An explainable artificial intelligence model for multiple lung diseases classification from chest X-ray images using fine-tuned transfer learning

Decision Analytics Journal, Journal Year: 2024, Volume and Issue: 12, P. 100499 - 100499

Published: July 2, 2024

Traditional deep learning models are often considered "black boxes" due to their lack of interpretability, which limits therapeutic use despite success in classification tasks. This study aims improve the interpretability diagnoses for COVID-19, pneumonia, and tuberculosis from X-ray images using an enhanced DenseNet201 model within a transfer framework. We incorporated Explainable Artificial Intelligence (XAI) techniques, including SHAP, LIME, Grad-CAM, Grad-CAM++, make model's decisions more understandable. To enhance image clarity detail, we applied preprocessing methods such as Denoising Autoencoder, Contrast Limited Adaptive Histogram Equalization (CLAHE), Gamma Correction. An ablation was conducted identify optimal parameters proposed approach. Our performance compared with other learning-based like EfficientNetB0, InceptionV3, LeNet evaluation metrics. The that included data augmentation techniques achieved best results, accuracy 99.20%, precision recall 99%. demonstrates critical role improving performance. SHAP LIME provided significant insights into decision-making process, while Grad-CAM Grad-CAM++ highlighted specific features regions influencing classifications. These transparency trust AI-assisted diagnoses. Finally, developed Android-based system most effective support medical specialists process.

Language: Английский

---

The Effect of Ultrasound Image Pre-Processing on Radiomics Feature Quality: A Study on Shoulder Ultrasound

Deleted Journal, Journal Year: 2025, Volume and Issue: unknown

Published: Feb. 6, 2025

Abstract Radiomics, the extraction of quantitative features from medical images, has shown great promise in enhancing diagnostic and prognostic models, particularly CT MRI. However, its application ultrasound (US) imaging, especially musculoskeletal (MSK) remains underexplored. The inherent variability ultrasound, influenced by operator dependency various imaging settings, presents significant challenges to reproducibility radiomic features. This study aims identify whether commonly used image pre-processing methods can increase radiomics features, increasing quality analysis. is performed with shoulder calcific tendinopathy as a case study. Ultrasound images 84 patients rotator cuff calcifications were retrospectively analysed. Three techniques—Histogram Equalization, Standard CLAHE, Advanced CLAHE—were applied adjust quality. Manual segmentation was performed, followed 849 these assessed using intraclass correlation coefficient (ICC), comparing results across within dataset. CLAHE method consistently yielded highest ICC values, indicating superior compared other methods. Wavelet-transformed GLCM GLRLM subgroups, demonstrated robust all techniques. Shape however, continued show lower reproducibility. significantly enhances calcifications. underscores importance achieving reliable analyses, operator-dependent modalities like ultrasound.

Language: Английский

---

Deep learning ensemble approach with explainable AI for lung and colon cancer classification using advanced hyperparameter tuning

BMC Medical Informatics and Decision Making, Journal Year: 2024, Volume and Issue: 24(1)

Published: Aug. 7, 2024

Lung and colon cancers are leading contributors to cancer-related fatalities globally, distinguished by unique histopathological traits discernible through medical imaging. Effective classification of these is critical for accurate diagnosis treatment. This study addresses challenges in the diagnostic imaging lung cancers, which among causes deaths worldwide. Recognizing limitations existing methods, often suffer from overfitting poor generalizability, our research introduces a novel deep learning framework that synergistically combines Xception MobileNet architectures. innovative ensemble model aims enhance feature extraction, improve robustness, reduce overfitting.Our methodology involves training hybrid on comprehensive dataset images, followed validation against balanced test set. The results demonstrate an impressive accuracy 99.44%, with perfect precision recall identifying certain cancerous non-cancerous tissues, marking significant improvement over traditional approach.The practical implications findings profound. By integrating Gradient-weighted Class Activation Mapping (Grad-CAM), offers enhanced interpretability, allowing clinicians visualize reasoning process. transparency vital clinical acceptance enables more personalized, treatment planning. Our not only pushes boundaries technology but also sets stage future aimed at expanding techniques other types cancer diagnostics.

Language: Английский

---

PediaPulmoDx: Harnessing Cutting Edge Preprocessing and Explainable AI for Pediatric Chest X-ray Classification with DenseNet121

Results in Engineering, Journal Year: 2025, Volume and Issue: unknown, P. 104320 - 104320

Published: Feb. 1, 2025

Language: Английский

---

Automated classification of chest X-rays: a deep learning approach with attention mechanisms

BMC Medical Imaging, Journal Year: 2025, Volume and Issue: 25(1)

Published: March 4, 2025

Pulmonary diseases such as COVID-19 and pneumonia, are life-threatening conditions, that require prompt accurate diagnosis for effective treatment. Chest X-ray (CXR) has become the most common alternative method detecting pulmonary COVID-19, lung opacity due to their availability, cost-effectiveness, ability facilitate comparative analysis. However, interpretation of CXRs is a challenging task. This study presents an automated deep learning (DL) model outperforms multiple state-of-the-art methods in diagnosing Lung Opacity, Viral Pneumonia. Using dataset 21,165 CXRs, proposed framework introduces seamless combination Vision Transformer (ViT) capturing long-range dependencies, DenseNet201 powerful feature extraction, global average pooling (GAP) retaining critical spatial details. results robust classification system, achieving remarkable accuracy. The methodology delivers outstanding across all categories: 99.4% accuracy F1-score 98.43% 96.45% 93.64% 99.63% 97.05% Pneumonia, 95.97% with 95.87% Normal subjects. achieves overall 97.87%, surpassing several reproducible objective outcomes. To ensure robustness minimize variability train-test splits, our employs five-fold cross-validation, providing reliable consistent performance evaluation. For transparency future comparisons, specific training testing splits have been made publicly accessible. Furthermore, Grad-CAM-based visualizations integrated enhance interpretability model, offering valuable insights into its decision-making process. innovative not only boosts but also sets new benchmark CXR-based disease diagnosis.

Language: Английский

---

Analysis of the Impact of Watermarking Technique in Neural Network Models to Predict Lung Diseases

Communications in computer and information science, Journal Year: 2025, Volume and Issue: unknown, P. 308 - 321

Published: Jan. 1, 2025

Language: Английский

---

PulmonNet V1: Leveraging the benefit of Leaky ReLU activation for the local and multi-scale global feature integration of chest radiographs to classify pulmonary diseases

Biomedical Signal Processing and Control, Journal Year: 2024, Volume and Issue: 96, P. 106600 - 106600

Published: July 2, 2024

Language: Английский

---

Residual learning for brain tumor segmentation: dual residual blocks approach

Neural Computing and Applications, Journal Year: 2024, Volume and Issue: unknown

Published: Sept. 30, 2024

Language: Английский

---

Identification of COVID-19 with CT scans using radiomics and DL-based features

Network Modeling Analysis in Health Informatics and Bioinformatics, Journal Year: 2024, Volume and Issue: 13(1)

Published: March 29, 2024

Language: Английский

---

Augmenting Radiological Diagnostics with AI for Tuberculosis and COVID-19 Disease Detection: Deep Learning Detection of Chest Radiographs

Diagnostics, Journal Year: 2024, Volume and Issue: 14(13), P. 1334 - 1334

Published: June 24, 2024

In this research, we introduce a network that can identify pneumonia, COVID-19, and tuberculosis using X-ray images of patients' chests. The study emphasizes tuberculosis, healthy lung conditions, discussing how advanced neural networks, like VGG16 ResNet50, improve the detection issues from images. To prepare for model's input requirements, enhanced them through data augmentation techniques training purposes. We evaluated performance by analyzing precision, recall, F1 scores across training, validation, testing datasets. results show ResNet50 model outperformed with accuracy resilience. It displayed superior ROC AUC values in both validation test scenarios. Particularly impressive were ResNet50's precision recall rates, nearing 0.99 all conditions set. On hand, also performed well during testing-detecting 0.93. Our highlights our deep learning method showcasing effectiveness over traditional approaches VGG16. This progress utilizes methods to enhance classification augmenting balancing them. positions approach as an advancement state-of-the-art applications imaging. By enhancing reliability diagnosing ailments such COVID-19 models have potential transform care treatment strategies, highlighting their role clinical diagnostics.

Language: Английский

---