An Unsupervised Brain Extraction Quality Control Approach for Efficient Neuro-Oncology Studies.

Area of Science:

• Neuroimaging
• Medical Image Analysis
• Computational Neuroscience

Background:

• Automated brain extraction is crucial for neuroimaging analysis and patient privacy.
• Manual brain mask creation is time-consuming, necessitating automated methods.
• Existing automated quality control (QC) methods lack flexibility and adaptability.

Purpose of the Study:

• To introduce a novel, unsupervised, and efficient QC method for automated brain extraction.
• To address limitations in flexibility and data adaptability of previous QC approaches.
• To improve the reliability of brain masks in large-scale neuroimaging studies.

Main Methods:

• Developed a QC approach based on statistical outlier detection.
• Evaluated performance using morphological features of brain masks from three automated tools.
• Tested on multi-institutional pre- and post-operative glioblastoma MRI scans.

Main Results:

• Achieved high accuracy: 0.9 for pre-operative and 0.87 for post-operative scans.
• Demonstrated the effectiveness of the unsupervised QC tool.
• Showcased the method's resource efficiency and minimal parameter tuning requirements.

Conclusions:

• The novel QC method significantly enhances flexibility and efficiency in brain mask validation.
• The tool is valuable for ensuring brain mask quality in neuroimaging.
• The approach has potential for adaptation to other applications requiring robust QC.