Sign in Subscribe
Machine learning

Advances in Machine Learning and Deep Learning in Glioblastoma Diagnosis and Prognosis

Strategic Insights

— 2 min read

Researchers in Algeria have made significant strides in using machine learning and deep learning algorithms to improve the diagnosis and prognosis of glioblastoma, a highly malignant primary brain cancer. Traditional diagnosis and prognosis are often subjective and time-consuming, whereas advances in machine learning and deep learning have accelerated research by enhancing tumor segmentation, molecular characterization, and survival prediction. A systematic review and meta-analysis of 44 studies published between 2021 and 2025 demonstrated that machine and deep learning models performed strongly across various clinical tasks, including overall survival prognosis, tumor segmentation, and molecular tests.

Key Takeaways:

  • The research analyzed 44 studies published between 2021 and 2025, using PubMed, Scopus, and Web of Science as sources.
  • Machine and deep learning models demonstrated strong performance across different clinical tasks, including overall survival prognosis, tumor segmentation, and molecular tests.
  • The pooled C-index for overall survival prognosis was 0.78 (95% CI 0.74-0.82, I = 68.5%).
  • The tumor segmentation models had a high average Dice Similarity Coefficient value of 0.91 (95% CI 0.87-0.94, I = 45.2%).
  • Molecular tests were highly accurate for predicting IDH1 mutation (pooled accuracy = 90.5%, 95% CI 88.1% to 92.8%) and MGMT methylation status (pooled accuracy = 97.8%, 95% CI 96.4% to 99.1%).
  • Transformer models outperformed CNN in segmentation, and radionics-based ML improved non-invasive molecular assessment.
  • The research concluded that, although AI techniques demonstrated encouraging results, substantial challenges still preclude efficient clinical applicability.

Statistics:

  • 44 studies were analyzed in the systematic review and meta-analysis.
  • 2021-2025: the publication period of the analyzed studies.
  • 0.78 (95% CI 0.74-0.82, I = 68.5%): pooled C-index for overall survival prognosis.
  • 0.91 (95% CI 0.87-0.94, I = 45.2%): average Dice Similarity Coefficient value for tumor segmentation models.
  • 90.5% (95% CI 88.1% to 92.8%): pooled accuracy for IDH1 mutation prediction.
  • 97.8% (95% CI 96.4% to 99.1%): pooled accuracy for MGMT methylation status prediction.

Sources:

  • NewsRx. Report Summarizes Personalized Medicine Study Findings from Hadja Fatima Tbahriti and Co-Researchers (Machine learning and deep learning in glioblastoma: a systematic review and meta-analysis of diagnosis, prognosis, and treatment). Drug Week. August 29, 2025; p 2448.
  • Discover Oncology. Machine learning and deep learning in glioblastoma: a systematic review and meta-analysis of diagnosis, prognosis, and treatment. Discover Oncology, 2025;16(1):1492.
  • Hadja Fatima Tbahriti, Ali Boukadoum, Meriem Benbernou, and Mohamed Belhocine. Machine learning and deep learning in glioblastoma: a systematic review and meta-analysis of diagnosis, prognosis, and treatment. Discover Oncology, 2025;16(1):1492.
  • Springer, One New York Plaza, Suite 4600, New York, Ny, United States.

Read more