Abstract:
Physiotherapists face difficulties in
detecting the exact problem of their patients. De
tecting the exact problem will make it easier for
physiotherapists to proceed with their treatment.
This study proposes the creation of a new Android
application that incorporates an AI diagnostic ca
pability into physiotherapy practice. The system
is based on the MobileNetV1 convolutional neural
network (CNN) architecture to detect bone fractures
and chest secretions (e.g., pneumonia) from X-ray
images. Through a REST-API backend architecture,
real-time predictions are produced using pre-trained
TensorFlow deep learning models. For fractures,
the model achieved an accuracy score of 96%, while
the chest secretion detection model attained 94%
accuracy on publicly available datasets. The system
provides physiotherapy practitioners with actionable
diagnostic clues by establishing a translation path
way from AI to physiotherapy research, especially
in low-resource settings.
