Federation of Brain Age Estimation in Structural Neuroimaging Data
Summary
Decentralized models enable accurate brain age prediction without sharing sensitive neuroimaging data. This approach maintains performance comparable to centralized methods, facilitating broader research in brain health.
Area of Science:
- Neuroscience
- Medical Imaging
- Artificial Intelligence
Background:
- Brain age estimation uses neuroimaging to assess neurological and psychiatric disorders' impact on brain development and aging.
- Accurate brain age prediction aids in early diagnosis and treatment monitoring.
- Data privacy and acquisition challenges limit access to large neuroimaging datasets.
Purpose of the Study:
- To propose and evaluate a decentralized approach for brain age prediction.
- To overcome data access limitations in neuroimaging research.
- To maintain prediction accuracy without centralizing sensitive brain data.
Main Methods:
- Utilized structural Magnetic Resonance Imaging (MRI) data.
- Developed and implemented a decentralized machine learning model for brain age prediction.
- Evaluated model performance against a centralized training approach.
Main Results:
- The decentralized brain age prediction model achieved performance comparable to models trained on centrally pooled data.
- Demonstrated the feasibility of decentralized learning for brain age estimation.
- Validated the model's effectiveness using extracted features from structural MRI.
Conclusions:
- Decentralized models offer a viable solution for brain age prediction, addressing data privacy and accessibility issues.
- This approach supports collaborative research in neuroimaging without compromising data security.
- The proposed method holds potential for advancing the diagnosis and monitoring of brain disorders.
