Published: Oct. 18, 2024
Language: Английский
Technologies, Journal Year: 2025, Volume and Issue: 13(5), P. 170 - 170
Published: April 23, 2025
Hyperspectral imaging (HSI) is an advanced technique that captures detailed spectral information across multiple fields. This review explores its applications in counterfeit detection, remote sensing, agriculture, medical imaging, cancer environmental monitoring, mining, mineralogy, and food processing, specifically highlighting significant achievements from the past five years, providing a timely update several It also presents cross-disciplinary classification framework to systematically categorize medical, environment, industry. In HSI identified fake currency with high accuracy 400–500 nm range achieved 99.03% F1-score for alcohol detection. Remote sensing include hyperspectral satellites, which improve forest by 50%, soil organic matter, prediction reaching R2 = 0.6. HSI-TransUNet model 86.05% crop classification, disease detection reached 98.09% accuracy. Medical benefits HSI’s non-invasive diagnostics, distinguishing skin 87% sensitivity 88% specificity. colorectal identification 86% 95% Environmental PM2.5 pollution 85.93% marine plastic waste 70–80% egg freshness 91%, pine nut 100% Despite advantages, faces challenges like costs complex data processing. Advances artificial intelligence miniaturization are expected accessibility real-time applications. Future advancements anticipated concentrate on integration of deep learning models automated feature extraction decision-making analysis. The development lightweight, portable devices will enable more on-site healthcare, monitoring. Moreover, processing methods enhance efficiency field deployment. These improvements seek accessibility, practicality, efficacy both industrial clinical environments.
Language: Английский
Citations
0
Cancers, Journal Year: 2025, Volume and Issue: 17(9), P. 1521 - 1521
Published: April 30, 2025
Cervical lesion classification is essential for early detection of cervical cancer. While deep learning methods have shown promise, most rely on single-modal data or require extensive manual annotations. This study proposes a novel Graph Neural Network (GNN)-based framework that integrates colposcopy images, segmentation masks, and graph representations improved classification. We developed fully connected graph-based architecture using GCNConv layers with global mean pooling optimized it via grid search. A five-fold cross-validation protocol was employed to evaluate performance before (1-100 epochs) after fine-tuning (101-151 epochs). Performance metrics included macro-average F1-score validation accuracy. Visualizations were used model interpretability. The achieved 89.4% accuracy 92.1% fine-tuning, which 94.56% 98.98%, respectively, fine-tuning. LIME-based visual explanations validated models focus discriminative regions. highlights the potential multi-modal analysis. Collaborating MNJ Institute Oncology, shows promise clinical use.
Language: Английский
Citations
0
World Journal of Gastrointestinal Oncology, Journal Year: 2025, Volume and Issue: 17(5)
Published: May 15, 2025
Esophageal cancer (EC), a common malignant tumor of the digestive tract, requires early diagnosis and timely treatment to improve patient prognosis. Automated detection EC using medical imaging has potential increase screening efficiency diagnostic accuracy, thereby significantly improving long-term survival rates quality life patients. Recent advances in deep learning (DL), particularly convolutional neural networks, have demonstrated remarkable performance analysis. These techniques shown significant progress automated identification tumors, quantitative analysis lesions, improvement accuracy efficiency. This article comprehensively examines research DL for EC, covering various modalities such as digital pathology, endoscopy, computed tomography, etc. It explores clinical value application prospects diagnosis. Additionally, addresses several critical challenges that must be overcome translation techniques, including constructing high-quality datasets, promoting multimodal feature fusion, optimizing artificial intelligence-clinical workflow integration. By providing detailed overview current state highlighting key future directions, this aims guide facilitate implementation technologies management, ultimately contributing better outcomes.
Language: Английский
Citations
0
Array, Journal Year: 2025, Volume and Issue: unknown, P. 100413 - 100413
Published: May 1, 2025
Language: Английский
Citations
0
Machine Learning with Applications, Journal Year: 2024, Volume and Issue: 18, P. 100605 - 100605
Published: Nov. 10, 2024
Language: Английский
Citations
3
BMC Medical Education, Journal Year: 2024, Volume and Issue: 24(1)
Published: Dec. 28, 2024
Language: Английский
Citations
3
Image and Vision Computing, Journal Year: 2024, Volume and Issue: 154, P. 105347 - 105347
Published: Nov. 30, 2024
Language: Английский
Citations
2
Cognitive Computation, Journal Year: 2024, Volume and Issue: 16(5), P. 2131 - 2153
Published: June 13, 2024
Language: Английский
Citations
1
IEEE Access, Journal Year: 2024, Volume and Issue: 12, P. 178401 - 178440
Published: Jan. 1, 2024
The emergence of generative adversarial networks (GANs) has ignited substantial interest in the domain synthesizing images from textual descriptions. This approach demonstrated remarkable versatility and user-friendliness producing conditioned images, showcasing notable progress areas like diversity, visual realism, semantic alignment recent years. Notwithstanding these developments, discipline still faces difficulties, such as high-resolution pictures with several objects developing trustworthy evaluation standards that are line human vision. goal this study is to provide a comprehensive overview state stochastic text-to-image creation models right now. It examines how they have changed over previous five years suggests classification system depending on degree supervision required. paper highlights shortcomings, provides critical current approaches for assessing models, further areas. These goals include improving training designs architecture, more reliable assessment criteria, fine-tuning datasets. review, which focuses synthesizing, useful addition earlier surveys offers guidance future studies subject.
Language: Английский
Citations
0
Cancers, Journal Year: 2024, Volume and Issue: 17(1), P. 29 - 29
Published: Dec. 25, 2024
Background/Objectives: Developing high-performance artificial intelligence (AI) models for rare diseases is challenging owing to limited data availability. This study aimed evaluate whether a novel three-class annotation method preparing training could enhance AI model performance in detecting osteosarcoma on plain radiographs compared conventional single-class annotation. Methods: We developed two methods the same dataset of 468 X-rays and 378 normal radiographs: (1C model) (3C that separately labeled intramedullary, cortical, extramedullary tumor components. Both used identical U-Net-based architectures, differing only their approaches. Performance was evaluated using an independent validation dataset. Results: Although both achieved high diagnostic accuracy (AUC: 0.99 vs. 0.98), 3C demonstrated superior operational characteristics. At standardized cutoff value 0.2, maintained balanced (sensitivity: 93.28%, specificity: 92.21%), whereas 1C showed compromised specificity (83.58%) despite sensitivity (98.88%). Notably, at 25th percentile threshold, false-negative rates significantly different values (3C: 0.661 1C: 0.985), indicating ability maintain substantially lower thresholds. Conclusions: anatomically informed can disease detection without requiring additional data. The improved stability thresholds suggests thoughtful strategies optimize training, particularly contexts where are limited.
Language: Английский
Citations
0