Neuro-oncology advances最新文献

{"title":"Exploring adult glioma through MRI: A review of publicly available datasets to guide efficient image analysis.","authors":"Meryem Abbad Andaloussi, Raphael Maser, Frank Hertel, François Lamoline, Andreas Dominik Husch","doi":"10.1093/noajnl/vdae197","DOIUrl":"10.1093/noajnl/vdae197","url":null,"abstract":"<p><strong>Background: </strong>Publicly available data are essential for the progress of medical image analysis, in particular for crafting machine learning models. Glioma is the most common group of primary brain tumors, and magnetic resonance imaging (MRI) is a widely used modality in their diagnosis and treatment. However, the availability and quality of public datasets for glioma MRI are not well known.</p><p><strong>Methods: </strong>In this review, we searched for public datasets of glioma MRI using Google Dataset Search, The Cancer Imaging Archive, and Synapse.</p><p><strong>Results: </strong>A total of 28 datasets published between 2005 and May 2024 were found, containing 62 019 images from 5515 patients. We analyzed the characteristics of these datasets, such as the origin, size, format, annotation, and accessibility. Additionally, we examined the distribution of tumor types, grades, and stages among the datasets. The implications of the evolution of the World Health Organization (WHO) classification on tumors of the brain are discussed, in particular the 2021 update that significantly changed the definition of glioblastoma.</p><p><strong>Conclusions: </strong>Potential research questions that could be explored using these datasets were highlighted, such as tumor evolution through malignant transformation, MRI normalization, and tumor segmentation. Interestingly, only 2 datasets among the 28 studied reflect the current WHO classification. This review provides a comprehensive overview of the publicly available datasets for glioma MRI currently at our disposal, providing aid to medical image analysis researchers in their decision-making on efficient dataset choice.</p>","PeriodicalId":94157,"journal":{"name":"Neuro-oncology advances","volume":"7 1","pages":"vdae197"},"PeriodicalIF":3.7,"publicationDate":"2025-01-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11773385/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143061789","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"AI-guided virtual biopsy: Automated differentiation of cerebral gliomas from other benign and malignant MRI findings using deep learning.","authors":"Mathias Holtkamp, Vicky Parmar, René Hosch, Luca Salhöfer, Hanna Styczen, Yan Li, Marcel Opitz, Martin Glas, Nika Guberina, Karsten Wrede, Cornelius Deuschl, Michael Forsting, Felix Nensa, Lale Umutlu, Johannes Haubold","doi":"10.1093/noajnl/vdae225","DOIUrl":"10.1093/noajnl/vdae225","url":null,"abstract":"<p><strong>Background: </strong>This study aimed to develop an automated algorithm to noninvasively distinguish gliomas from other intracranial pathologies, preventing misdiagnosis and ensuring accurate analysis before further glioma assessment.</p><p><strong>Methods: </strong>A cohort of 1280 patients with a variety of intracranial pathologies was included. It comprised 218 gliomas (mean age 54.76 ± 13.74 years; 136 males, 82 females), 514 patients with brain metastases (mean age 59.28 ± 12.36 years; 228 males, 286 females), 366 patients with inflammatory lesions (mean age 41.94 ± 14.57 years; 142 males, 224 females), 99 intracerebral hemorrhages (mean age 62.68 ± 16.64 years; 56 males, 43 females), and 83 meningiomas (mean age 63.99 ± 13.31 years; 25 males, 58 females). Radiomic features were extracted from fluid-attenuated inversion recovery (FLAIR), contrast-enhanced, and noncontrast T1-weighted MR sequences. Subcohorts, with 80% for training and 20% for testing, were established for model validation. Machine learning models, primarily XGBoost, were trained to distinguish gliomas from other pathologies.</p><p><strong>Results: </strong>The study demonstrated promising results in distinguishing gliomas from various intracranial pathologies. The best-performing model consistently achieved high area-under-the-curve (AUC) values, indicating strong discriminatory power across multiple distinctions, including gliomas versus metastases (AUC = 0.96), gliomas versus inflammatory lesions (AUC = 1.0), gliomas versus intracerebral hemorrhages (AUC = 0.99), gliomas versus meningiomas (AUC = 0.98). Additionally, across all these entities, gliomas had an AUC of 0.94.</p><p><strong>Conclusions: </strong>The study presents an automated approach that effectively distinguishes gliomas from common intracranial pathologies. This can serve as a quality control upstream to further artificial-intelligence-based genetic analysis of cerebral gliomas.</p>","PeriodicalId":94157,"journal":{"name":"Neuro-oncology advances","volume":"7 1","pages":"vdae225"},"PeriodicalIF":3.7,"publicationDate":"2025-01-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11773384/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143061842","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Therapeutic manipulation and bypass of the blood-brain barrier: powerful tools in glioma treatment.","authors":"Alexandra M Giantini-Larsen, Abhinav Pandey, Andrew L A Garton, Margherita Rampichini, Graham Winston, Jacob L Goldberg, Rajiv Magge, Philip E Stieg, Mark M Souweidane, Rohan Ramakrishna","doi":"10.1093/noajnl/vdae201","DOIUrl":"10.1093/noajnl/vdae201","url":null,"abstract":"<p><p>The blood-brain barrier (BBB) remains an obstacle for delivery of chemotherapeutic agents to gliomas. High grade and recurrent gliomas continue to portend a poor prognosis. Multiple methods of bypassing or manipulating the BBB have been explored, including hyperosmolar therapy, convection-enhanced delivery (CED), laser-guided interstitial thermal therapy (LITT), and Magnetic Resonance Guided Focused Ultrasound (MRgFUS) to enhance delivery of chemotherapeutic agents to glial neoplasms. Here, we review these techniques, currently ongoing clinical trials to disrupt or bypass the BBB in gliomas, and the results of completed trials.</p>","PeriodicalId":94157,"journal":{"name":"Neuro-oncology advances","volume":"7 1","pages":"vdae201"},"PeriodicalIF":3.7,"publicationDate":"2025-01-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11773386/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143061806","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"The value of AI for assessing longitudinal brain metastases treatment response.","authors":"Vincent Andrearczyk, Luis Schiappacasse, Matthieu Raccaud, Jean Bourhis, John O Prior, Michel A Cuendet, Andreas F Hottinger, Vincent Dunet, Adrien Depeursinge","doi":"10.1093/noajnl/vdae216","DOIUrl":"10.1093/noajnl/vdae216","url":null,"abstract":"<p><strong>Background: </strong>Effective follow-up of brain metastasis (BM) patients post-treatment is crucial for adapting therapies and detecting new lesions. Current guidelines (Response Assessment in Neuro-Oncology-BM) have limitations, such as patient-level assessments and arbitrary lesion selection, which may not reflect outcomes in high tumor burden cases. Accurate, reproducible, and automated response assessments can improve follow-up decisions, including (1) optimizing re-treatment timing to avoid treating responding lesions or delaying treatment of progressive ones, and (2) enhancing precision in evaluating responses during clinical trials.</p><p><strong>Methods: </strong>We compared manual and automatic (deep learning-based) lesion contouring using unidimensional and volumetric criteria. Analysis focused on (1) agreement in size and RANO-BM categories, (2) stability of measurements under scanner rotations and over time, and (3) predictability of 1-year outcomes. The study included 49 BM patients, with 184 MRI studies and 448 lesions, retrospectively assessed by radiologists.</p><p><strong>Results: </strong>Automatic contouring and volumetric criteria demonstrated superior stability (<i>P</i> < .001 for rotation; <i>P</i> < .05 over time) and better outcome predictability compared to manual methods. These approaches reduced observer variability, offering reliable and efficient response assessments. The best outcome predictability, defined as 1-year response, was achieved using automatic contours and volumetric measurements. These findings highlight the potential of automated tools to streamline clinical workflows and provide consistency across evaluators, regardless of expertise.</p><p><strong>Conclusion: </strong>Automatic BM contouring and volumetric measurements provide promising tools to improve follow-up and treatment decisions in BM management. By enhancing precision and reproducibility, these methods can streamline clinical workflows and improve the evaluation of response in trials and practice.</p>","PeriodicalId":94157,"journal":{"name":"Neuro-oncology advances","volume":"7 1","pages":"vdae216"},"PeriodicalIF":3.7,"publicationDate":"2025-01-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11786217/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143082098","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"High-dose MTX-based polychemotherapy for primary CNS lymphoma in younger patients: Long-term efficacy of the modified Bonn protocol.","authors":"Thomas Zeyen, Johannes Weller, Matthias Schneider, Anna-Laura Potthoff, Christina Schaub, Lea Roever, Eleni Gkika, Hartmut Vatter, Tobias A W Holderried, Peter Brossart, Ulrich Herrlinger, Niklas Schaefer","doi":"10.1093/noajnl/vdaf005","DOIUrl":"10.1093/noajnl/vdaf005","url":null,"abstract":"<p><strong>Background: </strong>Polychemotherapy based on high-dose methotrexate (HD-MTX) is the standard therapy for newly diagnosed younger patients (<65 years) with primary CNS lymphoma (PCNSL). In the modified Bonn protocol, consolidation therapy consists of intraventricular chemotherapy that is added to the continuation of HD-MTX-based chemotherapy. This study investigates the efficacy and toxicity of the modified Bonn protocol in first-line therapy of young patients with PCNSL.</p><p><strong>Methods: </strong>All consecutive immunocompetent patients aged <65 years who were newly diagnosed with PCNSL from 2012 to 2021 and started first-line therapy with the modified Bonn protocol at the Neurooncology Center Bonn were included in this retrospective analysis. Treatment comprised 3 courses of rituximab/HD-MTX/IFO followed by consolidation therapy with 2 courses of HD-AraC and 2 courses of HD-MTX/IFO, including intrathecal MTX and intrathecal AraC. Progression-free and overall survival were evaluated.</p><p><strong>Results: </strong>Forty-three patients were included. Thirty-seven patients (86%) reached intrathecal consolidation therapy. Grade 3/4 toxicity was observed in 58.1%. The median PFS was 102.8 months; 5-year OS rate was 76% (median not reached). Eighteen patients developing refractory/relapsed PCNSL received second-line therapy using the modified Freiburg protocol (AraC/TT +/- HD-MTX/rituximab followed by BCNU/TT-based HD-ASCT). A second relapse was observed in 11/18 patients (median follow-up of 17 months (IQR 5-43.7 months)).</p><p><strong>Conclusions: </strong>First-line treatment of PCNSL with the modified Bonn protocol is highly effective. The outcome compares well with other seemingly more toxic PCNSL protocols for younger patients. In patients with disease recurrence, second-line therapy according to the modified Freiburg protocol appears to be effective.</p>","PeriodicalId":94157,"journal":{"name":"Neuro-oncology advances","volume":"7 1","pages":"vdaf005"},"PeriodicalIF":3.7,"publicationDate":"2025-01-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11788595/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143124236","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Mapping glioma's impact on cognition: Insights from macrostructure, microstructure, and beyond.","authors":"Nuria Cayuela, Cristina Izquierdo, Lucía Vaquero, Estela Càmara, Jordi Bruna, Marta Simó","doi":"10.1093/noajnl/vdaf003","DOIUrl":"https://doi.org/10.1093/noajnl/vdaf003","url":null,"abstract":"<p><strong>Background: </strong>Cognitive impairment (CI) significantly impacts the quality of life of glioma patients. The main contributing risk factors include tumor characteristics, treatment-related factors, and their complex interplay. This review explores the role of advanced structural neuroimaging techniques in understanding CI in glioma patients.</p><p><strong>Methods: </strong>A literature search was conducted in PubMed, PsycINFO, and ISI Web of Knowledge using specific keywords. We included studies with advanced magnetic resonance imaging techniques and objective neuropsychological exams.</p><p><strong>Results: </strong>At diagnosis, during the pre-surgery phase, associations between glioma characteristics and cognitive outcomes have been described. Specifically, patients with isocitrate dehydrogenase (IDH)-wild-type gliomas exhibit more adverse cognitive outcomes, accompanied by disruptions in gray (GM) and white matter (WM) networks when compared to IDH-mutant. In addition, pre- and post-surgery imaging analyses highlight the importance of preserving specific WM tracts, such as the inferior longitudinal and arcuate fasciculus, in mitigating verbal memory and language processing decline. Furthermore, examining gliomas in perisylvian regions emphasizes deleterious effects on various cognitive domains. Additionally, it has been suggested that neuroplastic reorganization could serve as a compensatory mechanism against CI. Lastly, a limited number of studies suggest long-term CI linked to GM atrophy and leukoencephalopathy induced by radiotherapy ± chemotherapy in glioma survivors, highlighting the need for improving treatment approaches, particularly for patients with extended survival expectations.</p><p><strong>Conclusion: </strong>This review underscores the need for nuanced understanding and an individual approach in the management of glioma patients. Neuroplastic insights offer clinicians valuable guidance in surgical decision-making and personalized therapeutic approaches thus improving patient outcomes in neuro-oncology.</p>","PeriodicalId":94157,"journal":{"name":"Neuro-oncology advances","volume":"7 1","pages":"vdaf003"},"PeriodicalIF":3.7,"publicationDate":"2025-01-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11795312/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143257584","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"cIMPACT-NOW update 9: Recommendations on utilization of genome-wide DNA methylation profiling for central nervous system tumor diagnostics.","authors":"Kenneth Aldape, David Capper, Andreas von Deimling, Caterina Giannini, Mark R Gilbert, Cynthia Hawkins, Jürgen Hench, Thomas S Jacques, David Jones, David N Louis, Sabine Mueller, Brent A Orr, MacLean Nasrallah, Stefan M Pfister, Felix Sahm, Matija Snuderl, David Solomon, Pascale Varlet, Pieter Wesseling","doi":"10.1093/noajnl/vdae228","DOIUrl":"10.1093/noajnl/vdae228","url":null,"abstract":"<p><p>Genome-wide DNA methylation signatures correlate with and distinguish central nervous system (CNS) tumor types. Since the publication of the initial CNS tumor DNA methylation classifier in 2018, this platform has been increasingly used as a diagnostic tool for CNS tumors, with multiple studies showing the value and utility of DNA methylation-based classification of CNS tumors. A Consortium to Inform Molecular and Practical Approaches to CNS Tumor Taxonomy (cIMPACT-NOW) Working Group was therefore convened to describe the current state of the field and to provide advice based on lessons learned to date. Here, we provide recommendations for the use of DNA methylation-based classification in CNS tumor diagnostics, emphasizing the attributes and limitations of the modality. We emphasize that the methylation classifier is one diagnostic tool to be used alongside previously established diagnostic tools in a fully integrated fashion. In addition, we provide examples of the inclusion of DNA methylation data within the layered diagnostic reporting format endorsed by the World Health Organization (WHO) and the International Collaboration on Cancer Reporting. We emphasize the need for backward compatibility of future platforms to enable accumulated data to be compatible with new versions of the array. Finally, we outline the specific connections between methylation classes and CNS WHO tumor types to aid in the interpretation of classifier results. It is hoped that this update will assist the neuro-oncology community in the interpretation of DNA methylation classifier results to facilitate the accurate diagnosis of CNS tumors and thereby help guide patient management.</p>","PeriodicalId":94157,"journal":{"name":"Neuro-oncology advances","volume":"7 1","pages":"vdae228"},"PeriodicalIF":3.7,"publicationDate":"2025-01-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11788596/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143124235","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A machine learning-assisted systematic review of preclinical glioma modeling: Is practice changing with the times?","authors":"Theodore C Hirst, Emma Wilson, Declan Browne, Emily S Sena","doi":"10.1093/noajnl/vdae193","DOIUrl":"10.1093/noajnl/vdae193","url":null,"abstract":"<p><strong>Background: </strong>Despite improvements in our understanding of glioblastoma pathophysiology, there have been no major improvements in treatment in recent years. Animal models are a vital tool for investigating cancer biology and its treatment, but have known limitations. There have been advances in glioblastoma modeling techniques in this century although it is unclear to what extent they have been adopted.</p><p><strong>Methods: </strong>We searched Pubmed and EMBASE using terms designed to identify all publications reporting an animal glioma experiment, using a machine learning algorithm to assist with screening. We reviewed the full text of a sample of 1000 articles and then used the findings to inform a screen of all included abstracts to appraise the modeling applications across the entire dataset.</p><p><strong>Results: </strong>The search identified 26 201 publications of which 13 783 were included at screening. The automated screening had high sensitivity but limited specificity. We observed a dominance of traditional cell line paradigms and the emergence of advanced tumor model systems eclipsed by a large increase in the volume of cell line experiments. Few studies used more than 1 model in vivo and most publications did not verify critical genetic features.</p><p><strong>Conclusions: </strong>Advanced models have clear advantages in terms of tumor and disease recapitulation and have largely not replaced traditional cell lines which have a number of critical deficiencies that limit their viability in modern animal research. The judicious use of advanced models or more relevant cell lines might improve the translational relevance of future animal glioblastoma experimentation.</p>","PeriodicalId":94157,"journal":{"name":"Neuro-oncology advances","volume":"6 1","pages":"vdae193"},"PeriodicalIF":3.7,"publicationDate":"2024-12-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11680884/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142904627","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Cuproplasia-related gene signature: Prognostic insights for glioma therapy.","authors":"Toni Rose Jue, Joseph Descallar, Vu Viet Hoang Pham, Jessica Lilian Bell, Tyler Shai-Hee, Riccardo Cazzolli, Sumanth Nagabushan, Eng-Siew Koh, Orazio Vittorio","doi":"10.1093/noajnl/vdae233","DOIUrl":"10.1093/noajnl/vdae233","url":null,"abstract":"<p><strong>Background: </strong>Adult-type diffuse gliomas encompass nearly a quarter of all primary tumors found in the CNS, including astrocytoma, oligodendroglioma, and glioblastoma. Histopathological tumor grade and molecular profile distinctly impact patient survival. Despite treatment advancements, patients with recurrent glioma have a very poor clinical outcome, warranting improved risk stratification to determine therapeutic interventions. Various studies have shown that copper is a notable trace element that is crucial for biological processes and has been shown to display pro-tumorigenic functions in cancer, particularly gliomas.</p><p><strong>Methods: </strong>Differential gene expression, Cox regression, and least absolute shrinkage and selection operator regression were used to identify 19 copper-homeostasis-related gene signatures using TCGA lower-grade glioma and glioblastoma (GBM) cohorts. The GLASS Consortium dataset was used as an independent validation cohort. Enrichment analysis revealed the involvement of the signature in various cancer-related pathways and biological processes. Using this CHRG signature, a risk score model and a nomogram were developed to predict survival in glioma patients.</p><p><strong>Results: </strong>Our prognostic CHRG signature stratified patients into high- and low-risk groups, demonstrating robust predictive performance. High-risk groups showed poorer survival outcomes. The nomogram model integrating CHRG signature and clinical features accurately predicted 1-, 3-, and 5-year survival rates in both training and test sets.</p><p><strong>Conclusions: </strong>The identified 19-gene CHRG signature holds promise as a prognostic tool, enabling accurate risk stratification and survival prediction in glioma patients. Integrating this signature with clinical characteristics enhances prognostic accuracy, underscoring its potential clinical utility in optimizing therapeutic strategies and patient care in glioma management.</p>","PeriodicalId":94157,"journal":{"name":"Neuro-oncology advances","volume":"7 1","pages":"vdae233"},"PeriodicalIF":3.7,"publicationDate":"2024-12-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11786221/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143082958","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"The potential of GPT-4 advanced data analysis for radiomics-based machine learning models.","authors":"Martha Foltyn-Dumitru, Aditya Rastogi, Jaeyoung Cho, Marianne Schell, Mustafa Ahmed Mahmutoglu, Tobias Kessler, Felix Sahm, Wolfgang Wick, Martin Bendszus, Gianluca Brugnara, Philipp Vollmuth","doi":"10.1093/noajnl/vdae230","DOIUrl":"https://doi.org/10.1093/noajnl/vdae230","url":null,"abstract":"<p><strong>Background: </strong>This study aimed to explore the potential of the Advanced Data Analytics (ADA) package of GPT-4 to autonomously develop machine learning models (MLMs) for predicting glioma molecular types using radiomics from MRI.</p><p><strong>Methods: </strong>Radiomic features were extracted from preoperative MRI of <i>n</i> = 615 newly diagnosed glioma patients to predict glioma molecular types (IDH-wildtype vs IDH-mutant 1p19q-codeleted vs IDH-mutant 1p19q-non-codeleted) with a multiclass ML approach. Specifically, ADA was used to autonomously develop an ML pipeline and benchmark performance against an established handcrafted model using various MRI normalization methods (N4, Zscore, and WhiteStripe). External validation was performed on 2 public glioma datasets D2 (<i>n</i> = 160) and D3 (<i>n</i> = 410).</p><p><strong>Results: </strong>GPT-4 achieved the highest accuracy of 0.820 (95% CI = 0.819-0.821) on the D3 dataset with N4/WS normalization, significantly outperforming the benchmark model's accuracy of 0.678 (95% CI = 0.677-0.680) (<i>P</i> < .001). Class-wise analysis showed performance variations across different glioma types. In the IDH-wildtype group, GPT-4 had a recall of 0.997 (95% CI = 0.997-0.997), surpassing the benchmark's 0.742 (95% CI = 0.740-0.743). For the IDH-mut 1p/19q-non-codel group, GPT-4's recall was 0.275 (95% CI = 0.272-0.279), lower than the benchmark's 0.426 (95% CI = 0.423-0.430). In the IDH-mut 1p/19q-codel group, GPT-4's recall was 0.199 (95% CI = 0.191-0.206), below the benchmark's 0.730 (95% CI = 0.721-0.738). On the D2 dataset, GPT-4's accuracy was significantly lower (<i>P</i> < .001) than the benchmark's, with N4/WS achieving 0.668 (95% CI = 0.666-0.671) compared with 0.719 (95% CI = 0.717-0.722) (<i>P</i> < .001). Class-wise analysis revealed the same pattern as observed in D3.</p><p><strong>Conclusions: </strong>GPT-4 can autonomously develop radiomics-based MLMs, achieving performance comparable to handcrafted MLMs. However, its poorer class-wise performance due to unbalanced datasets shows limitations in handling complete end-to-end ML pipelines.</p>","PeriodicalId":94157,"journal":{"name":"Neuro-oncology advances","volume":"7 1","pages":"vdae230"},"PeriodicalIF":3.7,"publicationDate":"2024-12-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11707530/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142960798","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}