvant radiotherapy. The median follow-up duration was 5 years. Clinical and volumetric variables were analyzed using Cox proportional hazards regression. Time-dependent receiver operating characteristic analysis at 3 years was also performed. Tumor progression/recurrence occurred in 60
Central nervous system (CNS) tumours are the deadliest cancer for children and currently present limited treatment options. Chimeric antigen receptor (CAR)-T cell therapies have emerged as an innovative approach supported by encouraging clinical results. Current clinical trials using CAR-T cells in the treatment of paediatric CNS cancers differ in a number of variables, including the CAR-T cell route of delivery, presence of lymphodepletion, identified target antigen, and CAR engineering features. Considering early learnings across these areas is an essential step to developing more effective treatment options, especially given the challenges of immunosuppressive tumour microenvironments, various toxicities, CAR-T cell exhaustion, and tumour antigen heterogeneity. In sum, while there is a need for continued innovation, CAR-T cells represent a promising treatment approach for this devastating category of diseases.
Recurrence risk estimates underpin meningioma research, including molecular classification and clinical trial benchmarking, yet are often based on retrospective or historical data. The aim of this study was to assess the variation of recurrence risk estimates across calendar periods, WHO classification editions, geographical settings, and healthcare systems.thetermine We analyzed 4,111 patients with primary WHO-1/-2 meningiomas from 31 centers in 15 countries (1990–2019). Recurrence was defined according to local radiological assessment. The 5- and 10-year recurrence risks were estimated using regression standardization with inverse probability of censoring weights, adjusting for key clinical, surgical, and histopathological variables. Recurrence risk estimates varied across all domains examined. More recent calendar periods were associated with higher predicted recurrence risk, particularly for WHO-2 at 5 years (e.g., ≥ 2013 vs. ≤ 2007: RR 1.60, 95
Differentiating tumor recurrence from radiation necrosis (RN) after stereotactic radiosurgery (SRS) remains a major diagnostic challenge in brain metastasis. We aimed to validate established MRI-based tumor habitat analysis for distinguishing tumor from RN in an independent cohort with histopathological ground truth. This retrospective study included 104 patients (104 lesions) with pathologically confirmed recurrent metastatic tumors (n = 68) or RN (n = 36) who underwent structural and physiologic MRI. Tumor habitats were generated using an established unsupervised clustering model applied to normalized T1-weighted enhanced, T2-weighted, apparent diffusion coefficient, and cerebral blood volume maps. Structural habitats (enhancing tissue, solid low-enhancing, nonviable) and physiologic habitats (hypervascular, hypovascular cellular, nonviable) were quantified as absolute volumes and volume fractions. Logistic regression and receiver operating characteristics analysis evaluated the ability to differentiate tumor and RN. Composite habitat scores integrating structural and physiologic habitats were also developed. Recurrent metastatic tumors showed higher contrast-enhancing volume (P = .006), higher solid low-enhancing habitat volume (P = .029) and fraction (P = .04), higher hypervascular habitat volume (P = .02) and fraction (P = .03), and lower nonviable tissue habitat fractions on structural (P = .003) and physiologic MRI (P = .015), compared with RN. The combined structural and physiologic MRI habitat score showed the highest diagnostic performance (AUC, 0.80; 95
Chordomas are rare, locally aggressive tumors for which effective salvage options are limited once maximal surgical and radiation therapy have been exhausted. Laser interstitial thermal therapy (LITT) is a minimally invasive, MRI-guided ablative technique established for spinal metastases, but published experience with LITT for chordoma remains limited to isolated case reports. We report the largest clinical experience with spinal LITT (sLITT) for chordoma and review the existing literature. We performed a retrospective, IRB-approved single-institution case series of patients with histologically confirmed chordoma treated with sLITT. Clinical, operative, and radiographic data were abstracted, including tumor volumes, ablation parameters, and functional and oncologic outcomes. A systematic PubMed search was performed to identify all prior published reports of LITT for chordoma. Four patients underwent sLITT across five procedures, treating metastatic foci from primary sacral, clival, and cervical chordomas, as well as one primary thoracic lesion. sLITT was used as salvage treatment in four procedures and as primary treatment in one patient to avoid the morbidity of open surgery. Median time from diagnosis to LITT was 19.1 months (range 2.5–168.3), and pre-operative KPS was 90–100 in all patients. A median of four laser fibers and 16 ablations were used per procedure (12–15 W). Median case duration was 352.5 min and length of stay was two days. Postoperative SRS was administered in all but one procedure. All patients experienced volumetric tumor reduction on the earliest post-operative MRI (median 4.2 months, range 2.8–8.8), with a mean reduction of 25.7
Central nervous system hemangioblastomas usually achieve durable local control after gross-total resection (GTR), yet recurrence can occur even after apparently complete removal. We evaluated whether preoperative MRI-defined radiologic subtypes, particularly cyst wall enhancement (CWE), are associated with local recurrence in surgically treated intracranial hemangioblastomas. We retrospectively reviewed 52 surgically treated intracranial hemangioblastomas in 49 patients between 2010 and 2025. Tumors were classified as solid (S), extratumoral cystic (Ce), intratumoral cystic (Ci), or mixed cystic (Cm) type. Primary analysis was restricted to 47 first resections, including 44 tumors treated with gross-total resection (GTR). Local recurrence after GTR was evaluated primarily using Kaplan–Meier analysis and the log-rank test. Because no recurrence occurred among CWE-negative GTR tumors, exploratory odds ratios were estimated using Firth bias-reduced penalized logistic regression. Among the 44 primary tumors treated with GTR, local recurrence occurred in 8 of 14 CWE-positive tumors (57.1
Hypoxia-targeted blood oxygen level–dependent (BOLD) MRI is an emerging non-contrast imaging technique for characterizing tumor biology in isocitrate dehydrogenase-wildtype glioblastoma. Beyond the magnitude of BOLD signal changes, signal heterogeneity may provide valuable information about tumor aggressiveness and microenvironmental complexity. The aim was to assess both the magnitude and heterogeneity of hypoxia-targeted BOLD signal changes across subregions of glioblastomas. In this prospective pilot study, hypoxia-targeted BOLD MRI data from 12 patients with histopathologically confirmed glioblastoma were analyzed. Tumor subregions were defined using an automated segmentation framework (Oncohabitats) based on conventional MRI and dynamic susceptibility contrast (DSC) MR perfusion imaging. Voxel-wise BOLD signal changes were calculated within the following volumes of interest: highly angiogenic tumor (HAT), low angiogenic tumor (LAT), peritumoral edema, and mirrored contralateral control tissue (mCET). The magnitude of BOLD signal changes was assessed using median BOLD signal changes, while heterogeneity was quantified using interquartile ranges (IQR), and kernel density estimation-derived entropy. Friedman tests with Dunn post-hoc tests were used to assess parameter differences between the VOIs. HAT demonstrated the largest BOLD signal changes compared with LAT, edema, and mCET (all FDR-adjusted p ≤ 0.017). Heterogeneity metrics revealed significantly increased IQR and entropy in HAT compared with edema and mCET (all FDR-adjusted p < 0.008), with LAT showing intermediate values. Hypoxia-targeted BOLD MRI reveals both greater BOLD signal changes and elevated intratumoral heterogeneity in highly angiogenic and likely aggressive glioblastoma subregions. These findings suggest that heterogeneity metrics may provide complementary information to BOLD signal magnitude alone.
Postoperative hydrocephalus after central neurocytoma (CN) resection may persist despite tumor removal. This study evaluated anatomical and surgical factors associated with postoperative hydrocephalus-related outcomes. Patients with pathologically confirmed CN who underwent tumor resection between 2011 and 2025 were retrospectively analyzed. The primary outcomes were composite postoperative hydrocephalus, radiographic hydrocephalus at latest follow-up, and hydrocephalus-related revision surgery (HRRS). Associations were evaluated using multivariable logistic and Cox regression with consideration of surgeon-level clustering. Propensity score matching was used to improve comparability between the transcortical (TC) and interhemispheric transcallosal (ITC) groups, and spontaneous remission was additionally assessed using competing-risk analysis. Among 652 screened patients, 454 were included in the descriptive cohort and 406 in the complete-case analytical cohort; propensity score matching yielded 119 pairs. Mean follow-up was 85.50 ± 49.94 months. Immediate postoperative hydrocephalus rates were similar between approaches, whereas composite postoperative hydrocephalus was more frequent after TC in both the descriptive cohort (31.60
Brain metastases are associated with a poor prognosis. While systemic therapy primarily determines overall survival, surgical cytoreduction may improve local disease control. Preserving neurological function and independence is particularly important in patients with limited life expectancy. However, surgery is often avoided for metastases located in eloquent brain regions. This study evaluates treatment selection and functional outcomes in patients with metastases involving the primary motor cortex. We conducted a retrospective single-center analysis of patients treated for central motor eloquent brain metastases. Primary endpoints were motor function and KPS at discharge and at 6-month FU. Neurological change was categorized as improvement, stability, or deterioration in Medical Research Council (MRC) grade. Secondary variables included intracranial tumor burden, extracranial disease, prior systemic therapy, disease duration, which were used for risk stratification. Comparative analyses were performed using Wilcoxon rank-sum test, Fisher’s exact test, multivariable logistic, and Firth regression. A total of 152 patients (79 male, 73 female; mean age 65.5 years) were eligible. Treatment comprised primary surgery in 109 patients, definitive stereotactic radiotherapy in 24, salvage surgery after radiotherapy in 17, and best supportive care in two. In single metastases with higher volume surgery was favored, whereas extracranial disease, prior systemic therapy, and longer disease duration were associated with allocation to radiotherapy. At discharge improvement rate was 13
Meningiomas show biological heterogeneity not captured by WHO grading alone. TGF-β signaling has been implicated in meningioma pathobiology, yet a systematic, grade-stratified, isoform-resolved characterization of TGF-β1, TGF-β2, and TGF-β3 across transcriptional, epigenetic, post-transcriptional, and protein levels had not been performed. We prospectively enrolled 154 patients undergoing resection of WHO CNS5 grade 1 meningothelial (n = 124) or grade 2 atypical (n = 30) meningioma across two neurosurgical centers. TGF-β1–3 mRNA and six bioinformatically prioritized targeting microRNAs were quantified by RT-qPCR; promoter methylation was assessed by methylation-specific PCR; protein abundance was measured by ELISA, Western blotting, and immunohistochemistry. TGF-β2 and TGF-β3 mRNA were markedly upregulated in grade 2 (fold-change 5.68 and 5.23, respectively), corroborated by higher ELISA and IHC signal, though not by Western blot. TGF-β1 mRNA expression was markedly reduced in grade 2 (fold-change 0.039) with reduced protein, despite a predominantly unmethylated promoter (76.7
Targeted therapy for BRAF V600E mutant papillary craniopharyngioma (PCP) has been rapidly accepted, though given the rarity of the tumor, there is limited guidance for appropriate use. Here, we retrospectively evaluated current practice patterns for the implementation of targeted therapeutics and their outcomes in the treatment of papillary craniopharyngioma (PCP). Practice patterns at the institutions in the Registry for Adenomas of the Pituitary and Related Disorders (RAPID) for treatment of BRAF V600E mutated PCPs using targeted therapy were evaluated. Baseline clinical and demographic variables were retrospectively recorded. Imaging response was evaluated. Dosing, treatment course, adverse events (AEs) and other therapeutic strategies employing BRAF/MEK inhibitors were studied. Adjuvant and/or salvage therapies, including radiation and surgery were recorded. Most patients demonstrated at least partial response to BRAF and/or BRAF/MEK inhibition (13/19, 72.2
Transformation of vestibular schwannoma (VS) into malignant peripheral nerve sheath tumor (MPNST) is rare and poorly characterized. We have synthesized available evidence from the literature to define clinical features, transformation patterns, management strategies, and survival outcomes in patients with MPNST after histologically confirmed transformation from VS. A PRISMA-guided systematic review of PubMed, Scopus, and Google Scholar databases identified case reports and case series reporting histologically confirmed VSs transforming to MPNSTs with clinical and outcome data. Demographic, clinical, treatment, and survival variables were extracted and analyzed. Twenty-one studies comprised 24 patients (7 male) with MPNSTs (median age 52 years; 8
Since its introduction in 1963, the Ommaya reservoir has expanded beyond its original role as a ventricular access device to support intrathecal therapeutics, cerebrospinal fluid diagnostics, and emerging applications in precision neuro-oncology. To our knowledge, the evolution of this research landscape has not been quantitatively characterized. We analyzed Web of Science Core Collection publications on June 19, 2026, retrieving 896 articles, of which 862 were included (English, unique DOI, complete metadata). Bibliometric analyses included publication trends, author impact, international collaboration mapping, keyword co-occurrence networks with spectral clustering, temporal overlay visualization, and abstract n-gram analysis. Analyses were performed in MATLAB v2025a. Output increased from 1 to 3 articles annually in the 1960–1970 s to a peak of 67 in 2024, with nearly 90
