TITLE:
Tuberculosis Detection from Computed Tomography with Convolutional Neural Networks
AUTHORS:
Xudong Liu, Haoxiang Lei, Sicun Han
KEYWORDS:
Lungs, Tuberculosis Detection, Computed Tomography, Convolutional Neural Networks
JOURNAL NAME:
Advances in Computed Tomography,
Vol.8 No.4,
December
18,
2019
ABSTRACT: Convolutional neural network (CNN),
a class of deep neural networks (most commonly used in visual image analysis),
has become one of the most influential innovations in the field of computer
vision. In our research, we built a system which allows the computer to extract
the feature and recognize the image of human lungs and to automatically
conclude the health level of the lungs based on database. Here, we built a CNN
model to train the datasets. After the training, the system could do certain
preliminary analysis already. In addition, we used the fixed coordinate to
reduce the noise and combined the Canny algorithm and the Mask algorithm to
further improve the accuracy of the system. The final accuracy turned out to be
87.0%, which is convincing. Our system can contribute a lot to the efficiency
and accuracy of doctors’ analysis of the patients’ health level. In the future,
we will do more improvement to reduce noise and increase accuracy.