An Efficient Gaussian Noise Denoising in CT Images with Deep Convolutional Neural Networks

Biomedical & Pharmacology Journal, Journal Year: 2025, Volume and Issue: 18(December Spl Edition), P. 139 - 159

Published: Jan. 20, 2025

This paper introduces a new deep learning paradigm using the Denoising Convolutional-Neural Network (DnCNN) model for denoising Gaussian noise in Computed Tomography (CT) images. By nature, is inherently random and additive, potentially obscuring vital diagnostic features significantly reducing image quality, resulting difficulties medical interpretation. Initially, distorted images are sourced from addition of with different intensity levels (σ = 5,10,15,20). The process DnCNN employs convolutional neural network that maps noisy to clean image, focusing on residual prevent loss detail. CT obtained after assessed quantitative measures like Peak signal ratio (PSNR), Signal (SNR), Structural similarity index measure (SSIM) Entropy difference (ED). proposed evaluated metrics, such as PSNR, SNR, SSIM, ED, demonstrating better performance than standard algorithms, including Total Variation, BM3D, Guided, Bilateral, Anisotropic Diffusion filters. experimental results show outperforms conventional methods. achieves PSNR 35.66 dB, an SNR 30.16 SSIM 0.91 ED 0.35. Additionally, zooming analysis profile evaluations confirms method effectively suppresses while preserving sharper edges finer anatomical structures. ensures superior visual quality greater efficacy compared traditional These findings confirm robust strategy imaging predicting accurate outcomes.

Language: Английский

---

GMamba: State space model with convolution for Grape leaf disease segmentation

Computers and Electronics in Agriculture, Journal Year: 2024, Volume and Issue: 225, P. 109290 - 109290

Published: Aug. 6, 2024

Language: Английский

---

A method framework of cruciate ligaments segmentation and reconstruction from MRI images

Technology and Health Care, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 26, 2025

Segmenting anterior and posterior cruciate ligaments (ACL/PCL) presents challenges in medical imaging due to diverse characteristics, including size, shape, intensity. Our study uses superpixel-based spectral clustering for knee ligament segmentation 2D DICOM slices, renowned generating high-quality clusters. The proposed method addresses the by (i) identifying ligamentous region (ROI) through computation, (ii) extracting features (intensity-based, shape-based, geometric complexity, Scale-Invariant Feature Transform) from ROI, (iii) segmenting tissues using on extracted features. Superpixel-based challenge of constructing a dense similarity matrix significantly reduces computational burden. Furthermore, 3D visualization structures is performed Visualization Toolkit (VTK). We evaluated our approach dataset MRI assessing results via dice score, average surface distance (ASD), root mean squared error (RMSE) metrics. achieved an score 0.912 ACL 0.896 PCL segmentation, outperforming other methods. These scores showed enhancement 10.7% 14.9% accuracy PCL, respectively. reduced margins were demonstrated with ASD values 1.60 1.78 RMSE 1.76 1.86 show effectiveness its potential increasing speed, offering significant advantages over manual reducing time expertise.

Language: Английский

---

Computer-aided diagnosis of spinal deformities based on keypoints detection in human back depth images

Biomedical Signal Processing and Control, Journal Year: 2025, Volume and Issue: 106, P. 107764 - 107764

Published: March 7, 2025

Language: Английский

---

Dynamic monitoring of surface area and water volume of reservoirs using satellite imagery, computer vision and deep learning

Remote Sensing Applications Society and Environment, Journal Year: 2024, Volume and Issue: 35, P. 101205 - 101205

Published: April 21, 2024

Language: Английский

---

Evaluation of Speckle Noise Reduction Filters and Machine Learning Algorithms for Ultrasound Images

IEEE Access, Journal Year: 2024, Volume and Issue: 12, P. 81293 - 81312

Published: Jan. 1, 2024

Language: Английский

---

Leveraging smart image processing techniques for early detection of foot ulcers using a deep learning network

Polish Journal of Radiology, Journal Year: 2024, Volume and Issue: 89, P. 368 - 377

Published: July 31, 2024

Purpose To detect foot ulcers in diabetic patients by analysing thermal images of the using a deep learning model and estimate effectiveness proposed comparing it with some existing studies. Material methods Open-source were used for study. The dataset consists two types feet patients: normal abnormal images. contains 1055 total images; among these, 543 are images, others patient. study’s was converted into new pre-processed applying canny edge detection watershed segmentation. This then balanced enlarged data augmentation, after that, prediction, applied diagnosis an ulcer foot. After segmentation, can enhance model’s performance correct predictions reduce computational cost. Results Our model, utilizing ResNet50 EfficientNetB0, tested on both original results highly promising, achieving 89% 89.1% accuracy datasets, respectively, EfficientNetB0 surpassing this 96.1% 99.4% respectively. Conclusions study offers practical solution detection, particularly situations where expert analysis is not readily available. efficacy our models real they outperformed other available models, demonstrating their potential real-world application.

Language: Английский

---

Figure-ground segmentation based medical image denoising using deep convolutional neural networks

International Journal of Computers and Applications, Journal Year: 2024, Volume and Issue: unknown, P. 1 - 27

Published: Nov. 4, 2024

Medical image noise can have a substantial impact on both diagnosis accuracy and quality. Noise cause problems with tasks related to diagnosis, like defining object boundaries, which are essential for precise diagnosis. By applying denoising techniques, medical imaging professionals significantly improve quality, reduce errors, enhance the of diagnoses treatments. This article presents method reducing noises Gaussian noise, salt pepper speckle ring artefacts in images, such as magnetic resonance (MRI), computed tomography (CT), chest X-ray images. study explores integration adaptive CNNs guided filtering enhanced quality deep learning-driven Figure-Ground segmentation. The proposed method's performance is extensively evaluated MRI/CT images under diverse scenarios, comparisons existing techniques using established statistical metrics. results validate that approach attains superior performance, yielding highest values across these

Language: Английский

---

Enhancing Low-Light Medical Imaging through Deep Learning-Based Noise Reduction Techniques

Indian Journal of Science and Technology, Journal Year: 2024, Volume and Issue: 17(34), P. 3567 - 3579

Published: Sept. 2, 2024

Background/ Objectives: Low-light medical imaging is highly challenging in clinical diagnostics due to increased noise levels that mask or obscure important anatomical details. In this respect, conventional reduction methods such as Gaussian filtering and median usually lead a trade-off between suppression the preservation of features an image, thus resulting poor-quality images. More advanced wavelet-based denoising Non-Local Means exhibit superior but remain computationally intensive introduce artifacts. These challenges come with need develop more effective efficient noise-reduction techniques. Methods: This study proposes end-to-end deep learning framework for low-light image enhancement. We present comprehensive deep-learning enhance images by integrating Convolutional Neural Networks autoencoders build robust model. The CNN extracts feature from noisy input images, while autoencoder does so reconstruction clean through encoding lower-dimensional representation retaining critical information. Findings: validates proposed model rigorous quantitative metrics peak signal-to-noise ratio structural similarity index. are designed provide full assessment quality concerning capability details related structure. Our improves traditional PSNR about 5 dB on average SSIM 0.15, which means better A comparative analysis techniques has been included, pointing out advantages approaches. Experimental results depict significant improvements over previous For instance, reduces level up 40% facilitates clear sharp 30%. terms quantification, these manifest value 35 score 0.85 compared 30 0.70 using Furthermore, illustrates training dynamics, maps, evolution model's incremental process. Novelty: study's findings validate efficacy enhancing diagnosis accuracy improving patient outcomes imaging. Keywords: imaging, Noise reduction, Networks, Denoising autoencoders, Medical

Language: Английский

---

Deep residual learning-based denoiser for medical X-ray images

Evolving Systems, Journal Year: 2024, Volume and Issue: unknown

Published: Sept. 6, 2024

Language: Английский

---