Frequency-Differencing Method to Kickstart Waveform Inversion Without Cycle Skipping

Because of its fast reconstruction times, frequency-domain full waveform inversion is the clinical gold standard for high-resolution sound speed imaging in medical ultrasound tomography; however, it fundamentally relies on having sufficiently low-frequency data to avoiding cycle skipping. This motivates the development of new strategies that can overcome cycle skipping in the absence of low-frequency data. In this work, we synthesize approximate low-frequency content from existing high-frequency signals using the same frequency differencing method used in frequency-difference beamforming for underwater sonar. The approximate low-frequency content is then used to kickstart full-waveform inversion so that cycle skipping can be better avoided at higher frequencies. We demonstrate the efficacy of this new approach using in-vivo breast imaging cases. In each case, internal tissue structures are seen with much greater clarity without cycle skipping artifacts when frequency differencing is used to provide a starting model for conventional full-waveform inversion.