Machine Learning Tools for Image-Based Glioma Grading and the Quality of Their Reporting: Challenges and...

Area of Science:

• Radiology and Medical Imaging
• Artificial Intelligence in Medicine
• Neuro-oncology

Background:

• Machine learning (ML) tools are increasingly used to enhance radiological practice.
• ML applications in neuro-oncology, particularly for glioma grade prediction, have grown substantially.
• Pre-operative glioma grade prediction using medical imaging is a key area of ML focus.

Purpose of the Study:

• To review ML models for glioma grade prediction.
• To describe their development workflow and algorithms.
• To highlight challenges and suggest improvements for clinical implementation.

Main Methods:

• Literature review of ML models for glioma grade prediction.
• Analysis of model development workflows and classifier algorithms.
• Discussion of challenges including data sources, validation, and reporting.

Main Results:

• Numerous ML models for glioma grade prediction exist in the literature.
• Common developmental workflows and classifier algorithms are identified.
• Key challenges in data, validation, and reporting quality are highlighted.

Conclusions:

• Improving reporting guidelines and risk of bias tools is crucial.
• Addressing current challenges can facilitate the clinical implementation of ML glioma grade prediction.
• Future work should focus on robust validation and transparent reporting.