Fusion of transformer attention and CNN features for skin cancer detection

Applied Soft Computing, Journal Year: 2024, Volume and Issue: 164, P. 112013 - 112013

Published: July 18, 2024

Language: Английский

---

DicomOS: A Preliminary Study on a Linux-Based Operating System Tailored for Medical Imaging and Enhanced Interoperability in Radiology Workflows

Electronics, Journal Year: 2025, Volume and Issue: 14(2), P. 330 - 330

Published: Jan. 15, 2025

In this paper, we propose a Linux-based operating system, namely, DicomOS, tailored for medical imaging and enhanced interoperability, addressing user-friendly functionality the main critical needs in radiology workflows. Traditional systems clinical settings face limitations, such as fragmented software ecosystems platform-specific restrictions, which disrupt collaborative workflows hinder diagnostic efficiency. Built on Ubuntu 22.04 LTS, DicomOS integrates essential DICOM functionalities directly into OS, providing unified, cohesive platform image visualization, annotation, sharing. Methods include custom configurations development of graphical user interfaces (GUIs) command-line tools, making them accessible to professionals developers. Key applications ITK-SNAP 3D Slicer are seamlessly integrated alongside specialized GUIs that enhance usability without requiring extensive technical expertise. As preliminary work, demonstrates potential simplify workflows, reduce cognitive load, promote efficient data sharing across diverse settings. However, further evaluations, including structured tests broader deployment with distributable ISO image, must validate its effectiveness scalability real-world scenarios. The results indicate provides versatile adaptable solution, supporting radiologists routine tasks while facilitating customization advanced users. an open-source platform, has evolve needs, positioning it valuable resource enhancing workflow integration collaboration.

Language: Английский

---

Probing the Efficacy of Federated Parameter-Efficient Fine-Tuning of Vision Transformers for Medical Image Classification

Lecture notes in computer science, Journal Year: 2025, Volume and Issue: unknown, P. 236 - 245

Published: Jan. 1, 2025

Language: Английский

---

Fine-tuned deep transfer learning: an effective strategy for the accurate chronic kidney disease classification

PeerJ Computer Science, Journal Year: 2025, Volume and Issue: 11, P. e2800 - e2800

Published: April 8, 2025

Kidney diseases are becoming an alarming concern around the globe. Premature diagnosis of kidney disease can save precious human lives by taking preventive measures. Deep learning demonstrates a substantial performance in various medical disciplines. Numerous deep approaches suggested literature for accurate chronic classification compromising on architectural complexity, speed, and resource constraints. In this study, transfer is exploited incorporating unexplored yet effective variants ConvNeXt EfficientNetV2 efficient diseases. The benchmark computed tomography (CT)-based database containing 12,446 CT scans tumor, stone cysts, normal patients utilized to train designed fine-tuned networks. However, due highly imbalanced distribution images among classes, operation data trimming balancing number each class, which essential designing unbiased predictive network. By utilizing pre-trained models our specific task, training time reduced leading computationally inexpensive solution. After comprehensive hyperparameters tuning with respect changes rates, batch sizes, optimizers, it depicted that EfficientNetV2B0 network 23.8 MB size only 6.2 million parameters shows diagnostic achieving generalized test accuracy 99.75% balanced database. Furthermore, attains high precision, recall, F1-score 99.75%, 99.63%, respectively. Moreover, final ensures its scalability impressive 99.73% set original dataset as well. Through extensive evaluation proposed strategy, concluded design outperforms counterparts terms computational efficiency tasks. serves accurate, efficient, solution tailored real-time deployment or mobile edge devices.

Language: Английский

---

Automated explainable deep learning framework for multiclass skin cancer detection and classification using hybrid YOLOv8 and vision transformer (ViT)

Biomedical Signal Processing and Control, Journal Year: 2025, Volume and Issue: 108, P. 107934 - 107934

Published: April 29, 2025

Language: Английский

---

Least square-support vector machine based brain tumor classification system with multi model texture features

Frontiers in Applied Mathematics and Statistics, Journal Year: 2023, Volume and Issue: 9

Published: Dec. 6, 2023

Radiologists confront formidable challenges when confronted with the intricate task of classifying brain tumors through analysis MRI images. Our forthcoming manuscript introduces an innovative and highly effective methodology that capitalizes on capabilities Least Squares Support Vector Machines (LS-SVM) in tandem rich insights drawn from Multi-Scale Morphological Texture Features (MMTF) extracted T1-weighted MR underwent meticulous evaluation a substantial dataset encompassing 139 cases, consisting 119 cases aberrant 20 normal The outcomes we achieved are nothing short extraordinary. LS-SVM-based approach vastly outperforms competing classifiers, demonstrating its dominance exceptional accuracy rate 98.97%. This represents 3.97% improvement over alternative methods, accompanied by notable 2.48% enhancement Sensitivity 10% increase Specificity. These results conclusively surpass performance traditional classifiers such as (SVM), Radial Basis Function (RBF), Artificial Neural Networks (ANN) terms classification accuracy. outstanding our model realm tumor diagnosis signifies leap forward field, holding promise delivering more precise dependable tools for radiologists healthcare professionals their pivotal role identifying using imaging techniques.

Language: Английский

---

Distinguishing between Crohn’s disease and ulcerative colitis using deep learning models with interpretability

Pattern Analysis and Applications, Journal Year: 2024, Volume and Issue: 27(1)

Published: Jan. 25, 2024

Abstract Crohn’s disease and ulcerative colitis are two chronic diseases that cause inflammation in the tissues of entire gastrointestinal tract described by term inflammatory bowel disease. Gastroenterologists find it difficult to evaluate endoscopic images recognise characteristics diseases. Therefore, this work aims build a dataset with (collected from public datasets LIMUC, HyperKvasir CrohnIPI) train deep learning models (five CNNs six ViTs) develop tool capable helping doctors distinguish type In addition, as these architectures will be too heavy hospital context, work, we looking use knowledge distillation create lighter simpler same precision pre-trained used study. During process, is important interpret before resulting ensure can maintain performance information learnt both similar. It concluded possible reduce 25x number parameters while maintaining good reducing inference time 5.32 s. Allied this, through interpretability was after identify ulcers, bleeding situations, lesions caused

Language: Английский

---

Dual-task kidney MR segmentation with transformers in autosomal-dominant polycystic kidney disease

Computerized Medical Imaging and Graphics, Journal Year: 2024, Volume and Issue: 113, P. 102349 - 102349

Published: Feb. 7, 2024

Autosomal-dominant polycystic kidney disease is a prevalent genetic disorder characterized by the development of renal cysts, leading to enlargement and failure. Accurate measurement total volume through segmentation crucial assess severity, predict progression evaluate treatment effects. Traditional manual suffers from intra- inter-expert variability, prompting exploration automated approaches. In recent years, convolutional neural networks have been employed for magnetic resonance images. However, use Transformer-based models, which shown remarkable performance in wide range computer vision medical image analysis tasks, remains unexplored this area. With their self-attention mechanism, Transformers excel capturing global context information, accurate organ delineations. paper, we compare various convolutional-based, Transformers-based, hybrid convolutional/Transformers-based segmentation. Additionally, propose dual-task learning scheme, where common feature extractor followed per-kidney decoders, towards better generalizability efficiency. We extensively architectures schemes on heterogeneous imaging dataset collected 112 patients with disease. Our results highlight effectiveness models relevancy exploiting improve accuracy mitigate data scarcity issues. A promising ability accurately delineating kidneys especially presence cyst distributions adjacent cyst-containing organs. This work contribute advancement reliable delineation methods nephrology, paving way broad spectrum clinical applications.

Language: Английский

---

A novel approach for melanoma detection utilizing GAN synthesis and vision transformer

Computers in Biology and Medicine, Journal Year: 2024, Volume and Issue: 176, P. 108572 - 108572

Published: May 9, 2024

Language: Английский

---

The effect of hair removal and filtering on melanoma detection: a comparative deep learning study with AlexNet CNN

PeerJ Computer Science, Journal Year: 2024, Volume and Issue: 10, P. e1953 - e1953

Published: April 16, 2024

Melanoma is the most aggressive and prevalent form of skin cancer globally, with a higher incidence in men individuals fair skin. Early detection melanoma essential for successful treatment prevention metastasis. In this context, deep learning methods, distinguished by their ability to perform automated detailed analysis, extracting melanoma-specific features, have emerged. These approaches excel performing large-scale optimizing time, providing accurate diagnoses, contributing timely treatments compared conventional diagnostic methods. The present study offers methodology assess effectiveness an AlexNet-based convolutional neural network (CNN) identifying early-stage melanomas. model trained on balanced dataset 10,605 dermoscopic images, modified datasets where hair, potential obstructive factor, was detected removed allowing assessment how hair removal affects model’s overall performance. To removal, we propose morphological algorithm combined different filtering techniques comparison: Fourier, Wavelet, average blur, low-pass filters. evaluated through 10-fold cross-validation metrics accuracy, recall, precision, F1 score. results demonstrate that proposed performs best implemented both Wavelet filter algorithm. It has accuracy 91.30%, recall 87%, precision 95.19%, score 90.91%.

Language: Английский

---