Editorial: Recent Advances in Deep Learning and Medical Imaging for Cancer Treatment

Cancers, Journal Year: 2024, Volume and Issue: 16(4), P. 700 - 700

Published: Feb. 7, 2024

In the evolving landscape of medical imaging, escalating need for deep-learningmethods takes center stage, offering capability to autonomously acquire abstract datarepresentations crucial early detection and classification cancer treatment. Thecomplexities in handling diverse inputs, high-dimensional features, subtle patternswithin imaging data are acknowledged as significant challenges this technologicalpursuit. This Special Issue, “Recent Advances Deep Learning Medical Imagingfor Cancer Treatment”, has attracted 19 high-quality articles that cover state-of-the-artapplications technical developments deep learning, automaticdetection, classification, explainable artificial intelligence-enabled diagnosis cancertreatment. ever-evolving treatment, five pivotal themes haveemerged beacons transformative change. editorial delves into realms ofinnovation shaping future focusing on interconnectedthemes: use intelligence applications AI cancerdiagnosis addressing image analysis, advancementsin techniques, innovations skin classification.

Language: Английский

---

A deep learning-based model for detecting Leishmania amastigotes in microscopic slides: a new approach to telemedicine

BMC Infectious Diseases, Journal Year: 2024, Volume and Issue: 24(1)

Published: June 1, 2024

Abstract Background Leishmaniasis, an illness caused by protozoa, accounts for a substantial number of human fatalities globally, thereby emerging as one the most fatal parasitic diseases. The conventional methods employed detecting Leishmania parasite through microscopy are not only time-consuming but also susceptible to errors. Therefore, main objective this study is develop model based on deep learning, subfield artificial intelligence, that could facilitate automated diagnosis leishmaniasis. Methods In research, we introduce LeishFuNet, learning framework designed parasites in microscopic images. To enhance performance our same-domain transfer initially train four distinct models: VGG19, ResNet50, MobileNetV2, and DenseNet 169 dataset related another infectious disease, COVID-19. These trained models then utilized new pre-trained fine-tuned set 292 self-collected high-resolution images, consisting 138 positive cases 154 negative cases. final prediction generated fusion information analyzed these models. Grad-CAM, explainable intelligence technique, implemented demonstrate model’s interpretability. Results results utilizing amastigotes images follows: accuracy 98.95 1.4%, specificity 98 2.67%, sensitivity 100%, precision 97.91 2.77%, F1-score 98.92 1.43%, Area Under Receiver Operating Characteristic Curve 99 1.33. Conclusion newly devised system precise, swift, user-friendly, economical, thus indicating potential substitute prevailing leishmanial diagnostic techniques.

Language: Английский

---

Deep learning for enhanced porosity detection in AZ91 magnesium alloys using windowed perception and aggregated sensing

Journal of Materials Informatics, Journal Year: 2025, Volume and Issue: 5(2)

Published: March 19, 2025

In this study, we innovatively proposed a deep learning model architecture to address the industry challenges in detection of porosity magnesium alloys. Magnesium alloys, known for their lightweight and high-strength characteristics, are extensively utilized aerospace, automotive, biomedical fields. However, absorption hydrogen during production process leads formation pores, which not only reduce material’s strength durability but may also cause premature failure material. The pores typically occurs solidification stage where dissolved molten metal is released upon cooling, forming tiny gas pores. presence these significantly affects mechanical properties material, potentially leading crack initiation propagation under high stress. Therefore, accurate quantification crucial enhancing quality control Our developed integrates window-shaped perception blocks with convolutional neural networks, enhanced by aggregated sensing layers (ASLs) on long-range connections. Extensive training real samples demonstrated that our outperforms mainstream algorithms such as U-Net TransUNet across various evaluation metrics, particularly fine target tasks complex scenarios. Specifically, achieved Dice coefficient 74.77% an Intersection over Union index 71.00%, surpassing other models. Moreover, method superior accuracy pore edge prediction, effectively mitigating issues oversegmentation undersegmentation, especially small irregular An ablation study further confirmed effectiveness each component, ASL module showing particular feature extraction reducing upsampling loss. summary, research highlights significant potential technology material defect provides efficient, automated solution practical production, contributing advancements materials science industrial control.

Language: Английский

---

Generalisable deep Learning framework to overcome catastrophic forgetting

Intelligent Systems with Applications, Journal Year: 2024, Volume and Issue: 23, P. 200415 - 200415

Published: July 10, 2024

Generalisation across multiple tasks is a major challenge in deep learning for medical imaging applications, as it can cause catastrophic forgetting problem. One commonly adopted approach to address these challenges train the model from scratch, incorporating old and new data, classes, tasks. However, this solution comes with its downsides, time-consuming, requires high computational resources, susceptible bias, lacks flexibility. To effectively issues, paper introduces generalisable DL framework that consists of three key components: self-supervised learning, feature fusion single task, classes or Using proposed framework, models SVM classifier accurately detect abnormalities X-ray tasks, including humerus wrist, achieving an accuracy 92.71% 90.74%, respectively. These results were achieved using minimal training requirements when introduced. Another experiment was performed on chest X-rays, where added pre-existing ones. Without requiring retraining both our combined class 98.18%. This demonstrates has not forgotten data. The enhances performance brings flexibility efficiency process, saving time resources.

Language: Английский

---

Analyzing emotions in online classes: Unveiling insights through topic modeling, statistical analysis, and random walk techniques

International Journal of Cognitive Computing in Engineering, Journal Year: 2024, Volume and Issue: 5, P. 221 - 236

Published: Jan. 1, 2024

High dropout rates globally perpetuate educational disparities with various underlying causes. Despite numerous strategies to address this issue, more attention should be given understanding and addressing student emotions during classes. This lack of focus adversely affects learner engagement retention rates. While previous studies on online learning have primarily emphasized the effectiveness technology, infrastructure, cognition, motivation, economic benefits, there is still a gap in emotional aspects distance learning. First, study addresses by employing thematic modeling utilizing non-negative matrix factorization (NMF) for emotion recognition through students' deep techniques facial (FER). Second, statistical analysis these findings further augments depth study. Finally, research proposes mathematical model based random walk state transitions. The underscore importance considering environments their significant impact student's academic performance satisfaction. By acknowledging factors, educators can enhance engagement, promote positive emotions, mitigate negative learning, ultimately improve courses.

Language: Английский

---

Resnet Transfer Learning For Enhanced Medical Image Classification In Healthcare

Published: Dec. 29, 2023

This work overcomes the limitations of sparsely labeled data by optimizing ResNet transfer learning methods in medical classification images. Using a deductive approach along with interpretive philosophy, we optimize for better diagnostic performance on healthcare sets. Our team technical includes preprocessing datasets, configuring model architectures, and fine-tuning hyperparameters using secondary data. The improved as demonstrated results is confirmed metrics such precision, reliability, recall. Analyses comparisons demonstrate superiority over basic models. Upcoming tasks include working together to create standardized benchmarks, improving interpretability scalability, verifying actual clinical settings.

Language: Английский

---

FEC-IGE: An Efficient Approach to Classify Fracture Based on Convolutional Neural Networks and Integrated Gradients Explanation

International Journal of Advanced Computer Science and Applications, Journal Year: 2024, Volume and Issue: 15(6)

Published: Jan. 1, 2024

In this paper, we propose the FEC-IGE framework includes data preprocessing, augmentation, transfer learning, and fine-tuning of pre-trained model convolutional neural network (CNN) architecture for problem bone fracture classification. Bone fractures are a widespread medical issue globally, with significant prevalence imposing substantial burdens on individuals healthcare systems. The impact extends beyond physical injury, often leading to pain, reduced mobility, decreased quality life affected individuals. Moreover, can incur economic costs due expenses, rehabilitation, lost productivity. recent years, progress in machine learning methodologies has exhibited potential tackling issues pertaining diagnosis By harnessing capabilities deep frameworks, scholars aspire design precise effective mechanisms automatically detecting classifying from imaging data. study, demonstrated its strength when applied models CNN task X-ray images accuracies 98.48%, 96.92%, 97.24% three experimental scenarios. These outcomes consequence model's procedures an enhanced dataset including 1129 pictures classified into ten different kinds fractures: avulsion fracture, comminuted dislocation, greenstick hairline impacted longitudinal oblique pathological spiral fracture. To increase transparency understanding model, Integrated Gradients explanation was also study. Finally, metrics precision, recall, F1 score, confusion matrix were evaluate performance other in-depth analysis.

Language: Английский

---

Fusion of GoogleNet and Capsule Neural Network models for improving bone tumor classification

2022 IEEE 7th International conference for Convergence in Technology (I2CT), Journal Year: 2024, Volume and Issue: unknown

Published: April 5, 2024

Combining the strengths of GoogleNet and Capsule Neural Network (CNN) models represents a novel approach to enhance categorization bone cancers, critical aspect for accurate diagnosis effective therapy planning. Current classification techniques often encounter challenges in achieving both resilience high accuracy. In this study, we address these issues by leveraging distinct features extracted from tumor images using model. These are subsequently fed into CNN model, known its proficiency capturing complex spatial correlations within data. The proposed fusion strategy aims improve precision synergistically utilizing two models. efficacy is evaluated on publicly available dataset, revealing notable accuracy 96.7%. Comparative analysis against state-of-the-art underscores superior performance integrated conclusion, amalgamation presents promising avenue advancing classification, potentially leading more diagnoses treatment plans individuals with conditions.

Language: Английский

---

OMSF2: optimizing multi-scale feature fusion learning for pneumoconiosis staging diagnosis through data specificity augmentation

Complex & Intelligent Systems, Journal Year: 2024, Volume and Issue: 11(1)

Published: Dec. 30, 2024

Language: Английский

---

Deep-Learning-Based Multitask Ultrasound Beamforming

Information, Journal Year: 2023, Volume and Issue: 14(10), P. 582 - 582

Published: Oct. 23, 2023

In this paper, we present a new method for multitask learning applied to ultrasound beamforming. Beamforming is critical component in the image formation pipeline. Ultrasound images are constructed using sensor readings from multiple transducer elements, with each element typically capturing acquisitions per frame. Hence, beamformer crucial framerate performance and overall quality. Furthermore, post-processing, such as denoising, usually beamformed achieve high clarity diagnosis. This work shows fully convolutional neural network that can learn different tasks by applying weight normalization scheme. We adapt our model both frame rate requirements fitting parameters sub-sampling task denoising optimizing speckle reduction task. Our outperforms single-angle delay sum on pixel-level measures noise reduction, subsampling, reconstruction.

Language: Английский

---