search

francismontalbo / fused-densenet-tiny

A novel approach in training a DenseNet model for diagnosing COVID-19 Chest X-Rays.
https://francismontalbo.github.io
29 stars 5 forks source link
ai article artificial-intelligence chest-xray-images convolution-neural-network covid-19 dataset deep-learning densenet gpu image-classification imageclassification imageprocessing keras tensorflow tutorial

readme

Diagnosing Covid-19 Chest X-Rays with a Lightweight Truncated DenseNet with Partial Layer Freezing and Feature Fusion

Author: Francis Jesmar P. Montalbo

Affiliation: Batangas State University

Email: francismontalbo@ieee.org

Personal Webpage

PLEASE CONTANCT ME IF YOU ARE HAVING TROUBLE. I CAN OFFER ASSITANCE

:heavy_exclamation_mark:This GitHub repository serves as a support for a published article in Biomedical Signal Processing and Control (BSPC) ISSN: 1746-8094. https://www.sciencedirect.com/science/article/abs/pii/S1746809421001804

CITATION

Francis Jesmar P. Montalbo,

Diagnosing Covid-19 Chest X-Rays with a Lightweight Truncated DenseNet with Partial Layer Freezing and Feature Fusion,

Biomedical Signal Processing and Control, 2021, 102583, ISSN 1746-8094

https://doi.org/10.1016/j.bspc.2021.102583.

(https://www.sciencedirect.com/science/article/pii/S1746809421001804)

Abstract: Due to the unforeseen turn of events, our world has undergone another global pandemic from a highly contagious novel coronavirus named COVID-19. The novel virus inflames the lungs similarly to Pneumonia, making it challenging to diagnose. Currently, the common standard to diagnose the virus's presence from an individual is using a molecular real-time Reverse-Transcription Polymerase Chain Reaction (rRT-PCR) test from fluids acquired through nasal swabs. Such a test is difficult to acquire in most underdeveloped countries with a few experts that can perform the test. As a substitute, the widely available Chest X-Ray (CXR) became an alternative to rule out the virus. However, such a method does not come easy as the virus still possesses unknown characteristics that even experienced radiologists and other medical experts find difficult to diagnose through CXRs. Several studies have recently used computer-aided methods to automate and improve such diagnosis of CXRs through Artificial Intelligence (AI) based on computer vision and Deep Convolutional Neural Networks (DCNN), which some require heavy processing costs and other tedious methods to produce. Therefore, this work proposed the Fused-DenseNet-Tiny, a lightweight DCNN model based on a densely connected neural network (DenseNet) truncated and concatenated. The model trained to learn CXR features based on transfer learning, partial layer freezing, and feature fusion. Upon evaluation, the proposed model achieved a remarkable 97.99% accuracy, with only 1.2 million parameters and a shorter end-to-end structure. It has also shown better performance than some existing studies and other massive state-of-the-art models that diagnosed COVID-19 from CXRs.

CITATION

Francis Jesmar P. Montalbo,

Truncating a Densely Connected Convolutional Neural Network with Partial Layer Freezing and Feature Fusion for Diagnosing COVID-19 from Chest X-Rays,

MethodsX, 2021, ISSN 2215-0161

https://doi.org/10.1016/j.mex.2021.101408

(https://www.sciencedirect.com/science/article/pii/S2215016121002016)

Keywords: chest x-rays; computer-aided diagnosis; covid-19; deep learning; densely connected neural networks

This is also a support for another article in MethodsX submission with an ISSN of 2215-0161. This repository will be maintained and improved after all reviews are done for the use of the public. A video utorial will also be released soon for a faster and easier implementation

Graphical Abstract

Made with draw.io

Fused-DenseNet-Tiny

Dataset:

Curated Dataset for COVID-19 Posterior-Anterior Chest Radiography Images (X-Rays).

Citation

SAIT, UNAIS; k v, Gokul Lal; Prajapati, Sunny; Bhaumik, Rahul; Kumar, Tarun; S, Sanjana; Bhalla , Kriti (2020), “Curated Dataset for COVID-19 Posterior-Anterior Chest Radiography Images (X-Rays).”, Mendeley Data, V1, doi: 10.17632/9xkhgts2s6.1 http://dx.doi.org/10.17632/9xkhgts2s6.1

:heavy_exclamation_mark:For a faster method, you may download the already prepared dataset used in the given link below.

CLICK ME FOR THE PREPARED DATASET

How to use:

:heavy_exclamation_mark:If training the model, the dependencies included a tensorflow-gpu. You may change the tensorflow-gpu to tensorflow if no GPU is to be used. However, the results from the paper were produced using a GPU (GTX 1070)

Dependencies included in the requirements.txt:

:heavy_exclamation_mark: START HERE You may clone using git or download the repository and extract the files manually:

First (easier):

Second (difficult):

REMEMBER THIS IS A LONGER PROCESS (Second process) TO TEST AND SIMULATE THE MODEL

Performance Results

Model Accuracy Precision Recall F1-Score
Fused-DenseNet-Tiny 97.99% 98.38% 95.15% 95.26%

CITE ME

@article{MONTALBO2021102583, title = {Diagnosing Covid-19 Chest X-Rays with a Lightweight Truncated DenseNet with Partial Layer Freezing and Feature Fusion}, journal = {Biomedical Signal Processing and Control}, pages = {102583}, year = {2021}, issn = {1746-8094}, doi = {https://doi.org/10.1016/j.bspc.2021.102583}, url = {https://www.sciencedirect.com/science/article/pii/S1746809421001804}, author = {Francis Jesmar P. Montalbo}, keywords = {chest x-rays, computer-aided diagnosis, covid-19, deep learning, densely connected neural networks}

@article{MONTALBO2021101408, title = {Truncating a Densely Connected Convolutional Neural Network with Partial Layer Freezing and Feature Fusion for Diagnosing COVID-19 from Chest X-Rays}, journal = {MethodsX}, pages = {101408}, year = {2021}, issn = {2215-0161}, doi = {https://doi.org/10.1016/j.mex.2021.101408}, url = {https://www.sciencedirect.com/science/article/pii/S2215016121002016}, author = {Francis Jesmar P. Montalbo}, keywords = {Deep Convolutional Neural Networks, COVID-19, Feature Fusion, Medical Image Diagnosis, Image Classification}