Distinguishing glioblastoma from brain metastasis; a systematic review and meta-analysis on the performance of machine learning

Abstract Background The discrimination of glioblastoma and solitary metastasis brain tumor is challenging. Up now, several conventional and advanced imaging modalities were used for distinguishing between these tumors with different success rates. We systematically reviewed the studies reported the...

Full description

Saved in:
Bibliographic Details
Main Authors: Mohammad Amin Habibi, Reza Omid, Shafaq Asgarzade, Sadaf Derakhshandeh, Ali Soltani Farsani, Zohreh Tajabadi
Format: Article
Language:English
Published: SpringerOpen 2025-02-01
Series:Egyptian Journal of Neurosurgery
Subjects:
Online Access:https://doi.org/10.1186/s41984-025-00386-w
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1823862843295924224
author Mohammad Amin Habibi
Reza Omid
Shafaq Asgarzade
Sadaf Derakhshandeh
Ali Soltani Farsani
Zohreh Tajabadi
author_facet Mohammad Amin Habibi
Reza Omid
Shafaq Asgarzade
Sadaf Derakhshandeh
Ali Soltani Farsani
Zohreh Tajabadi
author_sort Mohammad Amin Habibi
collection DOAJ
description Abstract Background The discrimination of glioblastoma and solitary metastasis brain tumor is challenging. Up now, several conventional and advanced imaging modalities were used for distinguishing between these tumors with different success rates. We systematically reviewed the studies reported the performance of machine learning (ML) algorithms for accurately discrimination of these two entities. Method The search was conducted from inception to 1 June, 2023, in PubMed/Medline, Embase, Scopus, and Web of Science to find out the studies investigated the performance of ML-based algorithm for differentiation of glioblastoma and metastatic brain tumor. Results This study included 28 studies comprising a total of 2,860 patients. The meta-analysis model results revealed a pooled sensitivity and specificity estimate of 0.83 [0.80–0.86] and 0.87 [0.83–0.90], respectively, indicating a commendable overall diagnostic accuracy across all the studies. ResNet50 and ResNet50-LSTM have shown promising results with single-study sensitivities of up to 88.9% and 88.2%, respectively. Furthermore, the integration of CNNs and RNNs has demonstrated improved performance compared to standalone models in a significant portion of the studies. The ROC curve area was 0.90, indicating high discriminative ability. The positive likelihood ratio was 6.2, and the negative likelihood ratio was 0.20, providing helpful information on how test results modified pretest probability. Conclusion ML applied to routine neuroimaging shows high diagnostic potential for glioblastoma detection. While more research is needed before clinical deployment, preliminary results are encouraging.
format Article
id doaj-art-c511d61460f943deb503739a4cd32330
institution Kabale University
issn 2520-8225
language English
publishDate 2025-02-01
publisher SpringerOpen
record_format Article
series Egyptian Journal of Neurosurgery
spelling doaj-art-c511d61460f943deb503739a4cd323302025-02-09T12:25:11ZengSpringerOpenEgyptian Journal of Neurosurgery2520-82252025-02-0140111110.1186/s41984-025-00386-wDistinguishing glioblastoma from brain metastasis; a systematic review and meta-analysis on the performance of machine learningMohammad Amin Habibi0Reza Omid1Shafaq Asgarzade2Sadaf Derakhshandeh3Ali Soltani Farsani4Zohreh Tajabadi5Department of Neurosurgery, Shariati Hospital, Tehran University of Medical ScienceCenter of Orthopedic Trans-Disiplinary Applied Research (COTAR), School of Medicine, Tehran University of Medical ScienceSchool of Medicine, Tabriz University of Medical SciencesSchool of Medicine, Tehran Medical Sciences, Islamic Azad UniversitySchool of Medicine, Tehran University of Medical SciencesDigestive Disease Research Institute, Shariati Hospital, Tehran University of Medical ScienceAbstract Background The discrimination of glioblastoma and solitary metastasis brain tumor is challenging. Up now, several conventional and advanced imaging modalities were used for distinguishing between these tumors with different success rates. We systematically reviewed the studies reported the performance of machine learning (ML) algorithms for accurately discrimination of these two entities. Method The search was conducted from inception to 1 June, 2023, in PubMed/Medline, Embase, Scopus, and Web of Science to find out the studies investigated the performance of ML-based algorithm for differentiation of glioblastoma and metastatic brain tumor. Results This study included 28 studies comprising a total of 2,860 patients. The meta-analysis model results revealed a pooled sensitivity and specificity estimate of 0.83 [0.80–0.86] and 0.87 [0.83–0.90], respectively, indicating a commendable overall diagnostic accuracy across all the studies. ResNet50 and ResNet50-LSTM have shown promising results with single-study sensitivities of up to 88.9% and 88.2%, respectively. Furthermore, the integration of CNNs and RNNs has demonstrated improved performance compared to standalone models in a significant portion of the studies. The ROC curve area was 0.90, indicating high discriminative ability. The positive likelihood ratio was 6.2, and the negative likelihood ratio was 0.20, providing helpful information on how test results modified pretest probability. Conclusion ML applied to routine neuroimaging shows high diagnostic potential for glioblastoma detection. While more research is needed before clinical deployment, preliminary results are encouraging.https://doi.org/10.1186/s41984-025-00386-wDeep learningRadiomicsGBMGliomaMRIArtificial intelligence
spellingShingle Mohammad Amin Habibi
Reza Omid
Shafaq Asgarzade
Sadaf Derakhshandeh
Ali Soltani Farsani
Zohreh Tajabadi
Distinguishing glioblastoma from brain metastasis; a systematic review and meta-analysis on the performance of machine learning
Egyptian Journal of Neurosurgery
Deep learning
Radiomics
GBM
Glioma
MRI
Artificial intelligence
title Distinguishing glioblastoma from brain metastasis; a systematic review and meta-analysis on the performance of machine learning
title_full Distinguishing glioblastoma from brain metastasis; a systematic review and meta-analysis on the performance of machine learning
title_fullStr Distinguishing glioblastoma from brain metastasis; a systematic review and meta-analysis on the performance of machine learning
title_full_unstemmed Distinguishing glioblastoma from brain metastasis; a systematic review and meta-analysis on the performance of machine learning
title_short Distinguishing glioblastoma from brain metastasis; a systematic review and meta-analysis on the performance of machine learning
title_sort distinguishing glioblastoma from brain metastasis a systematic review and meta analysis on the performance of machine learning
topic Deep learning
Radiomics
GBM
Glioma
MRI
Artificial intelligence
url https://doi.org/10.1186/s41984-025-00386-w
work_keys_str_mv AT mohammadaminhabibi distinguishingglioblastomafrombrainmetastasisasystematicreviewandmetaanalysisontheperformanceofmachinelearning
AT rezaomid distinguishingglioblastomafrombrainmetastasisasystematicreviewandmetaanalysisontheperformanceofmachinelearning
AT shafaqasgarzade distinguishingglioblastomafrombrainmetastasisasystematicreviewandmetaanalysisontheperformanceofmachinelearning
AT sadafderakhshandeh distinguishingglioblastomafrombrainmetastasisasystematicreviewandmetaanalysisontheperformanceofmachinelearning
AT alisoltanifarsani distinguishingglioblastomafrombrainmetastasisasystematicreviewandmetaanalysisontheperformanceofmachinelearning
AT zohrehtajabadi distinguishingglioblastomafrombrainmetastasisasystematicreviewandmetaanalysisontheperformanceofmachinelearning