Nonalcoholic fatty liver disease diagnosis using convolutional and recurrent neural networks on radiofrequency signal spectrograms
He, Churan
This item is only available for download by members of the University of Illinois community. Students, faculty, and staff at the U of I may log in with your NetID and password to view the item. If you are trying to access an Illinois-restricted dissertation or thesis, you can request a copy through your library's Inter-Library Loan office or purchase a copy directly from ProQuest.
Permalink
https://hdl.handle.net/2142/113507
Description
Title
Nonalcoholic fatty liver disease diagnosis using convolutional and recurrent neural networks on radiofrequency signal spectrograms
Nonalcoholic fatty liver disease (NAFLD) is increasingly common around the world, and it is the
most common form of chronic liver disease in the United States. The gold standard of NAFLD
diagnosis, liver biopsy, is invasive and unideal for early-stage screening. Researchers used
deep learning methods of raw radiofrequency (RF) ultrasound signals to perform a non-invasive
assessment of patients' liver fat fractions. Han et al. proposed a method that used one-dimensional
convolutional neural networks (1D-CNN) to estimate the liver fat fraction and diagnose nonalcoholic
fatty liver disease (NAFLD) using time-domain RF signals. The neural network model detects
patterns in raw RF signals and finds subtle correlations between these patterns with the liver fat
fraction. In our research, we propose a new machine learning approach that can take advantage
of the characteristics of RF signals. Compared to Han et al.'s method that directly utilizes the time domain
RF signals, which contain the signal’s spectrum information along with time. With the
spectrograms, our model can analyze how frequency components evolve from the beginning to the
end of the signals using the recurrent neural network (RNN) and extract more information than the
1D-CNN model. The result demonstrates that CNN combined with RNN on the spectrogram we proposed and implemented is capable of achieving a higher Pearson correlation coefficient between
the estimated fat fraction and the magnetic resonance imaging derived proton density fat fraction
(MRI-PDFF) (r = 0.869) than the 1D-CNN model (r = 0.835).
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.
