COMPARISON OF DEEP NEURAL NETWORK ARCHITECTURES FOR ULTRASOUND FAT-FRACTION ESTIMATION IN NON-ALCOHOLIC FATTY LIVER DISEASE
Cai, William
Loading…
Permalink
https://hdl.handle.net/2142/114978
Description
Title
COMPARISON OF DEEP NEURAL NETWORK ARCHITECTURES FOR ULTRASOUND FAT-FRACTION ESTIMATION IN NON-ALCOHOLIC FATTY LIVER DISEASE
Author(s)
Cai, William
Issue Date
2022-05
Keyword(s)
deep learning; transformer; ultrasound; nonalcoholic fatty liver disease
Abstract
Nonalcoholic fatty liver disease (NAFLD) is the most common form of chronic liver disease,
affecting approximately one quarter of the United States’ population. With traditional methods
of diagnosis such as biopsies being both expensive and invasive, ultrasound imaging for NAFLD
is becoming increasingly popular. Deep learning models based on ultrasound radiofrequency
(RF) signals have shown promising results in medical applications such as non-invasively
quantifying the amount of fat in a human liver. A one-dimensional convolutional neural
network (1D-CNN) was recently proposed to estimate the liver fat content from ultrasound RF
signals. In this thesis, we compare the performance between the 1D-CNN and the popular
transformer architecture for this task. The transformer model is often used in the fields of
natural language processing and computer vision due to its self-attention mechanism, and it
outperforms many other models. The comparison between the 1D-CNN and transformer will
provide insights into architecture selection for deep learning using ultrasound RF signals.
Use this login method if you
don't
have an
@illinois.edu
email address.
(Oops, I do have one)
IDEALS migrated to a new platform on June 23, 2022. If you created
your account prior to this date, you will have to reset your password
using the forgot-password link below.