Leveraging transfer learning-driven convolutional neural network-based semantic segmentation model for medical image analysis using MRI images DOI Creative Commons
Amal Alshardan,

Nuha Alruwais,

Hamed Alqahtani

и другие.

Scientific Reports, Год журнала: 2024, Номер 14(1)

Опубликована: Дек. 18, 2024

Recognition and segmentation of brain tumours (BT) using MR images are valuable tedious processes in the healthcare industry. Earlier diagnosis localization BT provide timely options to select effective treatment plans for doctors can save lives. from Magnetic Resonance Images (MRI) is considered a big challenge owing difficulty tissues, segmenting them healthier tissue challenging when manual done through radiologists. Among recent proposals method, method based on machine learning (ML) image processing could be better. Thus, DL-based extensively applied, convolutional network has better effects. The deep model problem large loss information number parameters encoding decoding processes. With this motivation, article presents new Deep Transfer Learning with Semantic Segmentation Medical Image Analysis (DTLSS-MIA) technique MRI images. DTLSS-MIA aims segment affected area At first, presented utilizes Median filtering (MF) approach optimize quality remove noise. For semantic follows DeepLabv3 + backbone EfficientNet determining region. Moreover, CapsNet architecture employed feature extraction process. Lastly, crayfish optimization (CFO) diffusion variational autoencoder (D-VAE) used as classification mechanism, CFO effectively tunes D-VAE hyperparameter. simulation analysis validated benchmark dataset. performance validation exhibited superior accuracy value 99.53% over other methods.

Язык: Английский

Integrating pyramid vision transformer and topological data analysis for brain tumor DOI Creative Commons
Dhananjay Joshi, Bhupesh Kumar Singh,

Kapil Kumar Nagwanshi

и другие.

Frontiers in Computer Science, Год журнала: 2025, Номер 7

Опубликована: Апрель 10, 2025

Introduction Brain tumor (BT) classification is crucial yet challenging due to the complex and varied nature of these tumors. We present a novel approach combining Pyramid Vision Transformer (PVT) with an adaptive deformable attention mechanism Topological Data Analysis (TDA) address complexities BT detection. While PVT have been explored in prior work, we introduce key innovations enhance their performance for medical image analysis. Methods developed that dynamically adjusts receptive fields based on complexity, focusing critical regions MRI scans. The also incorporates sampling rate hierarchical dynamic position embeddings context-aware multi-scale feature extraction. Feature channels are partitioned into specialized groups via offset group improve diversity, strategy further integrates local global contexts yield refined representations. Additionally, applying TDA images extracts meaningful topological patterns, followed by Random Forest classifier final classification. Results method was evaluated Figshare brain dataset. It achieved 99.2% accuracy, 99.35% recall, 98.9% precision, 99.12% F1-score, Matthews correlation coefficient (MCC) 0.98, LogLoss 0.05, average processing time approximately 6 seconds per image. Discussion These results underscore method's ability combine detailed extraction insights, significantly improving accuracy efficiency proposed offers promising tool more reliable rapid diagnosis.

Язык: Английский

Процитировано

0
Enhanced Optimisation of MPLS Network Traffic using a Novel Adjustable Bat Algorithm with Loudness Optimizer DOI Creative Commons
Mohsin Masood,

Mohamed Mostafa Fouad,

Rashid Kamal

и другие.

Results in Engineering, Год журнала: 2025, Номер unknown, С. 104774 - 104774

Опубликована: Апрель 1, 2025

Язык: Английский

Процитировано

0
EPDTNet + -EM: Advanced Transfer Learning and SubNet Architecture for Medical Image Diagnosis DOI

K. Dhivya,

K Sangamithrai,

Indra Priyadharshini S

и другие.

Cognitive Computation, Год журнала: 2025, Номер 17(2)

Опубликована: Апрель 1, 2025

Язык: Английский

Процитировано

0
Enhanced BT segmentation with modified U-Net architecture: a hybrid optimization approach using CFO-SFO algorithm DOI

G. Yogalakshmi,

B. Sheela Rani

International Journal of Systems Assurance Engineering and Management, Год журнала: 2025, Номер unknown

Опубликована: Апрель 12, 2025

Язык: Английский

Процитировано

0
An Intelligent Medical Model for Classification of Brain Tumours and Stroke Lesions Using Machine Learning in Healthcare for Resource-Constrained Devices DOI
Ahed Abugabah

SN Computer Science, Год журнала: 2025, Номер 6(5)

Опубликована: Апрель 17, 2025

Язык: Английский

Процитировано

0
Advancements in Brain Tumour Analysis: A Review of Machine Learning, Deep Learning, Image Processing, and Explainable AI Techniques DOI

S. Venu Gopal,

Ch. Kavitha

Operations Research Forum, Год журнала: 2025, Номер 6(2)

Опубликована: Май 5, 2025

Язык: Английский

Процитировано

0
Optimizing Convolutional Neural Networks: A Comprehensive Review of Hyperparameter Tuning Through Metaheuristic Algorithms DOI
Mohamed F. Ibrahim, Nazar K. Hussein, David Guinovart-Sanjuán

и другие.

Archives of Computational Methods in Engineering, Год журнала: 2025, Номер unknown

Опубликована: Май 20, 2025

Язык: Английский

Процитировано

0
Bio-Inspired Metaheuristics in Deep Learning for Brain Tumor Segmentation: A Decade of Advances and Future Directions DOI Creative Commons
Shoffan Saifullah, Rafał Dreżewski, Anton Yudhana

и другие.

Information, Год журнала: 2025, Номер 16(6), С. 456 - 456

Опубликована: Май 29, 2025

Accurate segmentation of brain tumors in magnetic resonance imaging (MRI) remains a challenging task due to heterogeneous tumor structures, varying intensities across modalities, and limited annotated data. Deep learning has significantly advanced accuracy; however, it often suffers from sensitivity hyperparameter settings generalization. To overcome these challenges, bio-inspired metaheuristic algorithms have been increasingly employed optimize various stages the deep pipeline—including tuning, preprocessing, architectural design, attention modulation. This review systematically examines developments 2015 2025, focusing on integration nature-inspired optimization methods such as Particle Swarm Optimization (PSO), Genetic Algorithm (GA), Differential Evolution (DE), Grey Wolf Optimizer (GWO), Whale (WOA), novel hybrids including CJHBA BioSwarmNet into learning-based frameworks. A structured multi-query search strategy was executed using Publish or Perish Google Scholar Scopus databases. Following PRISMA guidelines, 3895 records were screened through automated filtering manual eligibility checks, yielding curated set 106 primary studies. Through bibliometric mapping, methodological synthesis, performance analysis, we highlight trends algorithm usage, application domains (e.g., architecture search), outcomes measured by metrics Dice Similarity Coefficient (DSC), Jaccard Index (JI), Hausdorff Distance (HD), ASSD. Our findings demonstrate that enhances accuracy robustness, particularly multimodal involving FLAIR T1CE modalities. The concludes identifying emerging research directions hybrid optimization, real-time clinical applicability, explainable AI, providing roadmap for future exploration this interdisciplinary domain.

Язык: Английский

Процитировано

0
Vital Characteristics Cellular Neural Network (VCeNN) for Melanoma Lesion Segmentation: A Biologically Inspired Deep Learning Approach DOI

Tongxin Yang,

Qilin Huang,

F.F. Cai

и другие.

Deleted Journal, Год журнала: 2024, Номер unknown

Опубликована: Сен. 16, 2024

Cutaneous melanoma is a highly lethal form of cancer. Developing medical image segmentation model capable accurately delineating lesions with high robustness and generalization presents formidable challenge. This study draws inspiration from cellular functional characteristics natural selection, proposing novel named the vital neural network. incorporates observed in multicellular organisms, including memory, adaptation, apoptosis, division. Memory module enables network to rapidly adapt input data during early stages training, accelerating convergence. Adaptation allows neurons select appropriate activation function based on varying environmental conditions. Apoptosis reduces risk overfitting by pruning low values. Division enhances network's learning capacity duplicating Experimental evaluations demonstrate efficacy this enhancing performance networks for segmentation. The proposed method achieves outstanding results across numerous publicly available datasets, indicating its potential contribute significantly field analysis facilitating accurate efficient imagery. an F1 score 0.901, Intersection over Union 0.841, Dice coefficient 0.913,

Язык: Английский

Процитировано

2
Convolutional neural network for oral cancer detection combined with improved tunicate swarm algorithm to detect oral cancer DOI Creative Commons

Xiao Wei,

Chanjuan Liu, Ke Jiang

и другие.

Scientific Reports, Год журнала: 2024, Номер 14(1)

Опубликована: Ноя. 19, 2024

Early Diagnosis of oral cancer is very important and can save you from some malignancies. However, while this approach aids in the rapid healing patients preservation their lives, there are several causes for poor wrong diagnosis cancer. In recent years, use computer-aided design tools as an auxiliary tool alongside clinicians has greatly benefited more accurate identification malignancy. The current study proposes a new identifying based on image processing deep learning. employs recently integrated model improved tunicate swarm algorithm to produce efficient improving convolutional neural network delivering diagnostic system. then implemented pictures dataset. validated by comparing it other published papers using various measurement markers. proposed achieved accuracy 98.70% recall 93.71% detecting cancerous lesions photographic images. also F1-score 90.08% precision 96.42%. final results demonstrate that offered exact be used conjunction with help diagnosing

Язык: Английский

Процитировано

2