Machine learning applications to neuroimaging for glioma detection and classification: An artificial intelligence...

Area of Science:

• Neuro-oncology
• Radiology
• Artificial Intelligence

Background:

• Glioma is a common brain tumor with poor high-grade survival rates.
• Magnetic resonance imaging (MRI) is crucial for glioma detection and monitoring.
• Definitive glioma diagnosis currently relies on surgical pathology.

Purpose of the Study:

• To systematically review machine learning (ML) applications in glioma MRI data analysis.
• To explore ML for automating systematic review processes in this field.
• To synthesize ML tools and data resources for future research.

Main Methods:

• Systematic review of 153 studies on ML and glioma MRI.
• Data extraction and analysis of ML applications.
• Natural language processing (NLP) for keyword extraction, topic modeling, and document classification.

Main Results:

• ML is applied to glioma grading, diagnosis, segmentation, biomarker identification, progression detection, and survival prediction.
• Top ML algorithms include convolutional neural networks, support vector machines, and random forests.
• Model performance metrics were generally strong (e.g., AUC 0.87 ± 0.09).
• NLP and transfer learning facilitated automated systematic review screening.

Conclusions:

• Machine learning demonstrates substantial utility in processing glioma MRI data.
• ML applications show potential to improve clinical practice and patient outcomes.
• Automated systematic reviews can accelerate the translation of research findings to clinical application.