Deep features to detect pulmonary abnormalities in chest X-rays due to infectious diseaseX: Covid-19, pneumonia, and tuberculosis

Information Sciences, Journal Year: 2022, Volume and Issue: 592, P. 389 - 401

Published: Feb. 4, 2022

Language: Английский

---

Incremental learning-based cascaded model for detection and localization of tuberculosis from chest x-ray images

Expert Systems with Applications, Journal Year: 2023, Volume and Issue: 238, P. 122129 - 122129

Published: Oct. 14, 2023

Language: Английский

---

COVID-Net CXR-2: An Enhanced Deep Convolutional Neural Network Design for Detection of COVID-19 Cases From Chest X-ray Images

Frontiers in Medicine, Journal Year: 2022, Volume and Issue: 9

Published: June 10, 2022

As the COVID-19 pandemic devastates globally, use of chest X-ray (CXR) imaging as a complimentary screening strategy to RT-PCR testing continues grow given its routine clinical for respiratory complaint. part COVID-Net open source initiative, we introduce CXR-2, an enhanced deep convolutional neural network design detection from CXR images built using greater quantity and diversity patients than original COVID-Net. We also new benchmark dataset composed 19,203 multinational cohort 16,656 at least 51 countries, making it largest, most diverse in access form. The CXR-2 achieves sensitivity positive predictive value 95.5 97.0%, respectively, was audited transparent responsible manner. Explainability-driven performance validation used during auditing gain deeper insights decision-making behavior ensure clinically relevant factors are leveraged improving trust usage. Radiologist conducted, where select cases were reviewed reported on by two board-certified radiologists with over 10 19 years experience, showed that critical consistent radiologist interpretations.

Language: Английский

---

SecureFed: federated learning empowered medical imaging technique to analyze lung abnormalities in chest X-rays

International Journal of Machine Learning and Cybernetics, Journal Year: 2023, Volume and Issue: 14(8), P. 2659 - 2670

Published: Feb. 14, 2023

Language: Английский

---

Big Data COVID-19 Systematic Literature Review: Pandemic Crisis

Electronics, Journal Year: 2021, Volume and Issue: 10(24), P. 3125 - 3125

Published: Dec. 16, 2021

The COVID-19 pandemic has frightened people worldwide, and coronavirus become the most commonly used phrase in recent years. Therefore, there is a need for systematic literature review (SLR) related to Big Data applications crisis. objective highlight technological advancements. Many studies emphasize area of Our study categorizes many manage control pandemic. There very limited SLR prospective with Data. picked five databases: Science direct, IEEE Xplore, Springer, ACM, MDPI. Before screening, following recommendation, Preferred Reporting Items Systematic Reviews Meta Analyses (PRISMA) were reported 893 from 2019, 2020 until September 2021. After 60 met inclusion criteria through data statistics, analysis was as search string. research’s findings successfully dealt healthcare risk diagnosis, estimation or prevention, decision making, drug problems. We believe that this will motivate research community perform expandable transparent against crisis COVID-19.

Language: Английский

---

TOPSIS aided ensemble of CNN models for screening COVID-19 in chest X-ray images

Scientific Reports, Journal Year: 2022, Volume and Issue: 12(1)

Published: Sept. 14, 2022

Abstract The novel coronavirus (COVID-19), has undoubtedly imprinted our lives with its deadly impact. Early testing isolation of the individual is best possible way to curb spread this virus. Computer aided diagnosis (CAD) provides an alternative and cheap option for screening said In paper, we propose a convolution neural network (CNN)-based CAD method COVID-19 pneumonia detection from chest X-ray images. We consider three input types identical base classifiers. To capture maximum complementary features, original RGB image, Red channel image stacked Robert's edge information. After that develop ensemble strategy based on technique order preference by similarity ideal solution (TOPSIS) aggregate outcomes overall framework, called TOPCONet, very light in comparison standard CNN models terms number trainable parameters required. TOPCONet achieves state-of-the-art results when evaluated publicly available datasets: (1) IEEE dataset + Kaggle Pneumonia Dataset, (2) Radiography (3) COVIDx.

Language: Английский

---

Domain Shifts in Machine Learning Based Covid-19 Diagnosis From Blood Tests

Journal of Medical Systems, Journal Year: 2022, Volume and Issue: 46(5)

Published: March 29, 2022

Many previous studies claim to have developed machine learning models that diagnose COVID-19 from blood tests. However, we hypothesize changes in the underlying distribution of data, so called domain shifts, affect predictive performance and reliability are a reason for failure such clinical application. Domain shifts can be caused, e.g., by disease prevalence (spreading or tested population), refined RT-PCR testing procedures (way taking samples, laboratory procedures), virus mutations. Therefore, diagnosing other diseases may not reliable degrade over time. We investigate whether present datasets how they methods. further set out estimate mortality risk based on routinely acquired tests hospital setting throughout pandemics under shifts. reveal evaluating large-scale dataset with different assessment strategies, as temporal validation. novel finding strongly diagnosis deteriorate their credibility. frequent re-training re-assessment indispensable robust enabling utility.

Language: Английский

---

CovidConvLSTM: A fuzzy ensemble model for COVID-19 detection from chest X-rays

Expert Systems with Applications, Journal Year: 2022, Volume and Issue: 206, P. 117812 - 117812

Published: June 16, 2022

Language: Английский

---

Detection of COVID-19, pneumonia, and tuberculosis from radiographs using AI-driven knowledge distillation

Heliyon, Journal Year: 2024, Volume and Issue: 10(5), P. e26801 - e26801

Published: Feb. 28, 2024

Chest radiography is an essential diagnostic tool for respiratory diseases such as COVID-19, pneumonia, and tuberculosis because it accurately depicts the structures of chest. However, accurate detection these from radiographs a complex task that requires availability medical imaging equipment trained personnel. Conventional deep learning models offer viable automated solution this task. high complexity often poses significant obstacle to their practical deployment within applications, including mobile apps, web cloud-based platforms. This study addresses resolves dilemma by reducing neural networks using knowledge distillation techniques (KDT). The proposed technique trains network on extensive collection chest X-ray images propagates smaller capable real-time detection. To create comprehensive dataset, we have integrated three popular radiograph datasets with tuberculosis. Our experiments show approach outperforms conventional methods in terms computational performance disease Specifically, our system achieves impressive average accuracy 0.97, precision 0.94, recall 0.97.

Language: Английский

---

Neural architecture search for pneumonia diagnosis from chest X-rays

Scientific Reports, Journal Year: 2022, Volume and Issue: 12(1)

Published: July 4, 2022

Pneumonia is one of the diseases that causes most fatalities worldwide, especially in children. Recently, pneumonia-caused deaths have increased dramatically due to novel Coronavirus global pandemic. Chest X-ray (CXR) images are readily available and common imaging modality for detection identification pneumonia. However, pneumonia from chest radiography a difficult task even experienced radiologists. Artificial Intelligence (AI) based systems great potential assisting quick accurate diagnosis X-rays. The aim this study develop Neural Architecture Search (NAS) method find best convolutional architecture capable detecting We propose Learning by Teaching framework inspired teaching-driven learning methodology humans, conduct experiments on dataset with over 5000 images. Our proposed yields an area under ROC curve (AUC) 97.6% detection, which improves upon previous NAS methods 5.1% (absolute).

Language: Английский

---