Tissue Recognition with Deep Ensemble Learning of Ultrasound Wavelet Spectra

Conventional ultrasound imaging presents the scattering amplitude of echo signal by local brightness whereas neglecting the spectral information. However, ultrasound harmonic imaging demonstrated that ultrasound echoes carry abundant spectral information that varies among organs, tissues, and microbubble contrasts. We hypothesize that the spectral pattern potentially acts as a fingerprint to identify the tissues in various organs, which may lead to high contrast ultrasonic imaging and easier interpretation. In this research, we propose to recognize each tissue by its spectral patterns, and use color-coding to exhibit the recognized tissues for a high-contrast and high-specificity ultrasound imaging