Article
Deep Learning Approaches to Automated Video Classification
of Upper Limb Tension Test
Wansuk Choi
1
and Seoyoon Heo
2,
*
Citation: Choi, W.; Heo, S. Deep
Learning Approaches to Automated
Video Classification of Upper Limb
Tension Test. Healthcare 2021, 9, 1579.
https://doi.org/10.3390/
healthcare9111579
Academic Editors: Keun Ho Ryu and
Nipon Theera-Umpon
Received: 25 September 2021
Accepted: 15 November 2021
Published: 18 November 2021
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2021 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
1
Department of Physical Therapy, International University of Korea, Jinju 52833, Korea; y3korea@gmail.com
2
Department of Occupational Therapy, School of Medical and Health Science, Kyungbok University,
Namyangju-si 12051, Korea
* Correspondence: syheo@kbu.ac.kr; Tel.: +82-31-539-5351
Abstract:
The purpose of this study was to classify ULTT videos through transfer learning with
pre-trained deep learning models and compare the performance of the models. We conducted
transfer learning by combining a pre-trained convolution neural network (CNN) model into a
Python-produced deep learning process. Videos were processed on YouTube and 103,116 frames
converted from video clips were analyzed. In the modeling implementation, the process of importing
the required modules, performing the necessary data preprocessing for training, defining the model,
compiling, model creation, and model fit were applied in sequence. Comparative models were Xcep-
tion, InceptionV3, DenseNet201, NASNetMobile, DenseNet121, VGG16, VGG19, and ResNet101, and
fine tuning was performed. They were trained in a high-performance computing environment, and
validation and loss were measured as comparative indicators of performance. Relatively low valida-
tion loss and high validation accuracy were obtained from Xception, InceptionV3, and DenseNet201
models, which is evaluated as an excellent model compared with other models. On the other hand,
from VGG16, VGG19, and ResNet101, relatively high validation loss and low validation accuracy
were obtained compared with other models. There was a narrow range of difference between the
validation accuracy and the validation loss of the Xception, InceptionV3, and DensNet201 models.
This study suggests that training applied with transfer learning can classify ULTT videos, and that
there is a difference in performance between models.
Keywords:
deep structured learning; supervised machine learning; automated feature extraction;
Brachial Plexus Tension Tests; rehabilitation medicine; human action recognition
1. Introduction
Whereas research into classifying videos using deep-learning approaches has been
inclined to be tentative in rehabilitation medicine fields, recent advances in technologies
have accelerated research into analyzing overwhelmed video data. Human action recog-
nition has been expected to achieve a more refined and more scientific educational effect
in the environment of the recent academic supply, which is described and consumed in
images or motion pictures. Video (including images or motion pictures) data are regarded
as a spatiotemporal generalization of image data from a traditional neural network’s point
of view [
1
], and all neural network structures for image classification have been naturally
extended and discussed to a three-dimensional version beyond two dimensions [
2
]. The
machine learning process is a given for deriving insights or making classifications and
predictions. It refers to the way the data fit into a mathematical model [
3
]. Particularly,
machine learning discovers patterns that do not involve human subjective judgment or
other possible biases from a large amount of data, having high predictive power [
4
]. Since
the introduction of a video classification method using a dimensional convolutional neural
network (CNN) [
5
,
6
], a 3D CNN has been applied to large-scale video classification. Inter-
estingly, however, the performance of the 3D CNN was only slightly better than that of the
CNN, which classified each frame of a video as a 2D convolution. As a result of this, it was
Healthcare 2021, 9, 1579. https://doi.org/10.3390/healthcare9111579 https://www.mdpi.com/journal/healthcare
