Efficient pneumonia detection using Vision Transformers on chest X-rays

Scientific Reports, Journal Year: 2024, Volume and Issue: 14(1)

Published: Jan. 30, 2024

Abstract Pneumonia is a widespread and acute respiratory infection that impacts people of all ages. Early detection treatment pneumonia are essential for avoiding complications enhancing clinical results. We can reduce mortality, improve healthcare efficiency, contribute to the global battle against disease has plagued humanity centuries by devising deploying effective methods. Detecting not only medical necessity but also humanitarian imperative technological frontier. Chest X-rays frequently used imaging modality diagnosing pneumonia. This paper examines in detail cutting-edge method detecting implemented on Vision Transformer (ViT) architecture public dataset chest available Kaggle. To acquire context spatial relationships from X-ray images, proposed framework deploys ViT model, which integrates self-attention mechanisms transformer architecture. According our experimentation with Transformer-based framework, it achieves higher accuracy 97.61%, sensitivity 95%, specificity 98% X-rays. The model preferable capturing context, comprehending relationships, processing images have different resolutions. establishes its efficacy as robust solution surpassing convolutional neural network (CNN) based architectures.

Language: Английский

---

Brain Tumor MRI Classification Using a Novel Deep Residual and Regional CNN

Biomedicines, Journal Year: 2024, Volume and Issue: 12(7), P. 1395 - 1395

Published: June 23, 2024

Brain tumor classification is essential for clinical diagnosis and treatment planning. Deep learning models have shown great promise in this task, but they are often challenged by the complex diverse nature of brain tumors. To address challenge, we propose a novel deep residual region-based convolutional neural network (CNN) architecture, called Res-BRNet, using magnetic resonance imaging (MRI) scans. Res-BRNet employs systematic combination regional boundary-based operations within modified spatial blocks. The blocks extract homogeneity, heterogeneity, boundary-related features tumors, while significantly capture local global texture variations. We evaluated performance on challenging dataset collected from Kaggle repositories, Br35H, figshare, containing various categories, including meningioma, glioma, pituitary, healthy images. outperformed standard CNN models, achieving excellent accuracy (98.22%), sensitivity (0.9811), F1-score (0.9841), precision (0.9822). Our results suggest that promising tool classification, with potential to improve efficiency

Language: Английский

---

Lung Cancer Classification using Optimized Attention-based Convolutional Neural Network with DenseNet-201 Transfer Learning Model on CT image

Biomedical Signal Processing and Control, Journal Year: 2024, Volume and Issue: 95, P. 106330 - 106330

Published: April 25, 2024

Language: Английский

---

A new deep boosted CNN and ensemble learning based IoT malware detection

Computers & Security, Journal Year: 2023, Volume and Issue: 133, P. 103385 - 103385

Published: July 7, 2023

Security issues are threatened in various types of networks, especially the Internet Things (IoT) environment that requires early detection. IoT is network real-time devices like home automation systems and can be controlled by open-source android devices, which an open ground for attackers. Attackers access credentials, initiate a different kind security breach, compromises control. Therefore, timely detecting increasing number sophisticated malware attacks challenge to ensure credibility protection. In this regard, we have developed new detection framework, Deep Squeezed-Boosted Ensemble Learning (DSBEL), comprised novel Boundary-Region Split-Transform-Merge (SB-BR-STM) CNN ensemble learning. The proposed STM block employs multi-path dilated convolutional, Boundary, regional operations capture homogenous heterogeneous global malicious patterns. Moreover, diverse feature maps achieved using transfer learning multi-path-based squeezing boosting at initial final levels learn minute pattern variations. Finally, boosted discriminative features extracted from deep SB-BR-STM provided classifiers (SVM, MLP, AdabooSTM1) improve hybrid generalization. performance analysis DSBEL framework against existing techniques been evaluated IOT_Malware dataset on standard measures. Evaluation results show progressive as 98.50% accuracy, 97.12% F1-Score, 91.91% MCC, 95.97 % Recall, 98.42 Precision. robust helpful activity suggests future strategies.

Language: Английский

---

Improving prediction of cervical cancer using KNN imputer and multi-model ensemble learning

PLoS ONE, Journal Year: 2024, Volume and Issue: 19(1), P. e0295632 - e0295632

Published: Jan. 3, 2024

Cervical cancer is a leading cause of women's mortality, emphasizing the need for early diagnosis and effective treatment. In line with imperative intervention, automated identification cervical has emerged as promising avenue, leveraging machine learning techniques to enhance both speed accuracy diagnosis. However, an inherent challenge in development these systems presence missing values datasets commonly used detection. Missing data can significantly impact performance models, potentially inaccurate or unreliable results. This study addresses critical identification-handling datasets. The present novel approach that combines three models into stacked ensemble voting classifier, complemented by use KNN Imputer manage values. proposed model achieves remarkable results 0.9941, precision 0.98, recall 0.96, F1 score 0.97. examines distinct scenarios: one involving deletion values, another utilizing imputation, third employing PCA imputing research significant implications medical field, offering experts powerful tool more accurate therapy enhancing overall effectiveness testing procedures. By addressing challenges achieving high accuracy, this work represents valuable contribution detection, ultimately aiming reduce disease on health healthcare systems.

Language: Английский

---

COVID-19 detection from chest X-ray images using CLAHE-YCrCb, LBP, and machine learning algorithms

BMC Bioinformatics, Journal Year: 2024, Volume and Issue: 25(1)

Published: Jan. 17, 2024

Abstract Background COVID-19 is a disease that caused contagious respiratory ailment killed and infected hundreds of millions. It necessary to develop computer-based tool fast, precise, inexpensive detect efficiently. Recent studies revealed machine learning deep models accurately using chest X-ray (CXR) images. However, they exhibit notable limitations, such as large amount data train, larger feature vector sizes, enormous trainable parameters, expensive computational resources (GPUs), longer run-time. Results In this study, we proposed new approach address some the above-mentioned limitations. The model involves following steps: First, use contrast limited adaptive histogram equalization (CLAHE) enhance CXR resulting images are converted from CLAHE YCrCb color space. We estimate reflectance chrominance Illumination–Reflectance model. Finally, normalized local binary patterns generated (Cr) YCb classification vector. Decision tree, Naive Bayes, support machine, K-nearest neighbor, logistic regression were used algorithms. performance evaluation on test set indicates superior, with accuracy rates 99.01%, 100%, 98.46% across three different datasets, respectively. probabilistic algorithm, emerged most resilient. Conclusion Our method uses fewer handcrafted features, affordable resources, less runtime than existing state-of-the-art approaches. Emerging nations where radiologists in short supply can adopt prototype. made both coding materials datasets accessible general public for further improvement. Check manuscript’s availability under declaration section access.

Language: Английский

---

COVID-19 severity detection using chest X-ray segmentation and deep learning

Scientific Reports, Journal Year: 2024, Volume and Issue: 14(1)

Published: Aug. 27, 2024

COVID-19 has resulted in a significant global impact on health, the economy, education, and daily life. The disease can range from mild to severe, with individuals over 65 or those underlying medical conditions being more susceptible severe illness. Early testing isolation are vital due virus's variable incubation period. Chest radiographs (CXR) have gained importance as diagnostic tool their efficiency reduced radiation exposure compared CT scans. However, sensitivity of CXR detecting may be lower. This paper introduces deep learning framework for accurate classification severity prediction using images. U-Net is used lung segmentation, achieving precision 0.9924. Classification performed Convulation-capsule network, high true positive rates 86% COVID-19, 93% pneumonia, 85% normal cases. Severity assessment employs ResNet50, VGG-16, DenseNet201, DenseNet201 showing superior accuracy. Empirical results, validated 95% confidence intervals, confirm framework's reliability robustness. integration advanced techniques radiological imaging enhances early detection assessment, improving patient management resource allocation clinical settings.

Language: Английский

---

Utilization of Deep Convolutional Neural Networks for Accurate Chest X-Ray Diagnosis and Disease Detection

Interdisciplinary Sciences Computational Life Sciences, Journal Year: 2023, Volume and Issue: 15(3), P. 374 - 392

Published: March 26, 2023

Language: Английский

---

COVID-19 infection analysis framework using novel boosted CNNs and radiological images

Scientific Reports, Journal Year: 2023, Volume and Issue: 13(1)

Published: Dec. 9, 2023

Abstract COVID-19, a novel pathogen that emerged in late 2019, has the potential to cause pneumonia with unique variants upon infection. Hence, development of efficient diagnostic systems is crucial accurately identifying infected patients and effectively mitigating spread disease. However, system poses several challenges because limited availability labeled data, distortion, complexity image representation, as well variations contrast texture. Therefore, two-phase analysis framework been developed scrutinize subtle irregularities associated COVID-19 contamination. A new Convolutional Neural Network-based STM-BRNet developed, which integrates Split-Transform-Merge (STM) block Feature map enrichment (FME) techniques first phase. The STM captures boundary regional-specific features essential for detecting infectious CT slices. Additionally, by incorporating FME Transfer Learning (TL) concept into blocks, multiple enhanced channels are generated capture minute illumination texture specific COVID-19-infected images. residual multipath learning used improve capacity progressively increase feature representation boosting at high level through TL. In second phase analysis, scans processed using newly SA-CB-BRSeg segmentation CNN delineate infection method utilizes approach combines smooth heterogeneous processes both encoder decoder. These operations structured patterns, including region-homogenous, variation, border. By these techniques, demonstrates its ability analyze segment related data. Furthermore, model incorporates CB decoder, where additional combined TL enhance low regions. models achieve impressive results, an accuracy 98.01%, recall 98.12%, F-score 98.11%, Dice Similarity 96.396%, IOU 98.85%. proposed will alleviate workload radiologist's decision-making region evaluating severity stages

Language: Английский

---

Malaria parasitic detection using a new Deep Boosted and Ensemble Learning framework

Complex & Intelligent Systems, Journal Year: 2024, Volume and Issue: 10(4), P. 4835 - 4851

Published: April 9, 2024

Abstract Malaria is a potentially fatal plasmodium parasite injected by female anopheles mosquitoes that infect red blood cells and cause millions of lifelong disability worldwide yearly. However, specialists’ manual screening in clinical practice laborious prone to error. Therefore, novel Deep Boosted Ensemble Learning (DBEL) framework, comprising the stacking new Boosted-BR-STM convolutional neural networks (CNN) ensemble ML classifiers, developed screen malaria images. The proposed based on dilated-convolutional block-based Split Transform Merge (STM) feature-map Squeezing–Boosting (SB) ideas. Moreover, STM block uses regional boundary operations learn parasite’s homogeneity, heterogeneity, with patterns. Furthermore, diverse boosted channels are attained employing Transfer Learning-based SB blocks at abstract, medium, conclusion levels minute intensity texture variation parasitic pattern. Additionally, enhance learning capacity foster more representation features, boosting final stage achieved through TL utilizing multipath residual learning. DBEL framework implicates prominent provides generated discriminative features classifiers. improves discrimination ability generalization deep feature spaces customized CNNs fed into classifiers for comparative analysis. outperforms existing techniques NIH dataset enhanced using discrete wavelet transform enrich space. Accuracy (98.50%), Sensitivity (0.9920), F-score (0.9850), AUC (0.9960), which suggests it be utilized screening.

Language: Английский

---