Breast cancer detection using interferometric MUSIC: experimental and numerical assessment

Area of Science:

• Microwave imaging for biomedical applications.
• Advanced signal processing techniques for radar systems.
• Breast cancer detection and characterization.

Background:

• Proximity of sensors in microwave breast cancer detection causes antenna coupling, degrading system performance.
• A priori characterization of radio frequency systems is challenging due to coupling, impacting detection efficacy.
• Existing methods like migration and wideband MUSIC (WB-MUSIC) struggle with unknown antenna responses.

Purpose of the Study:

• To demonstrate the advantages of the interferometric multiple signal classification (I-MUSIC) approach.
• To highlight I-MUSIC's reduced dependence on a priori antenna information.
• To compare I-MUSIC performance against migration and WB-MUSIC using SCR, SMR, and SD metrics.

Main Methods:

• Acquired data using a multi-monostatic radar system on a synthetic breast phantom (1-3 GHz).
• Phantom mimicked dielectric properties of breast tissues, including a 5mm tumor.
• Processed data with I-MUSIC, noncoherent migration, and WB-MUSIC; benchmarked with numerical simulations (CST Microwave Studio).

Main Results:

• I-MUSIC demonstrated superior performance over WB-MUSIC and noncoherent migration in both experimental and numerical datasets.
• Experimental results showed I-MUSIC achieved at least 5x better SCR and 2.3x better SMR.
• Numerical simulations reduced delocalization effect with I-MUSIC, achieving an SD of 1.61 with an antipodal Vivaldi antenna.

Conclusions:

• I-MUSIC consistently generated high signal-to-clutter levels (above 5.46 dB experimentally, 7.61 dB numerically).
• The method achieved up to 7.84 dB SCR with specific antennas, outperforming comparative techniques.
• I-MUSIC successfully detected a 5mm tumor with a permittivity contrast of 1.5, proving its efficacy.