Predicting chemo-brain in breast cancer survivors using multiple MRI features and machine-learning

Area of Science:

• Neuroscience
• Oncology
• Medical Imaging

Background:

• Breast cancer (BC) is the most prevalent cancer globally in women.
• Chemotherapy, while vital for BC treatment, can induce cognitive impairment known as chemo-brain.
• Early detection of chemo-brain is crucial for managing treatment side effects.

Purpose of the Study:

• To develop and validate machine-learning (ML) models for identifying chemo-brain in breast cancer survivors.
• To detect subtle brain alterations in post-chemotherapy patients using advanced imaging techniques.

Main Methods:

• Recruited 19 breast cancer patients undergoing chemotherapy and 20 healthy controls (HCs).
• Acquired resting-state functional MRI and generalized q-sampling imaging (GQI) data from all participants.
• Employed ML algorithms including logistic regression (LR), decision tree classifier (CART), and XGBoost (XGB) for classification.

Main Results:

• Multiple ML models, utilizing GQI indices, regional homogeneity, fractional anisotropy, and quantitative anisotropy, demonstrated strong classification performance.
• Leave-one-out cross-validation achieved a peak accuracy of 84% in distinguishing between chemo-brain and HC groups.
• LR, CART, and XGB models, using various feature sets, were most effective in this classification task.

Conclusions:

• Successfully constructed ML models capable of differentiating chemo-brains from healthy brains.
• These findings hold promise for future clinical applications in monitoring chemo-brain effects in breast cancer patients.