Uncertainty-aware diabetic retinopathy detection using deep learning enhanced by Bayesian approaches

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: Jan. 8, 2025

Abstract Deep learning-based medical image analysis has shown strong potential in disease categorization, segmentation, detection, and even prediction. However, high-stakes complex domains like healthcare, the opaque nature of these models makes it challenging to trust predictions, particularly uncertain cases. This sort uncertainty can be crucial analysis; diabetic retinopathy is an example where slight errors without indication confidence have adverse impacts. Traditional deep learning rely on single-point limiting their ability provide measures essential for robust clinical decision-making. To solve this issue, Bayesian approximation approaches evolved are gaining market traction. In work, we implemented a transfer approach, building upon DenseNet-121 convolutional neural network detect retinopathy, followed by extensions trained model. techniques, including Monte Carlo Dropout, Mean Field Variational Inference, Deterministic were applied represent posterior predictive distribution, allowing us evaluate model predictions. Our experiments combined dataset (APTOS 2019 + DDR) with pre-processed images showed that Bayesian-augmented outperforms state-of-the-art test accuracy, achieving 97.68% Dropout model, 94.23% 91.44% We also measure how certain predictions are, using entropy standard deviation metric each approach. evaluated both AUC accuracy scores at multiple data retention levels. addition overall performance boosts, results highlight does not only improve classification detection but reveals beneficial insights about estimation help build more trustworthy decision-making solutions.

Language: Английский

---

VisionDeep-AI: Deep learning-based retinal blood vessels segmentation and multi-class classification framework for eye diagnosis

Biomedical Signal Processing and Control, Journal Year: 2024, Volume and Issue: 94, P. 106273 - 106273

Published: March 28, 2024

Language: Английский

---

Gradient-Weighted Class Activation Mapping Based Deep Transfer Learning For Glaucoma Disease Prediction

Research Square (Research Square), Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 2, 2025

Glaucoma is a progressive eye disease characterized by damage to optic nerve. Early detection and management are crucial preserving vision, making prediction of glaucoma risk. To improve accurate prediction, Gradient-weighted Class Activation Mapped Deep Transfer Learning (GWCAMDTL) model developed. The main aim the enhance accuracy while minimizing time consumption. Retinal fundus images collected from dataset for in image acquisition phase. transfer learning involves adapting pre-trained deep performing prediction. In proposed model, Multilayer Perceptron classifier used as analyzing given large number training images. Then, new constructed along with its Initially, layers usually frozen preserve learned features infected regions. Transferring information previously results mode tasks has potential significantly feature efficiency applying congruence correlation coefficient. Mapping generates visual explanations predictions made model. Fine-tuning part learning. During fine-tuning weights certain updated better fit specific characteristics dataset, leading reduction both validation error. This approach improves strengths it clinical retinal process helps make extensively F1-score. Experimental conducted using various evaluation metrics. Results GWCAMDTL achieve higher reduced well error compared existing methods.

Language: Английский

---

Grad-CAM based explanations for multiocular disease detection using Xception net

Image and Vision Computing, Journal Year: 2025, Volume and Issue: unknown, P. 105419 - 105419

Published: Jan. 1, 2025

Language: Английский

---

Cnn-trans model: A parallel dual-branch network for fundus image classification

Biomedical Signal Processing and Control, Journal Year: 2024, Volume and Issue: 96, P. 106621 - 106621

Published: July 13, 2024

The existence of fundus diseases not only endangers people's vision, but also brings serious economic burden to the society. Fundus images are an objective and standard basis for diagnosis diseases. With continuous advancement computer science, deep learning methods dominated by convolutional neural networks (CNN) have been widely used in image classification. However, current CNN-based classification research still has a lot room improvement: CNN cannot effectively avoid interference repeated background information limited ability model whole world. In response above findings, this paper proposes CNN-Trans model. is parallel dual-branch network, which two branches CNN-LSTM Vision Transform (ViT). branch uses Xception after transfer learning. As original feature extractor, LSTM responsible dealing with gradient disappearance problem network iterations before head, then introduces new type lightweight attention mechanism between LSTM: Coordinate Attention, so as emphasize key related suppress less useful information; while self-attention ViT local interactions, it can establish long-distance dependence on target extract global features. Finally, concatenation (Concat) operation fuse features branches. extracted form complementary advantages. After fusion, more comprehensive sent layer. large number experimental tests comparisons, results show that: achieved accuracy 80.68% task, that comparable state-of-the-art methods. performance..

Language: Английский

---

A Novel Approach Analysis of Heart and Eye Disease Coherence Detection using Deep Learning Techniques

Journal of Machine and Computing, Journal Year: 2025, Volume and Issue: unknown, P. 015 - 029

Published: Jan. 3, 2025

One of the major factors contributing to rising death rate is cardiovascular disease. Analyzing clinical data has made it harder predict To solve aforementioned problems, an improved DenseNet model presented in this study. The proposed approach forecasts Central Retinal Artery Occlusion (CRAO) and Coronary Disease (CAD) simultaneously by using patient's from eye cardiac examinations. Then, coherence relationship calculated with help Pearson’s correlation coefficient for both diseases. As far as we are aware, first study use DL techniques between CRAO CAD. While predicting CAD, Improved 97.5% accuracy when compared benchmarked models like ResNet 50 VGG16.

Language: Английский

---

A novel eye disease segmentation and classification model using advanced deep learning network

Biomedical Signal Processing and Control, Journal Year: 2025, Volume and Issue: 105, P. 107565 - 107565

Published: Feb. 10, 2025

Language: Английский

---

Diagnose eyes diseases using deep learning algorithms

AIP conference proceedings, Journal Year: 2025, Volume and Issue: 3264, P. 040004 - 040004

Published: Jan. 1, 2025

---

Comparison of Classical and Deep Learning-Based Feature Representations for Age-Related Macular Degeneration

Lecture notes in electrical engineering, Journal Year: 2025, Volume and Issue: unknown, P. 505 - 514

Published: Jan. 1, 2025

Language: Английский

---

Artificial Intelligence-Driven Eye Disease Classification Model

Applied Sciences, Journal Year: 2023, Volume and Issue: 13(20), P. 11437 - 11437

Published: Oct. 18, 2023

Eye diseases can result in various challenges and visual impairments. These affect an individual’s quality of life general health well-being. The symptoms eye vary widely depending on the nature severity disease. Early diagnosis protect individuals from impairment. Artificial intelligence (AI)-based disease classification (EDC) assists physicians providing effective patient services. However, complexities fundus image classifier’s performance. There is a demand for practical EDC identifying earlier stages. Thus, author intends to build model using deep learning (DL) technique. Denoising autoencoders are used remove noises artifacts images. single-shot detection (SSD) approach generates key features. whale optimization algorithm (WOA) with Levy Flight Wavelet search strategy followed selecting In addition, Adam optimizer (AO) applied fine-tune ShuffleNet V2 classify Two benchmark datasets, ocular intelligent recognition (ODIR) utilized performance evaluation. proposed achieved accuracy Kappa values 99.1 96.4, 99.4 96.5, ODIR respectively. It outperformed recent models. findings highlight significance classifying complex Healthcare centers implement improve their standards serve more significant number patients. future, be extended identify comprehensive range diseases.

Language: Английский

---