Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
Multicenter Study
. 2025 May 15;27(4):1102-1115.
doi: 10.1093/neuonc/noae260.

Machine learning-based prognostic subgrouping of glioblastoma: A multicenter study

Hamed Akbari  1 Spyridon Bakas  2   3   4   5 Chiharu Sako  6   7   8 Anahita Fathi Kazerooni  9   10 Javier Villanueva-Meyer  11 Jose A Garcia  6   7   8 Elizabeth Mamourian  6   7   8 Fang Liu  12   13 Quy Cao  12   13 Russell T Shinohara  6   8   12   13 Ujjwal Baid  2   3 Alexander Getka  6   7   8 Sarthak Pati  2 Ashish Singh  6   7   8 Evan Calabrese  14 Susan Chang  15 Jeffrey Rudie  16 Aristeidis Sotiras  17 Pamela LaMontagne  17 Daniel S Marcus  17 Mikhail Milchenko  17 Arash Nazeri  17 Carmen Balana  18 Jaume Capellades  19 Josep Puig  20 Chaitra Badve  21 Jill S Barnholtz-Sloan  22   23   24 Andrew E Sloan  25   26 Vachan Vadmal  27 Kristin Waite  24   27 Murat Ak  28 Rivka R Colen  28   29 Yae Won Park  30 Sung Soo Ahn  30 Jong Hee Chang  31   32 Yoon Seong Choi  33   34 Seung-Koo Lee  30 Gregory S Alexander  35 Ayesha S Ali  35 Adam P Dicker  35 Adam E Flanders  36 Spencer Liem  35 Joseph Lombardo  37   35 Wenyin Shi  35 Gaurav Shukla  6   8   35   38 Brent Griffith  39 Laila M Poisson  40   41 Lisa R Rogers  41 Aikaterini Kotrotsou  42 Thomas C Booth  43   44 Rajan Jain  45   46 Matthew Lee  45 Abhishek Mahajan  47   48 Arnab Chakravarti  49 Joshua D Palmer  49 Dominic DiCostanzo  49 Hassan Fathallah-Shaykh  50 Santiago Cepeda  51 Orazio Santo Santonocito  52 Anna Luisa Di Stefano  52 Benedikt Wiestler  53 Elias R Melhem  54 Graeme F Woodworth  54   55 Pallavi Tiwari  56   57 Pablo Valdes  58 Yuji Matsumoto  59 Yoshihiro Otani  59 Ryoji Imoto  59 Mariam Aboian  7 Shinichiro Koizumi  60 Kazuhiko Kurozumi  60 Toru Kawakatsu  60 Kimberley Alexander  61   62   63 Laveniya Satgunaseelan  61   62   63 Aaron M Rulseh  64 Stephen J Bagley  65   66 Michel Bilello  6   7   8 Zev A Binder  9   66 Steven Brem  9   66 Arati S Desai  66 Robert A Lustig  67 Eileen Maloney  9 Timothy Prior  9 Nduka Amankulor  9   66 MacLean P Nasrallah  6   8   66   68 Donald M O'Rourke  9   66 Suyash Mohan  6   7   8 Christos Davatzikos  6   7   8 ReSPOND consortium
Collaborators, Affiliations
Multicenter Study

Machine learning-based prognostic subgrouping of glioblastoma: A multicenter study

Hamed Akbari et al. Neuro Oncol. .

Abstract

Background: Glioblastoma (GBM) is the most aggressive adult primary brain cancer, characterized by significant heterogeneity, posing challenges for patient management, treatment planning, and clinical trial stratification.

Methods: We developed a highly reproducible, personalized prognostication, and clinical subgrouping system using machine learning (ML) on routine clinical data, magnetic resonance imaging (MRI), and molecular measures from 2838 demographically diverse patients across 22 institutions and 3 continents. Patients were stratified into favorable, intermediate, and poor prognostic subgroups (I, II, and III) using Kaplan-Meier analysis (Cox proportional model and hazard ratios [HR]).

Results: The ML model stratified patients into distinct prognostic subgroups with HRs between subgroups I-II and I-III of 1.62 (95% CI: 1.43-1.84, P < .001) and 3.48 (95% CI: 2.94-4.11, P < .001), respectively. Analysis of imaging features revealed several tumor properties contributing unique prognostic value, supporting the feasibility of a generalizable prognostic classification system in a diverse cohort.

Conclusions: Our ML model demonstrates extensive reproducibility and online accessibility, utilizing routine imaging data rather than complex imaging protocols. This platform offers a unique approach to personalized patient management and clinical trial stratification in GBM.

Keywords: glioblastoma; machine learning; mpMRI; prognostic subgrouping; survival.

PubMed Disclaimer

Figures

Figure 1.
Figure 1.
Schematic illustration of machine learning (ML)-based subgrouping analysis pipeline.
Figure 2.
Figure 2.
Overall survival map of all deceased ReSPOND patients. Brain regions associated with longer overall survival are shown on one end of the spectrum, while regions linked to shorter overall survival are shown on the opposite end. The color bar indicates the distribution of survival durations across the patient population. Abbreviation: ReSPOND = Radiomics Signatures for PrecisiON Diagnostics.
Figure 3.
Figure 3.
Kaplan–Meier survival curves. Actual survival on the x-axis is compared among each of the 3 groups based on subgroups and integrated risk assessments. The top left panel shows Kaplan–Meier survival curves for 3 subgroups; the top right panel is boxplots of the subgroups versus the patients’ survival. On each box, the central line indicates the median and the bottom and top edges of the box indicate the 25th and 75th percentiles, respectively. The whiskers extend to the most extreme data points not considered outliers. The bottom panel shows Kaplan–Meier survival curves of integrated risk stratification with different combinations of predictors. Abbreviations: EOR = extent of resection; MGMT = O6-methylguanine-DNA methyltransferase; SPI = survival prediction index.
Figure 4.
Figure 4.
Features versus actual survival of deceased patients in the full ReSPOND cohort. (A) The Concordance Index between patients’ survival and each feature. (B) The Concordance Index between SPI and OS per institution. (C) Boxplots of Z-scored features for short, medium, and long survival patients. On each box, the central mark indicates the median, and the bottom and top edges of the box indicate the 25th and 75th percentiles, respectively. Abbreviations: BV = brain volume; CAT = Catalan Institute of Oncology; CWRU = Case Western Reserve University; Dist2Vent = distance to ventricles; ED = volume of edema; ET = volume of enhancing tumor; HF = Henry Ford Health; KCL = King’s College London; MDA = MD Anderson Cancer Center; NC = volume of non-enhancing core; NYU = New York University; OS = overall survival; OSU = Ohio State University; Penn = University of Pennsylvania; ReSPOND = Radiomics Signatures for PrecisiON Diagnostics; RH = University Hospital Río Hortega; SM = overall survival map; SPI = survival prediction index; TC = volume of tumor core; TJU = Thomas Jefferson University; TMC = Tata Memorial Center; UAB = University of Alabama at Birmingham; UCSF = University of California, San Francisco; UPMC = University of Pittsburgh Medical Center; WashU = Washington University in St. Louis; WT = Volume of Whole Tumor.

References

    1. Ostrom QT, Cioffi G, Waite K, Kruchko C, Barnholtz-Sloan JS. CBTRUS statistical report: primary brain and other central nervous system tumors diagnosed in the United States in 2014–2018. Neuro-Oncol. 2021;23(12 Suppl 2):iii1–iii105. - PMC - PubMed
    1. Stupp R, Taillibert S, Kanner A, et al. Effect of tumor-treating fields plus maintenance temozolomide vs maintenance temozolomide alone on survival in patients with glioblastoma: a randomized clinical trial. JAMA. 2017;318(23):2306–2316. - PMC - PubMed
    1. Board WCoTE. WHO Classification of Tumours. Vol 6. 5th ed. Lyon (France): International Agency for Research on Cancer; 2021.
    1. Telloni SM. Tumor staging and grading: a primer. In: Espina V, ed. Molecular Profiling: Methods and Protocols. New York, NY: Springer New York; 2017:1–17. - PubMed
    1. Ludwig JA, Weinstein JN. Biomarkers in cancer staging, prognosis and treatment selection. Nat Rev Cancer. 2005;5(11):845–856. - PubMed

Publication types