Advertise on podcast: Neuro-oncology Talks | The Deep Dive

This deep dive based on PubMed published literature from 2022 to 2025 on artificial intelligence in neurosurgery. It explore the transformative impact of artificial intelligence and robotic systems on modern neurosurgery, specifically in the realms of neuro-oncology, vascular intervention, and spinal surgery. These technologies enhance patient care through automated tumor segmentation, precise image-guided navigation, and machine learning models that predict clinical outcomes with high accuracy. While platforms like the da Vinci system and deep learning architectures offer superior precision compared to traditional methods, the literature notes that high operational costs, algorithmic transparency, and ethical concerns remain significant barriers to adoption. Furthermore, research highlights the importance of compensating for brain shift during procedures and the potential for these tools to revolutionize resident education. Ultimately, the sources advocate for a balanced approach where technological innovation supplements, rather than replaces, a neurosurgeon’s specialized judgment.

This podcast is related to the technical report published by the Brain Tumour Charity and York Health Economics Consortium 2025. It evaluates the economic and societal impact of brain tumours in the United Kingdom, projecting a total lifetime burden of £18.7 billion for patients diagnosed in 2025. The analysis incorporates diverse factors such as direct healthcare expenses, lost productivity due to early death, and the monetary valuation of diminished quality of life. Research indicates that malignant tumours drive the majority of this financial weight, while children facing the disease suffer significant long-term educational disadvantages. The findings emphasize that premature mortality is the primary contributor to the overall societal cost. Ultimately, the document suggests that earlier diagnosis and targeted policy interventions could substantially alleviate the human and financial strain caused by these conditions. This highlights the need for new technologies and research into AI which will likely address a major part of the unmet need of early detection along with other aspects. Link to the report: https://www.thebraintumourcharity.org/

Discussion based on the paper titled "From artificial to organic: Rethinking the roots of intelligence for digital health" (2025) This academic article argues for a conceptual shift in how we define intelligence, moving away from the "artificial" label toward a framework of organic versus inorganic systems. The authors contend that current computational models are not separate from nature but are actually distillations of human cognition, trained on data and biological principles rooted in organic life. By recognizing AI as an inorganic extension of organic wisdom, the text suggests we can better address ethical accountability and bias, particularly within the field of digital health. Ultimately, the source advocates for an interdisciplinary approach that treats intelligence as a continuum of adaptation rather than a divide between man and machine. This perspective aims to improve clinical safety and organizational efficiency as medical technology moves toward more autonomous systems. Paper Link: https://doi.org/10.1371/journal.pdig.0001109 Journal: PLOS Digital Health (open access) Authors: P Ghimire and K Ashkan

Summary of paper titled "Metabolic signatures derived from whole‑brain MR‑spectroscopy identify early tumor progression in high‑grade gliomas using machine learning" by Rivera et al published in Journal of Neuro-oncology (Open Access) Link to the paper: https://link.springer.com/article/10.1007/s11060-024-04812-1 This research article investigates the use of whole-brain magnetic resonance spectroscopy (WB-MRS) combined with machine learning (ML) to predict the progression of high-grade gliomas. The study utilizes a novel WB-MRS protocol to analyze five key metabolites within seven brain regions, leveraging ML models to identify patterns indicative of tumor growth. Gradient boosting, a specific ML algorithm, demonstrated the highest accuracy in predicting tumor progression, particularly within fluid-attenuated inversion recovery (FLAIR) signals, achieving a mean area under the curve (AUC) of 0.86. The researchers developed an interactive web application, the Miami Glioma Prediction Map (MGPM), to disseminate their findings and facilitate further research. The study concludes that this ML-based WB-MRS approach offers a promising, non-invasive method for early detection of glioma recurrence, potentially improving patient outcomes by guiding timely treatment decisions.

Summary of the paper "Predicting intraoperative 5-ALA-induced tumor fluorescence via MRI and deep learning in gliomas with radiographic lower-grade characteristics" by Molina et al published in Journal of Neuro-Oncology (2024) Link for the paper: https://rdcu.be/d2vpq This research article investigates the use of a deep learning model to predict intraoperative fluorescence in lower-grade gliomas using preoperative MRI. The model, combining a U-Net and a Random Forest classifier, achieved a mean balanced accuracy of 80% in predicting fluorescence, which is induced by the administration of 5-aminolevulinic acid (5-ALA). This prediction could improve surgical planning by identifying optimal candidates for 5-ALA use, thereby facilitating more precise tumor resection and reducing the risk of undergrading. The study used a large dataset of patients with lower-grade gliomas, incorporating data from multiple MRI scanners, and addressed the challenges of data heterogeneity through careful preprocessing. Despite its success, further improvements to the model and dataset are suggested for future research.

Summary of the paper "Foundation models for fast, label-free detection of glioma infiltration" by Kondepudi et al in Nature (Open Access). Link to the paper: https://doi.org/10.1038/s41586-024-08169-3 This study presents FastGlioma, a visual foundation model for the rapid and accurate detection of glioma infiltration in fresh surgical tissue. This model was trained on a massive dataset of stimulated Raman histology (SRH) images and then fine-tuned to predict the degree of tumor infiltration within a whole-slide image. The model’s performance was tested in a prospective, multicenter cohort of patients with diffuse glioma, and it demonstrated excellent accuracy in identifying tumor infiltration, outperforming traditional image-guided and fluorescence-guided surgical adjuncts. The study also highlights the model’s interpretability through few-shot visualizations, its potential for zero-shot generalization to other brain tumor diagnoses, and its potential for improving the safety and efficacy of glioma surgery.

Summary of the paper "Bayesian networks for Risk Assessment and postoperative deficit prediction in intraoperative neurophysiology for brain surgery" by Pescador et al in Journal of Clinical Monitoring and Computing (2024) Link for the paper | https://doi.org/10.1007/s10877-024-01159-w This study explores the effectiveness of Intraoperative Neurophysiological Monitoring (IONM) during brain surgery using Bayesian Networks (a type of machine learning/AI technique). While traditional randomized controlled trials are difficult to conduct in this field, Bayesian Networks offer a mathematical approach to assess the value of IONM by analyzing prior data and predicting postoperative outcomes. The study found that IONM, especially when corrective actions are taken in response to signal changes, can significantly reduce neurological deficits in patients, improving their recovery and reducing healthcare costs. However, the study highlights limitations, such as the relatively small sample size and the need for further research to better understand the impact of time on neurological outcomes.

Summary of the paper "Ethnicity in neuro-oncology research: How are we doing and how can we do better?"by Mirza et al (2024) published in Journal of Neuro-Oncology (open access) Link for the paper | https://doi.org/10.1007/s11060-024-04769-1 This research article examines the underrepresentation of ethnic minorities in neuro-oncology clinical trials and investigates the impact of ethnicity on treatment outcomes. The authors conducted a systematic review and meta-analysis of Phase III and IV trials, finding a significant lack of ethnicity data reporting, particularly regarding outcomes. While their meta-analysis did not reveal significant differences in survival by ethnicity, the authors acknowledge that the limited data restricts their ability to draw definitive conclusions. The study highlights the critical need for more inclusive recruitment strategies, improved reporting standards, and further research to understand the role of ethnicity in neuro-oncology.

Summary of the paper "Accuracy of Radiomics in Predicting IDH Mutation Status inDiffuse Gliomas: A Bivariate Meta-Analysis" published in Radiology: Artificial Intelligence by Salle et al. Dec 2023 (open access) Link for the paper | https://doi.org/10.1148/ryai.220257 This paper reviews 26 studies that utilized radiomics to predict the presence of an isocitrate dehydrogenase (IDH) mutation in patients with diffuse gliomas (a type of brain tumor). Radiomics involves analyzing medical images with artificial intelligence to extract features and predict clinical outcomes. The authors found that radiomics achieved a pooled sensitivity of 79% and specificity of 80% in detecting IDH mutations. However, the quality of the included studies was generally low, which potentially introduced bias into the results. Despite the promising potential of radiomics, the study concludes that further research is needed to improve the methodology and performance of these algorithms before widespread clinical adoption.

Summary of the paper "Intraoperative Neuromonitoring of the Visual Pathway in Asleep Neuro-Oncology Surgery" by Soumpasis, Ghimire, Lavrador et al published in Cancers (Open access). Link to the paper | https://doi.org/10.3390/cancers15153943 This research article examines the effectiveness of intraoperative visual evoked potential (VEP) monitoring during brain tumor surgery in visually eloquent areas. The study assessed both transcranial and direct cortical VEP recordings, and the results indicate that direct cortical recordings, when feasible, are more strongly correlated with postoperative visual outcomes. Notably, the study found that invasion of the optic radiation by the tumor is a significant predictor of poor visual field outcomes, independent of the tumor type. The authors conclude that utilizing a combination of VEP monitoring and tractography can contribute to better prediction of long-term visual field outcomes, although the technique is prone to false warnings and requires further investigation.

Summary of paper "Intraoperative mapping of pre-central motor cortex and subcortex: a proposal for supplemental cortical and novel subcortical maps to Penfield’s motor homunculus" by Ghimire et al. (Open Access) Link for the paper | https://doi.org/10.1007/s00429-021-02274-z This study proposes supplemental maps of the motor cortex and corticospinal tract, aiming to improve the accuracy and safety of surgeries involving these areas. The authors reviewed past research on motor mapping and analyzed data from 180 patients who underwent surgery for brain lesions, specifically focusing on the representation of intercostal muscles, which had not been previously described in detail. The study uses advanced imaging techniques like fMRI and DTI alongside intraoperative stimulation mapping to create these maps. These updated maps will help surgeons understand the complex functional anatomy of the motor pathway and make better decisions during surgeries involving eloquent brain regions.

Summary of paper "Cortical resting motor threshold difference in asleep-awake craniotomy for motor eloquent gliomas: WHO grading influences motor pathway excitability" by Pescador, Lavrador, Ghimire et al. (open access) Link for the paper | https://doi.org/10.1093/cercor/bhad493 This research paper, published in the journal Cerebral Cortex, explores the relationship between tumor grade and cortical motor excitability during awake and asleep phases of surgery in patients with gliomas. The researchers used intraoperative neuromonitoring to measure resting motor threshold (RMT) differences in the motor cortex, and found that higher tumor grade correlated with a greater excitability difference between awake and asleep states. ROC analysis identified a 3 mA difference as the best predictor of high-grade glioma, while a Bayesian Network analysis confirmed that an excitability difference above 3 mA indicated a 75.8% probability of a high-grade glioma. These findings suggest that intraoperative neuromonitoring could provide valuable information about tumor grade and potentially guide surgical decisions.

Summary of the paper "MGMT methylation and its prognostic significance in inoperable IDH-wildtype glioblastoma: the MGMT-GBM study" by Ghimire et al published in Acta Neurocirúrgica (Open access). Link for the paper | https://doi.org/10.1007/s00701-024-06300-x This research article examines the prognostic significance of MGMT methylation in patients with inoperable IDH-wildtype glioblastoma who undergo biopsy and systemic treatment. The authors analyze data from a retrospective study of 142 patients, grouping them based on their MGMT methylation status and the type of treatment they received. Their findings indicate that methylated MGMT status is associated with better overall survival, regardless of the treatment modality. The authors also develop a Cox proportional hazard model to predict overall survival in these patients, identifying MGMT status and treatment categories as statistically significant predictors. The study highlights the importance of MGMT methylation as a predictive biomarker for overall survival in this patient population and underscores the potential benefits of tailoring treatment based on methylation status.

Summary of "Radiogenomic biomarkers for immunotherapy in glioblastoma: A systematic review of magnetic resonance imaging studies" by Ghimire et al published in Neuro-oncology Advances (open access paper). Link for the paper | https://doi.org/10.1093/noajnl/vdae055 This systematic review examines the current state of research on radiogenomic biomarkers for immunotherapy in glioblastoma. The authors searched multiple databases, including PubMed, MEDLINE, and Embase, to identify relevant studies published between 1990 and 2023. The search was conducted according to PRISMA guidelines, and the quality of included studies was assessed using QUADAS 2 and CLAIM checklists. The review identified nine studies, all retrospective, exploring the relationship between MRI features and immune-related genomic markers in glioblastoma. Seven of the nine studies were classified as radiogenomic.