Computer-based quantitative image texture analysis using multi-collinearity diagnosis in chest X-ray images DOI Creative Commons
Antonio Quintero-Rincón, Ricardo Di-Pasquale,

Karina Quintero-Rodríguez

et al.

PLoS ONE, Journal Year: 2025, Volume and Issue: 20(4), P. e0320706 - e0320706

Published: April 14, 2025

Despite tremendous efforts devoted to the area, image texture analysis is still an open research field. This paper presents algorithm and experimental results demonstrating feasibility of developing automated tools detect abnormal X-ray images based on tissue attenuation. Specifically, this work proposes using variability characterised by singular values conditional indices extracted from value decomposition (SVD) as features. In addition, introduces a “tuning weight" parameter consider attenuation in tissues affected pathologies. weight estimated coefficient variation minimum covariance determinant bandwidth yielded non-parametric distribution variance-decomposition proportions SVD. When multiplied two features (singular indices), single acts tuning weight, reducing misclassification improving classic performance metrics, such true positive rate, false negative predictive values, discovery area-under-curve, accuracy total cost. The proposed method implements ensemble bagged trees classification model classify chest COVID-19, viral pneumonia, lung opacity, or normal. It was tested challenging, imbalanced public dataset. show 88% without applying 99% with its application. outperforms state-of-the-art methods, attested all metrics.

Language: Английский

COVID-19 recognition from chest X-ray images by combining deep learning with transfer learning DOI Creative Commons
Changjiang Zhang,

Lu-Ting Ruan,

Li Ji

et al.

Digital Health, Journal Year: 2025, Volume and Issue: 11

Published: Jan. 1, 2025

Objective Based on the current research status, this paper proposes a deep learning model named Covid-DenseNet for COVID-19 detection from CXR (computed tomography) images, aiming to build with smaller computational complexity, stronger generalization ability, and excellent performance benchmark datasets other different sample distribution features sizes. Methods The proposed first extracts obtains of multiple scales input image through transfer learning, followed by assigning internal weights extracted attention mechanism enhance important suppress irrelevant features; finally, fuses these multi-scale fusion architecture we designed obtain richer semantic information improve modeling efficiency. Results We evaluated our compared it advanced models three publicly available chest radiology types, one which is baseline dataset, constructed Covid-DenseNet, recognition accuracy test set was 96.89%, respectively. With 98.02% 96.21% two datasets, performs better than models. In addition, further external sets, trained data sets balanced (experiment 1) unbalanced 2), identified same set, DenseNet121. in experiment 1 2 80% 77.5% respectively, 3.33% 4.17% higher that DenseNet121 set. On basis, also changed number samples 2, impact change training results showed when increased became more abundant, performed robust. Conclusion Compared models, has achieved effect quite good, good robustness, enrichment features, robustness improved, clinical practice ability.

Language: Английский

Citations

0
Radiologist-inspired Symmetric Local-Global Multi-Supervised Learning for early diagnosis of pneumoconiosis DOI
Jiarui Wang, Meiyue Song, Deng-Ping Fan

et al.

Expert Systems with Applications, Journal Year: 2025, Volume and Issue: unknown, P. 127173 - 127173

Published: March 1, 2025

Language: Английский

Citations

0
Computer-based quantitative image texture analysis using multi-collinearity diagnosis in chest X-ray images DOI Creative Commons
Antonio Quintero-Rincón, Ricardo Di-Pasquale,

Karina Quintero-Rodríguez

et al.

PLoS ONE, Journal Year: 2025, Volume and Issue: 20(4), P. e0320706 - e0320706

Published: April 14, 2025

Despite tremendous efforts devoted to the area, image texture analysis is still an open research field. This paper presents algorithm and experimental results demonstrating feasibility of developing automated tools detect abnormal X-ray images based on tissue attenuation. Specifically, this work proposes using variability characterised by singular values conditional indices extracted from value decomposition (SVD) as features. In addition, introduces a “tuning weight" parameter consider attenuation in tissues affected pathologies. weight estimated coefficient variation minimum covariance determinant bandwidth yielded non-parametric distribution variance-decomposition proportions SVD. When multiplied two features (singular indices), single acts tuning weight, reducing misclassification improving classic performance metrics, such true positive rate, false negative predictive values, discovery area-under-curve, accuracy total cost. The proposed method implements ensemble bagged trees classification model classify chest COVID-19, viral pneumonia, lung opacity, or normal. It was tested challenging, imbalanced public dataset. show 88% without applying 99% with its application. outperforms state-of-the-art methods, attested all metrics.

Language: Английский

Citations

0