Transfer learning in deep neural network-based receiver coil sensitivity map estimation.

Area of Science:

• Magnetic Resonance Imaging (MRI)
• Deep Learning
• Image Reconstruction

Background:

• Accurate receiver coil sensitivity maps are crucial for parallel MRI algorithms like SENSitivity Encoding (SENSE).
• Deep learning methods for sensitivity map estimation face challenges with training dataset size and generalization.
• Retraining neural networks from scratch due to data mismatches is inefficient.

Purpose of the Study:

• To address data scarcity and generalization issues in deep learning-based receiver coil sensitivity map estimation.
• To evaluate the efficacy of a transfer learning approach for 3T MRI applications.

Main Methods:

• A U-Net model pre-trained on 1.5T sensitivity maps was assessed for 3T estimation.
• End-to-end fine-tuning of the pre-trained U-Net was performed for 3T receiver coil sensitivity map estimation.

Main Results:

• The proposed transfer learning method effectively estimated 3T receiver coil sensitivity maps.
• SENSE reconstruction using the estimated maps demonstrated successful outcomes.

Conclusions:

• The transfer learning approach overcomes limitations of data scarcity and improves generalization for deep learning-based sensitivity map estimation.
• The method enables accurate SENSE reconstruction in 3T MRI, validated by PSNR, RMSE, and image profiles.