Data-Driven Quadratic Kernel Synthesis for Nonlinear Ultrasound Imaging

Abhishek Sahoo, Emad S. Ebbini

Research output: Contribution to journalConference articlepeer-review

1 Scopus citations


Tissue Harmonic Imaging (THI) is a well studied method that reconstructs images using second harmonic echo data. However, the use of a band-pass filter to extract harmonic components reduces the signal bandwidth and hence degrades the axial resolution. Pulse Inversion (PI) is a promising technique that overcomes the issue of bandwidth and preserves axial resolution, but introduction of image artifacts due to tissue motion makes this method somewhat undesirable for processing in vivo data. In order to address the above mentioned issues, we have proposed a data-driven quadratic Volterra filter capable of capturing specific nonlinear frequency interactions to selectively enhance image quality without losing spatial resolution. The kernel synthesis is a multi-step procedure which analyzes the individual SVD decomposed kernels, filters them with 2D Gaussian filters to retain the desired frequency interactions and finally sums up the kernels coherently to reclaim the bandwidth. An in vivo dataset with and without contrast agents is used to demonstrate image quality improvement with no loss of axial resolution. The reconstructed nonlinear images also report enhancement in specificity of the heterogeneous perfusion of the tumor. Additionally, the proposed method is observed to reject acoustic artifacts and improve lateral resolution.

Original languageEnglish (US)
JournalIEEE International Ultrasonics Symposium, IUS
StatePublished - 2021
Event2021 IEEE International Ultrasonics Symposium, IUS 2021 - Virtual, Online, China
Duration: Sep 11 2011Sep 16 2011

Bibliographical note

Publisher Copyright:
© 2021 IEEE.


  • Bifrequency
  • Contrast Imaging
  • Diagnostic Ultrasound
  • Polynomial Signal Processing
  • Volterra Filter


Dive into the research topics of 'Data-Driven Quadratic Kernel Synthesis for Nonlinear Ultrasound Imaging'. Together they form a unique fingerprint.

Cite this