Steven Brem
VerifiedUniversity of Pennsylvania · Rehabilitation Medicine
Active 1972–2025
About
Steven Brem, M.D., is a Professor and Chief of Neurosurgery at the University of Pennsylvania Perelman School of Medicine. He serves as the Director of Medical Student Education in the Department of Neurosurgery and is the Medical Director of the Center for Precision Surgery at the Abramson Cancer Center. His clinical expertise includes complex brain tumor surgery, surgery of intrinsic and metastatic brain tumors, pituitary tumors, acoustic neuromas, meningiomas, skull base surgery, brain mapping, and health outcomes. His research focuses on angiogenesis, neuro-oncology, translational drug discovery, personalized medicine, genomics, biomarkers, experimental therapeutics, brain mapping, and clinical guidelines related to brain tumors and metastases. Brem has contributed to the field through numerous publications and is involved in innovative approaches such as vagus nerve stimulation for cancer treatment and machine learning-based prognostic subgrouping of glioblastoma.
Research topics
- Oncology
- Medicine
- Artificial Intelligence
- Computer Science
- Internal medicine
- Cancer research
- Pathology
- Surgery
- Computational biology
- Bioinformatics
- Biology
- Machine Learning
- Genetics
- Dermatology
Selected publications
International Journal of Radiation Oncology*Biology*Physics · 2025-09-01
articleJCO Precision Oncology · 2025-09-01 · 1 citations
articleOpen accessPURPOSE Because of tumor heterogeneity and sampling error, next-generation sequencing (NGS) of glioblastoma (GBM) tumors may provide an incomplete picture of the somatic mutational landscape. We hypothesized that simultaneous targeted NGS of matched tumor tissue and cerebrospinal fluid (CSF), obtained during craniotomy for resection of GBM, would lead to identification of clinically relevant variants not detected by tissue NGS alone. METHODS We enrolled 50 patients undergoing resection of newly diagnosed (n = 15) or recurrent (n = 35) GBM. CSF was collected intraoperatively via the subarachnoid space (n = 25) or lateral ventricle (n = 25) and assayed by NGS using a hybrid capture liquid biopsy panel. Matched tumor tissue also underwent large panel hybrid capture NGS testing. RESULTS CSF samples from 28 of 50 patients (56%) passed quality control metrics. At least one CSF variant was detected in 25 of 28 patients (89%), and 22 of 28 patients had matched tissue sequencing results available. In these 22 patients (primary analysis cohort), the median number of variants detected in CSF was higher than in tissue (3 v 2 variants, respectively; P = .0035), and 15 of 22 patients (68%) had ≥1 CSF variant not detected in matched tissue, including clinically relevant alterations in EGFR , PMS2 , PIK3CA , and TP53 . CONCLUSION The addition of intraoperatively acquired CSF liquid biopsy to tissue NGS in patients with GBM may improve detection of clinically relevant variants, potentially improving selection of patients for clinical trials.
International Journal of Radiation Oncology*Biology*Physics · 2025-09-01
articleBrain Sciences · 2025-01-31 · 1 citations
articleOpen accessPurpose: Early treatment response assessments are crucial, and the results are known to better correlate with prognosis and survival outcomes. The present study was conducted to differentiate true progression (TP) from pseudoprogression (PsP) in long-term-surviving glioblastoma patients using our previously established multiparametric MRI-based predictive model, as well as to identify clinical factors impacting survival outcomes in these patients. Methods: We report six patients with glioblastoma that had an overall survival longer than 5 years. When tumor specimens were available from second-stage surgery, histopathological analyses were used to classify between TP (>25% characteristics of malignant neoplasms; n = 2) and PsP (<25% characteristics of malignant neoplasms; n = 2). In the absence of histopathology, modified RANO criteria were assessed to determine the presence of TP (n = 1) or PsP (n = 1). The predictive probabilities (PPs) of tumor progression were measured from contrast-enhancing regions of neoplasms using a multiparametric MRI-based prediction model. Subsequently, these PP values were used to define each lesion as TP (PP ≥ 50%) or PsP (PP < 50%). Additionally, detailed clinical information was collected. Results: Our predictive model correctly identified all patients with TP (n = 3) and PsP (n = 3) cases, reflecting a significant concordance between histopathology/modified RANO criteria and PP values. The overall survival varied from 5.1 to 12.3 years. Five of the six glioblastoma patients were MGMT promoter methylated. All patients were female, with a median age of 56 years. Moreover, all six patients had a good functional status (KPS ≥ 70), underwent near-total/complete resection, and received alternative therapies. Conclusions: Multiparametric MRI can aid in assessing treatment response in long-term-surviving glioblastoma patients.
Neuro-Oncology · 2025-11-01
articleAbstract PURPOSE Glioblastoma is the most aggressive adult brain tumor, with a median overall survival of approximately 15 months. It is important to build accurate prognostic models for glioblastoma patients to inform clinical management and trials. This study proposes a self-supervised learning-based approach with multimodal data integration for survival prediction and prognostic stratification of glioblastoma patients on the ReSPOND consortium. METHODS We curated a multi-parametric MRI dataset (T1, T1CE, T2, FLAIR) of 3,119 glioblastoma patients from 22 institutions across 3 continents. Masked autoencoder (MAE) was adapted to pretrain a Vision Transformer (ViT) encoder by reconstructing the masked image patches. The encoder was utilized for extracting patch embeddings for survival tasks, with cross-attention mechanism to incorporate the molecular and clinical information (age, sex, extent of resection, MGMT) to guide imaging feature aggregation. Imaging and clinical embeddings were fused through a multi-layer perceptron (MLP) for log-risk hazard estimation, optimized using Cox partial likelihood. Model performance and generalizability were assessed via k-fold cross-validation on the ReSPOND consortium and the leave-one-site-out validation was performed on 11 institutions comparing with CoxPH, DeepSurv and DeepHit. Prognostic risk stratification via Kaplan-Meier analysis divided the patients into low-, medium- and high-risk subgroups per site. RESULTS Multimodal data integration using the proposed framework achieved the highest C-index (0.674 ± 0.017) on the ReSPOND consortium. Integration of clinical information and MGMT consistently boosted the performance of the proposed model across sites (0.615 ± 0.046 vs. 0.662 ± 0.044). The imaging-based approaches, i.e., radiomics and convolutional neural network (CNN) features performed less robustly. The Kaplan-Meier curves and log-rank tests suggested the proposed framework achieved more separable prognostic subgroups. CONCLUSION The proposed self-supervised multimodal learning framework shows promise for survival prediction and prognostic risk stratification in glioblastoma. It highlights the challenge for clinical model deployment due to the data heterogeneity in multi-institutional cohort.
Neuro-Oncology · 2025-11-01
articleOpen accessAbstract PURPOSE This study aims to identify distinct imaging subtypes of glioblastoma using multi-modal MRIs from the ReSPOND consortium, providing insights into tumor heterogeneity to inform personalized treatment approaches. METHODS We analyzed 3,145 subjects with multi-modal MRIs (T1, T2, T1Gd, FLAIR) from 16 geographically distinct institutions across North America, Europe, and Asia. Radiomic features were extracted using Masked Auto-Encoder (MAE), a deep learning architecture trained through self-supervised learning on 23,608 MRIs across 11-dimensional MRI-derived imaging measures (conventional, diffusion, and perfusion protocols). For clustering analysis, we used features extracted from the four structural modalities available in all subjects. Following feature extraction, we applied ComBat harmonization to remove institutional batch effects, performed dimensionality reduction using cross-validated PCA, and employed K-medoids clustering with consensus analysis (100 random data partitions with a 90/10 split for clustering/validation) to identify stable subtypes, with optimal cluster number determined using adjusted rand index. Additionally, we extracted morphological, intensity, and textural features to characterize the identified subtypes in an interpretable manner. RESULTS Our analysis revealed three reproducible glioblastoma subtypes. Statistical analysis demonstrated significant differences between subtypes in tumor morphology and spatial location features. Subtype 1 showed centrally located, spherical tumors; Subtype 2 exhibited peripheral, irregular morphologies; Subtype 3 had mixed features (p &lt; 0.001). Key features included right hemisphere ratio, centroid Z position, sphericity, and shape convexity ratio (p &lt; 0.001). Survival analysis indicated median overall survival of 12.3, 14.5, and 16.8 months for Subtypes 1, 2, and 3, respectively (p = 0.042, borderline). Preliminary genomic analysis showed frequent TP53 and RB1 co-mutations, with distinct molecular patterns per subtype. CONCLUSION This study demonstrates that multi-modal MRIs can successfully identify glioblastoma subtypes characterized by deep learning derived radiomic features. Future work will focus on validation with comprehensive molecular and tissue-based profiles, survival outcomes, and treatment responses to support personalized medicine.
Neuro-Oncology · 2025-03-17 · 5 citations
articleOpen accessBACKGROUND: For newly diagnosed glioblastoma (GBM), a combination of upfront surgical immunotherapy with aglatimagene besadenovec (CAN-2409), followed by chemoradiation and then adjuvant nivolumab has not been tested. The aim of this study was to test the safety of this regimen and determine metrics of immune activation that may correlate with clinical outcomes. METHODS: Forty-one patients with suspected newly diagnosed GBM by imaging were enrolled in this multi-institutional, open-label, phase 1b clinical trial before surgical resection. Frozen section confirmation of high-grade glioma was required for administration of CAN-2409. This was then followed with chemoradiation and adjuvant nivolumab. Tumor and blood were assayed for genetic and immune markers before and during treatment. RESULTS: The regimen was well tolerated and generated measurable immune activation. Factors linked to survival were identified, such as baseline mutated gene pairs (eg, MED15/HRC), tumor immune cell composition, and changes in systemic cytokine, immune cells, and T-cell diversity. The most significant serial systemic immune changes were observed in a long-term survivor subset of patients with gross total resection (GTR)/methylated methylguanine methyltransferase (MGMT) promoter tumors. Median overall survival (mOS) in these patients was 30.6 months, while it was less for patients with unmethylated or subtotal resections. CONCLUSIONS: These findings suggest the opportunity for patient stratification and the potential for more durable antitumor immune responses in future clinical trials of this multimodal standard of care and combined immunotherapy regimen. ClinicalTrials.gov identifier: NCT03576612.
Neuro-Oncology Advances · 2024-01-01 · 4 citations
articleOpen accessAbstract Background It is imperative to differentiate true progression (TP) from pseudoprogression (PsP) in glioblastomas (GBMs). We sought to investigate the potential of physiologically sensitive quantitative parameters derived from diffusion and perfusion magnetic resonance imaging (MRI), and molecular signature combined with machine learning in distinguishing TP from PsP in GBMs in the present study. Methods GBM patients (n = 93) exhibiting contrast-enhancing lesions within 6 months after completion of standard treatment underwent 3T MRI. Final data analyses were performed on 75 patients as O6-methylguanine-DNA-methyltransferase (MGMT) status was available only from these patients. Subsequently, patients were classified as TP (n = 55) or PsP (n = 20) based on histological features or mRANO criteria. Quantitative parameters were computed from contrast-enhancing regions of neoplasms. PsP datasets were artificially augmented to achieve balanced class distribution in 2 groups (TP and PsP). A random forest algorithm was applied to select the optimized features. The data were randomly split into training and testing subsets in an 8:2 ratio. To develop a robust prediction model in distinguishing TP from PsP, several machine-learning classifiers were employed. The cross-validation and receiver operating characteristic (ROC) curve analyses were performed to determine the diagnostic performance. Results The quadratic support vector machine was found to be the best classifier in distinguishing TP from PsP with a training accuracy of 91%, cross-validation accuracy of 86%, and testing accuracy of 85%. Additionally, ROC analysis revealed an accuracy of 85%, sensitivity of 70%, and specificity of 100%. Conclusions Machine learning using quantitative multiparametric MRI may be a promising approach to distinguishing TP from PsP in GBMs.
Proceedings on CD-ROM - International Society for Magnetic Resonance in Medicine. Scientific Meeting and Exhibition/Proceedings of the International Society for Magnetic Resonance in Medicine, Scientific Meeting and Exhibition · 2024-08-14
articleIDH mutation has become one of the most important prognostic biomarkers in glioma management, regardless of histopathological features. The oncometabolite 2HG has been proposed as a biomarker for IDH-specific genetic profiles for gliomas. We report clinical utility of SVS and 1H-MRSI using a long TE (97 ms) in assessing IDH-mutant gliomas by detecting the characteristic resonances of 2HG. Our results from 25 patients showed sensitivity and specificity of 77% and 83%, respectively. In conclusion, 1H-MRS with optimized TE can accurately detect 2HG levels, which has significant clinical implications for determining prognosis and evaluating therapeutic efficacy for targeted and/or alternative therapies.
Neuro-Oncology · 2024-11-01 · 1 citations
articleOpen accessAbstract BACKGROUND The purpose of this study was to assess the feasibility and effectiveness of AI-guided personalized precision radiation therapy (PPRT) with targeted dose escalation in enhancing outcomes for patients with newly diagnosed glioblastoma. METHODS An open-label trial was conducted at the University of Pennsylvania (NCT03477513) from August 2018 to April 2023, enrolling 20 patients with IDH-wildtype glioblastoma who underwent maximal safe resection. They were matched with a contemporaneous cohort of glioblastoma patients based on age, sex, extent of resection (EOR), O6-methylguanine-DNA methyltransferase promoter (MGMTp) methylation status, and IDH status. Propensity score matching (PSM) was employed to select the control group, utilizing one-to-four nearest neighbor matching. One patient was excluded due to concurrent treatment with tumor treating fields (TTFields). The PPRT group received personalized radiation dose escalation to 75 Gy in 30 fractions guided by AI-based predictive modeling of recurrence, along with temozolomide chemotherapy. The control group received standard-of-care chemoradiotherapy, 60 Gy in 30 fractions. A previously published and evaluated (retrospectively and prospectively) predictive AI model, which has demonstrated high predictive value for neoplastic cell infiltration and future tumor recurrence using preoperative, multi-parametric MRIs, was utilized. RESULTS Median overall survival was 24.3 months in the PPRT-temozolomide group compared to 17.5 months in the standard-of-care treatment group (hazard ratio [HR]=0.30; 95% CI: 0.16-0.57; p&lt;0.001). Excluding two patients with leptomeningeal disease and bone marrow metastasis, the median survival was 35.4 months (HR=0.25; 95% CI: 0.12-0.51; p&lt;0.001). CONCLUSION AI-guided PPRT for newly diagnosed glioblastoma patients demonstrated feasibility in routine clinical practice and significantly improved overall survival compared to matched controls receiving standard-of-care treatment. These findings underscore the potential of personalized precision radiation therapy, with focused dose escalation, in improving outcomes for glioblastoma patients and emphasize the need for prospective validation in a randomized controlled clinical trial.
Recent grants
Connectomics meets neuro-oncology: mapping the brain for treatment planning
NIH · $2.7M · 2016–2022
NIH · $100k · 1994
NIH · $454k · 2003
Frequent coauthors
- 147 shared
Donald M. O’Rourke
University of Pennsylvania
- 120 shared
Stephen Gately
Translational Genomics Research Institute
- 104 shared
Stephen Bagley
University of Pennsylvania
- 101 shared
Arati Desai
University of Alabama at Birmingham
- 95 shared
Shingo Takano
University of Tsukuba Hospital
- 90 shared
MacLean P. Nasrallah
University of Pennsylvania
- 66 shared
Ana-Maria C. Tsanaclis
Universidade de São Paulo
- 66 shared
Michael Klagsbrun
- Resume-aware match score
- Save to shortlist
- AI-drafted outreach
See your match with Steven Brem
PhdFit ranks faculty by your research interests, methods, and publications — grounded in their actual work, not templates.
- Free to start
- No credit card
- 30-second signup