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

Scientific Reports, Год журнала: 2025, Номер 15(1)

Опубликована: Янв. 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.

Язык: Английский

---

Advances in Deep Learning for Medical Image Analysis: A Comprehensive Investigation

Journal of Statistical Theory and Practice, Год журнала: 2025, Номер 19(1)

Опубликована: Янв. 23, 2025

Язык: Английский

---

AI-Based Financial Transaction Monitoring and Fraud Prevention with Behaviour Prediction

Опубликована: Июль 16, 2024

In this study, we explored the application of deep learning techniques for credit card fraud detection, aiming to improve performance and reliability anomaly detection methods in financial transactions. We first utilized Isolation Forest algorithm, achieving a accuracy 26% top 1000 Subsequently, experimented with Autoencoder an unsupervised neural network model, which enhanced 33.6% best case despite some fluctuations. The results demonstrate models' strong feature extraction capability adaptability, highlighting their potential surpass traditional methods. However, high imbalance dataset, only 0.17% transactions being fraudulent, poses significant challenge. This study underscores necessity further experimentation optimization structures hyperparameters achieve more stable efficient detection. findings provide valuable insights reference points future research using methodologies.

Язык: Английский

---

LIU-NET: lightweight Inception U-Net for efficient brain tumor segmentation from multimodal 3D MRI images

PeerJ Computer Science, Год журнала: 2025, Номер 11, С. e2787 - e2787

Опубликована: Март 31, 2025

Segmenting brain tumors is a critical task in medical imaging that relies on advanced deep-learning methods. However, effectively handling complex tumor regions requires more comprehensive and strategies to overcome challenges such as computational complexity, the gradient vanishing problem, variations size visual impact. To these challenges, this research presents novel computationally efficient method termed lightweight Inception U-Net (LIU-Net) for accurate segmentation task. LIU-Net balances model complexity load provide consistent performance uses blocks capture features at different scales, which makes it relatively lightweight. Its capability efficiently precisely segment tumors, especially challenging-to-detect regions, distinguishes from existing models. This Inception-style convolutional block assists capturing multiscale while preserving spatial information. Moreover, proposed utilizes combination of Dice loss Focal handle class imbalance issue. The was evaluated benchmark BraTS 2021 dataset, where generates remarkable outcomes with score 0.8121 enhancing (ET) region, 0.8856 whole (WT) 0.8444 core (TC) region test set. evaluate robustness architecture, cross-validated an external cohort 2020 dataset. obtained 0.8646 ET 0.9027 WT 0.9092 TC These results highlight effectiveness integrating into making promising candidate image segmentation.

Язык: Английский

---

Enhancing Energy Efficiency in Green Buildings Through Artificial Intelligence

Опубликована: Авг. 20, 2024

Artificial Intelligence (AI) is poised to revolutionize the architectural design and energy management of green buildings, offering significant advancements in sustainability efficiency. This paper explores transformative impact AI on improving efficiency reducing carbon emissions commercial buildings. By leveraging algorithms, architects can optimize building performance through advanced environmental analysis, automation repetitive tasks, real-time data-driven decision-making. facilitates precise consumption forecasting integration renewable sources, enhancing overall Our study demonstrates that reduce CO2 by approximately 8% 19%, respectively, typical mid-size office buildings 2050 compared conventional methods. Further, combination with policies low-emission production projected yield reductions up 40% 90% emissions. provides a systematic approach for quantifying AI's benefits across various types climate zones, valuable insights decision-makers construction industry.

Язык: Английский

---

Enhancing skin lesion classification: a CNN approach with human baseline comparison

PeerJ Computer Science, Год журнала: 2025, Номер 11, С. e2795 - e2795

Опубликована: Апрель 15, 2025

This study presents an augmented hybrid approach for improving the diagnosis of malignant skin lesions by combining convolutional neural network (CNN) predictions with selective human interventions based on prediction confidence. The algorithm retains high-confidence CNN while replacing low-confidence outputs expert assessments to enhance diagnostic accuracy. A model utilizing EfficientNetB3 backbone is trained datasets from ISIC-2019 and ISIC-2020 SIIM-ISIC melanoma classification challenges evaluated a 150-image test set. model’s are compared against 69 experienced medical professionals. Performance assessed using receiver operating characteristic (ROC) curves area under curve (AUC) metrics, alongside analysis resource costs. baseline achieves AUC 0.822, slightly below performance experts. However, improves true positive rate 0.782 reduces false 0.182, delivering better minimal involvement. offers scalable, resource-efficient solution address variability in image analysis, effectively harnessing complementary strengths humans CNNs.

Язык: Английский

---

Early Detection of Gynecological Malignancies Using Ensemble Deep Learning Models: ResNet50 and Inception V3

Informatics in Medicine Unlocked, Год журнала: 2025, Номер unknown, С. 101620 - 101620

Опубликована: Янв. 1, 2025

Язык: Английский

---

Hybrid deep learning framework for diabetic retinopathy classification with optimized attention AlexNet

Computers in Biology and Medicine, Год журнала: 2025, Номер 190, С. 110054 - 110054

Опубликована: Март 29, 2025

Язык: Английский

---

The Contribution of Federated Learning to AI Development

Опубликована: Июль 5, 2024

With the widespread application of artificial intelligence technology in various industries, users' attention to privacy and data security has increased significantly. Federated learning, as a new paradigm combining privacy-enhanced computing intelligence, resolves contradiction between open sharing. This paper presents benefits federated learning terms privacy, real-time processing, model robustness, compliance cross-industry applications. At same time, when combined with Edge AI technology, promotes decentralisation intelligent systems, improving protection accuracy. also discusses cases medical field, through local processing training, effectively protecting user realizing sharing optimization, promoting development intelligence.

Язык: Английский

---

Urban Planning and Green Building Technologies Based on Artificial Intelligence: Principles, Applications, and Global Case Study Analysis

Scientific Journal of Technology, Год журнала: 2024, Номер 6(8), С. 9 - 21

Опубликована: Авг. 21, 2024

The application of AI technology in urban planning covers multiple levels, such as data analysis, decision support, and automated planning. Urban research relies on to understand summarize the law growth improve analysis evolution trend space. Planning design use explore relevant factors affecting development their weights discuss critical role green building sustainable construction industry. With increase global energy consumption carbon emissions, traditional methods can no longer meet environmental protection requirements efficient resources. As a solution, has been paid more attention adopted by people. These technologies focus not only efficiency impact buildings but also resource utilization load over entire life cycle driven machine learning. This paper details basic principles applications technologies, including AI-driven reduction negative impacts, improvement occupant health, resources, optimization indoor quality. focuses LEED assessment system developed U.S. Green Building Council advancing practices. In addition, analyzes vital points water design, learning-driven wind environment optimization, solar application, practical cases these scale.

Язык: Английский

---