LSTM-based Prediction of QQ Model Parameters and OEF Estimation via Brain MRI Signal Decay Modeling

Key Points

• The aim is to accurately estimate oxygen extraction fraction (OEF) using MRI signal decay modeling.
• Modeling mGRE signal decay as an exponential function.
• Estimating general parameters independent of scanning conditions.
• Using an LSTM-based deep learning model to predict five QQ model parameters.
• Calculating OEF based on predicted parameters.
• Conducting experiments on data from 18 patients.
• Proposed method reduced prediction error for parameters compared to existing clustering methods.
• OEF root mean square error (RMSE) decreased to approximately 5.23±0.035, a reduction of about 0.12.
• Demonstrated robustness against variations in scanning conditions.

Abstract