Application of Spectral Approach Combined with U-NETs for Quantitative Microwave Breast Imaging

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Electromagnetics

Background:

• Accurate breast imaging is crucial for early disease detection.
• Quantitative dielectric property estimation can improve diagnostic accuracy.
• Current imaging modalities have limitations in speed and quantitative accuracy.

Purpose of the Study:

• To develop a real-time quantitative human breast imaging system.
• To leverage a spectral approach combined with deep learning for improved imaging.
• To validate the proposed method using numerical simulations.

Main Methods:

• A spectral approach based on the Fourier diffraction theorem was employed.
• A pair of U-NETs were trained for spectral transformation.
• Training data was generated using a spectral database of anthropomorphic cavities.
• Optimization was performed using the Adam optimizer with WMAPE loss.

Main Results:

• The U-NET pair successfully transformed induced current spectra to contrast dielectric spectra.
• Numerical results demonstrated the feasibility and accuracy of the proposed imaging concept.
• The system achieved real-time quantitative imaging capabilities.

Conclusions:

• The proposed spectral imaging technique with U-NETs shows significant promise for real-time quantitative breast imaging.
• This approach offers a potential advancement over existing breast imaging methods.
• Further validation in clinical settings is warranted.