UNAS-Net: A deep convolutional neural network for predicting Covid-19 severity

Syarif, Abdusy; Azman, Novi; Repi, Viktor Vekky Ronald; Sinaga, Ernawati; Asvial, Muhamad

doi:10.1016/j.imu.2021.100842

Cited by 9 publications

(6 citation statements)

References 17 publications

(26 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…f t controls the information of memory cells from time t−1 to time t. It controls the input information to the memory cells at time t. o t controls the information of the memory cells at time t until the hidden state h t . Equations ( 17), (18), and (19), respectively show the formulation of f t , i t , and o t in the LSTM network.…”

Section: Pandemic Predictionmentioning

confidence: 99%

See 1 more Smart Citation

Enhancing CNN‐LSTM neural networks using jellyfish search algorithm for pandemic modeling

Feriz,

Jalali,

Forghani

2024

Concurrency and Computation

View full text Add to dashboard Cite

SummaryThis paper presents a comprehensive three‐step approach (CNN‐JSO‐LSTM) for predictive modeling using a pandemic such as COVID‐19 as a test case. Initially, a Convolutional Neural Network (CNN) is employed to extract crucial features pertinent to the pandemic. Subsequently, the Jellyfish Search Optimizer (JSO) algorithm is applied for feature selection, identifying the most relevant factors. These chosen features are then inputted into a Long Short‐Term Memory (LSTM) network, responsible for classifying samples into “healthy” and “diseased” categories. Our method enhances LSTM performance using the Jellyfish Search optimizer, resulting in exceptional prediction accuracy. Our experiments achieved remarkable metrics, with an accuracy of 95.32%, high sensitivity (94.87%), and precision (94.28%), surpassing alternative methods. In conclusion, our study presents a promising and highly accurate approach for pandemic prediction, harnessing deep learning and swarm intelligence techniques. These findings suggest a potential for more effective pandemic management and intervention strategies.

show abstract

Section: Pandemic Predictionmentioning

confidence: 99%

“…In this study, DNN‐based methods are proposed for the diagnosis of COVID‐19. In Reference 19, a deep learning architecture is proposed to predict the cases of COVID‐19 in India. In this research, combined CNN + LSTM have been used to predict COVID‐19.…”

Section: Related Workmentioning

confidence: 99%

Enhancing CNN‐LSTM neural networks using jellyfish search algorithm for pandemic modeling

Feriz,

Jalali,

Forghani

2024

Concurrency and Computation

View full text Add to dashboard Cite

show abstract

“…Mangalmurti and Wattanapongsakorn 35 described effective ML approaches for the automatic detection/classification of COVID‐19 and other lung diseases. Syarif et al 36 employed DL to detect COVID‐19. The model recommended using CXR.…”

Section: Preliminariesmentioning

confidence: 99%

IFCnCov: An IoT‐based smart diagnostic architecture for COVID‐19

et al. 2023

View full text Add to dashboard Cite

Performing a coronary disease diagnosis remotely is challenging now‐a‐days. COVID‐19 is a worldwide pandemic, and methods for detecting COVID‐19 are hampered by insufficient data and a lack of validation testing. Internet of Things (IoT) applications that rely on cloud computing (CC) are being studied in an effort to improve e‐Healthcare systems, even though CC presents substantial latency, bandwidth, energy consumption, security and privacy issues and so forth. The extension to CC, fog computing (FC), can overcome these said limitations. This study aims to diagnose COVID‐19 patients to fight the outbreak remotely. This study proposes IFCnCov, which enables remote users to diagnose COVID‐19 disease in real‐time using integrated IoT, FC, and CC principles, as well as ensemble learning (EL) and deep learning (DL). The proposed system is a two‐layered architecture, trained with DL approaches to two different datasets: a symptom‐based dataset and a chest x‐rays imaging dataset obtained from the Kaggle repository employing several evaluative measures. From various experiments, this proposed IFCnCov achieves comparatively enhanced accuracies of 97.71% and 98.64%, precision of 97.38% and 98.52%, sensitivity of 98.19% and 98.92%, specificity of 97.21% and 98.34%, and F1‐scores of 97.79% and 98.72% in the first and second stages respectively, which also outperforms some other considered state‐of‐the‐art works. Additionally, this work is validated in terms of several network parameters, including scalability, energy consumption, network utilization, jitter, processing time, throughput, and arbitration time.

show abstract

“…Jha et al, [12] reported 90% efficiency for COVID-19 CXR images segmentation in a datasets from Kaggle and GitHub [13,14] with the multilevel threshold value of 75, 145, 185, 195 and 230 respectively, and the algorithm used is Falcon speed based Adaptive opposition Slime mould Algorithm (FS-AOSMA) with the falcon speed of 0.05 to 1 m/sec. Abdusy Syarif et al, [15] applied Otsu thresholding to binarize 12,467 CXR images, and archived the classification of 97.36% accuracy and 95.24% sensitivity by using Universitas National Network (UNAS-Net) deep learning model.…”

Section: Introductionmentioning

confidence: 99%

Robust Segmentation of COVID-19 Chest X-Ray Images: Analysis of Variant k-Means Based Clustering Algorithms

Ooi Wei Herng,

Aimi Salihah Abdul Nasir,

Abdul Syafiq Abdull Sukor

2024

ARASET

View full text Add to dashboard Cite

Computer aided diagnosis (CADx) become one the most famous method in diagnostic medical field due to the high reliability and efficiency. Recently, the coronavirus disease (COVID-19) has become severe global pandemic. Particularly, the Chest X-ray (CXR) imaging has become an essentiality in COVID-19 detection. As a result, the convergence of CADx technology with Chest X-ray analysis has achieved great efficiency in COVID-19 diagnosis. Therefore, the research value of CADx in COVID-19 diagnosis is exceptionally high. This study aims to evaluate different k-means based clustering algorithms and identifying the one with the highest overall accuracy. First of all, 150 COVID-19 CXR open-source images are acquired from Kaggle and Github. All the images will be unified into a same image size with 1000*1000 pixels and quality during the image pre-processing. Next, the resized images are enhanced by the Modified Global Contrast Stretching (MGCS) enhancement method to increase the quality of images. Then, the traditional k-means, k-medians, k-medoids and fast k-means clustering methods have been implemented in the image segmentation. At the same time, five different numbers <2, 4, 6, 8, 10> of clusters also tested out in this study. Lastly, all the segmented is proceeded to the segmentation performance based on sensitivity, specificity, accuracy, precision, recall and F-score. The result proves that the k-medoids clustering algorithm with 2 clusters archived the best overall segmentation performance as it obtained the highest sensitivity, accuracy, recall and F-score with 66.14%, 87.98%, 0.6614 and 0.7327.

show abstract

UNAS-Net: A deep convolutional neural network for predicting Covid-19 severity

Cited by 9 publications

References 17 publications

Enhancing CNN‐LSTM neural networks using jellyfish search algorithm for pandemic modeling

Enhancing CNN‐LSTM neural networks using jellyfish search algorithm for pandemic modeling

IFCnCov: An IoT‐based smart diagnostic architecture for COVID‐19

Robust Segmentation of COVID-19 Chest X-Ray Images: Analysis of Variant k-Means Based Clustering Algorithms

Contact Info

Product

Resources

About