Advancing Acoustic Droplet Vaporization for Tissue Characterization Using Quantitative Ultrasound and Transfer Learning

Acoustic droplet vaporization (ADV) is an emerging technique with expanding applications in biomedical ultrasound. ADV-generated bubbles can function as microscale probes that provide insights into the mechanical properties of their surrounding microenvironment. This study investigated the acoustic...

Full description

Saved in:
Bibliographic Details
Published in:IEEE transactions on biomedical engineering Vol. 72; no. 6; pp. 1897 - 1908
Main Authors: Kaushik, Anuj, Fabiilli, Mario L., Myers, Daniel D., Fowlkes, J. Brian, Aliabouzar, Mitra
Format: Journal Article
Language:English
Published: United States IEEE 01.06.2025
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Subjects:
Acoustic droplet vaporization
Acoustic responses
Acoustics
Artificial neural networks
Biological tissues
Biomedical engineering
Biomedical imaging
Bubbles
Cavitation
Deep Learning
Density
Droplets
Fibrin
Fibrin - chemistry
Homogeneity
Humans
Hydrogels
Hydrogels - chemistry
Hydrophones
Image Processing, Computer-Assisted - methods
Imaging techniques
Lipidomics
Liquids
Mean
Mechanical properties
Microenvironments
Neural networks
Phantoms, Imaging
phase-shift droplets
Probes
Radiology
Texture
tissue characterization
Tissue engineering
Transfer learning
Ultrasonic imaging
Ultrasonography - methods
Ultrasound
Vaporization
Volatilization
ISSN:0018-9294, 1558-2531, 1558-2531
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Acoustic droplet vaporization (ADV) is an emerging technique with expanding applications in biomedical ultrasound. ADV-generated bubbles can function as microscale probes that provide insights into the mechanical properties of their surrounding microenvironment. This study investigated the acoustic and imaging characteristics of phase-shift nanodroplets in fibrin-based, tissue-mimicking hydrogels using passive cavitation detection and active imaging techniques, including B-mode and contrast-enhanced ultrasound. The findings demonstrated that the backscattered signal intensities and pronounced nonlinear acoustic responses, including subharmonic and higher harmonic frequencies, of ADV-generated bubbles correlated inversely with fibrin density. Additionally, we quantified the mean echo intensity, bubble cloud area, and second-order texture features of the generated ADV bubbles across varying fibrin densities. ADV bubbles in softer hydrogels displayed significantly higher mean echo intensities, larger bubble cloud areas, and more heterogeneous textures. In contrast, texture uniformity, characterized by variance, homogeneity, and energy, correlated directly with fibrin density. Furthermore, we incorporated transfer learning with convolutional neural networks, adapting AlexNet into two specialized models for differentiating fibrin hydrogels. The integration of deep learning techniques with ADV offers great potential, paving the way for future advancements in biomedical diagnostics.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
content type line 23
ISSN:0018-9294
1558-2531
1558-2531
DOI:10.1109/TBME.2025.3527141