Learning Pixel Level Affinity with Class Labels for Weakly Supervised Segmentation of Lung Cavities

Research Square (Research Square), Journal Year: 2025, Volume and Issue: unknown

Published: April 7, 2025

Accurately annotating lung cavities (LCs) at the pixel level in computed tomography (CT) images presents a significant challenge due to their diverse shapes and sizes. To address this limitation, weakly supervised semantic segmentation (WSSS) methods utilizing sparse annotations, such as image-level labels, have emerged promising trend. This paper proposes novel scribble-supervised framework for LCs that leverages annotation-driven affinity. The introduces bidirectional interaction Mamba UNet model, named MambaUNeLCsT, designed inefficiency of transformer models processing long sequences. refine coarse pseudo-labels, an attention-based affinity pseudo-label refinement module is incorporated, employing algorithm establish associations between unlabeled pseudo-labeled samples. approach infers labels samples by computing sample similarities. Additionally, overcome limited spatial supervision provided scribble-based included, effectively capturing complete morphology boundary information LCs. enhances model’s capability recognize process fine structures. Experimental results demonstrate MambaUNeLCsT achieves state-of-the-art performance 3D medical image segmentation, outperforming existing WSSS tasks.

Language: Английский

---

An adaptive convolution neural network model for tuberculosis detection and diagnosis using semantic segmentation

Polish Journal of Radiology, Journal Year: 2025, Volume and Issue: 90, P. 124 - 137

Published: March 14, 2025

Purpose Tuberculosis (TB) continues to be a major cause of death from infectious diseases globally. TB is treatable with antibiotics, but it often misdiagnosed or left untreated, particularly in rural and resource-constrained regions. While chest X-rays are key tool diagnosis, their effectiveness hindered by the variability radiological presentations lack trained radiologists high-prevalence areas. Deep learning-based imaging techniques offer promising approach computer-aided diagnosis for TB, enabling precise timely detection while alle­viating burden on healthcare professionals. This study aims enhance X-ray images developing deep learning models. We have observed upper lower lobe consolidation, pleural effusion, calcification, cavity formation military nodules. A proposed preprocessing technique has been also introduced our work based gamma correction gradient contrast enhancement. leverage Res-UNet architecture image segmentation introduce novel network classification, targeting improved accuracy precision diagnostic performance. Material methods model was using 704 sourced Montgomery County Shenzhen Hospital datasets. Following training, applied segment lung regions 1400 scans, encompassing both cases normal controls, obtained National Institute Allergy Infectious Diseases (NIAID) Portal program dataset. The segmented were subsequently classified as either model. used enhancement capturing intensity changes comparing each pixel its neighbour pyramid reduction unique mapping histogram matching along used. integrated classification images. Classification done customised convolutional neural network, visualisation Grad-CAM. Results demonstrated excellent performance segmentation, achieving an 98.18%, recall 98.40%, 97.45%, F1-score 97.97%, Dice coefficient 96.33%, Jaccard index 96.05%. Similarly, exhibited outstanding results, 99.45%, 99.29%, AUC 99.9%. Enhanced method showed ambe 16.51, entropy 6.7370, CII 86.80, psnr 28.71, ssim 86.83 which quite satisfactory. Conclusions findings demonstrate efficiency system diagnosing X-rays, potentially surpassing clinician-level precision. underscores tool, resourcelimited settings restricted access expertise. Additionally, modified superior compared standard U-Net, highlighting potential greater dia­gnostic accuracy.

Language: Английский

---

DLSAC-Net: An automated enhanced segmentation and classification network for lung diseases detection using chest X-Ray images

Multimedia Tools and Applications, Journal Year: 2025, Volume and Issue: unknown

Published: April 10, 2025

Language: Английский

---

DeepPulmoTB: A benchmark dataset for multi-task learning of tuberculosis lesions in lung computerized tomography (CT)

Heliyon, Journal Year: 2024, Volume and Issue: 10(4), P. e25490 - e25490

Published: Feb. 1, 2024

Tuberculosis (TB) remains a significant global health challenge, characterized by high incidence and mortality rates on scale. With the rapid advancement of computer-aided diagnosis (CAD) tools in recent years, CAD has assumed an increasingly crucial role supporting TB diagnosis. Nonetheless, development for heavily relies well-annotated computerized tomography (CT) datasets. Currently, available annotations CT datasets are still limited, which turn restricts to some extent. To address this limitation, we introduce DeepPulmoTB, multi-task learning dataset explicitly designed demonstrate advantages propose novel model, DeepPulmoTBNet (DPTBNet), joint segmentation classification lesion tissues images. The architecture DPTBNet comprises two subnets: SwinUnetR task, lightweight multi-scale network task. Furthermore, enhance model's capacity capture features, improved iterative optimization algorithm that refines feature maps integrating probability obtained previous iterations. Extensive experiments validate effectiveness practicality DeepPulmoTB dataset.

Language: Английский

---

Early diagnosis and meta-agnostic model visualization of tuberculosis based on radiography images

Scientific Reports, Journal Year: 2023, Volume and Issue: 13(1)

Published: Dec. 20, 2023

Despite being treatable and preventable, tuberculosis (TB) affected one-fourth of the world population in 2019, it took lives 1.4 million people 2019. It 1.2 children around same year. As is an infectious bacterial disease, early diagnosis TB prevents further transmission increases survival rate person. One standard methods sputum culture test. Diagnosing rapid test results usually take one to eight weeks 24 h. Using posterior-anterior chest radiographs (CXR) facilitates a more cost-effective tuberculosis. Due intraclass variations interclass similarities images, prognosis from CXR difficult. We proposed system (tbXpert) based on deep learning methods. Deep Fused Linear Triangulation (FLT) considered for images reconcile variation similarities. To improve robustness approach, information must be obtained minimal radiation uneven quality images. The advanced FLT method accurately visualizes infected region without segmentation. fused are trained by network (DLN) with residual connections. largest database, comprised 3500 normal utilized training validating recommended model. Specificity, sensitivity, Accuracy, AUC estimated determine performance systems. demonstrates maximum testing accuracy 99.2%, sensitivity 98.9%, specificity 99.6%, precision 99.4%, all which pretty high when compared current state-of-the-art approaches lessen radiologist's time, effort, reliance level competence specialist, suggested named tbXpert can deployed as computer-aided technique

Language: Английский

---

A Systematic Review: Classification of Lung Diseases from Chest X-Ray Images Using Deep Learning Algorithms

SN Computer Science, Journal Year: 2024, Volume and Issue: 5(4)

Published: April 6, 2024

Language: Английский

---

An accurate prediction for respiratory diseases using deep learning on bronchoscopy diagnosis images

Journal of Advanced Research, Journal Year: 2024, Volume and Issue: unknown

Published: Nov. 1, 2024

Language: Английский

---

Detection of Chest X-ray Abnormalities Using CNN Based on Hyperparameter Optimization

Published: Nov. 15, 2023

The chest X-ray (CXR) is a commonly used diagnostic imaging test that requires significant expertise and careful observation due to the complex nature of pathology fine texture lung lesions. Despite long-term clinical training professional guidance provided radiologists, there still possibility errors in diagnosis. Therefore, we have developed novel approach using convolutional neural network (CNN) model detect abnormalities CXR images. was optimized algorithms such as Adam RMSprop. Also, several hyperparameters were optimized, including pooling layer, dropout target size, epochs. Hyperparameter optimization aims improve model's accuracy by testing various combinations hyperparameter values algorithms. To evaluate performance, scenario modeling create 32 models tested them confusion matrix. results indicated best achieved 97.94%. This based on data 4538 findings suggest can CNN accurately identifying abnormalities. this study has important implications for improving reliability image interpretation, which could ultimately benefit patients detection treatment diseases. Acknowledging dataset constraints, address future steps improvement.

Language: Английский

---

Detection of TB from Chest X-ray: A Study with EfficientNet

2021 International Conference on System, Computation, Automation and Networking (ICSCAN), Journal Year: 2023, Volume and Issue: unknown

Published: Nov. 17, 2023

The lung is one of the prime organs, and any disease in causes mild to severe breathing problems; untreated will lead several complications. Tuberculosis (TB) a ailment that needs premature recognition handling. primary objective employ deep-learning (DL) based TB detection using chest $X$ -rays. Various stages proposed scheme consist (i) data collection resizing, (ii) DL-supported feature extraction, (iii) binary classification five-fold cross-validation, (iv) comparison with earlier results confirming merit scheme. This research implements EfficientNet (EN) variants classify chosen xmlns:xlink="http://www.w3.org/1999/xlink">$\mathrm{X}$ -rays into healthy/TB classes SoftMax classifier. EN_B2 (ENB2) has been successful providing an accuracy xmlns:xlink="http://www.w3.org/1999/xlink">$96{\% }$ as far considered when compared other methods. superiority suggested strategy also confirmed by analysis most recent technology, which confirms worth system on -ray imagery.

Language: Английский

---

Exploring Deep Learning Models for Accurate Alzheimer's Disease Classification based on MRI Imaging

EAI Endorsed Transactions on Pervasive Health and Technology, Journal Year: 2024, Volume and Issue: 10

Published: March 25, 2024

INTRODUCTION: Alzheimer's disease (AD), a complex neurodegenerative condition, presents significant challenges in early and accurate diagnosis. Early prediction of AD severity holds the potential for improved patient care timely interventions. This research investigates use deep learning methodologies to forecast utilizing data extracted from Magnetic Resonance Imaging (MRI) scans. OBJECTIVES: study aims explore efficacy models predicting using MRI data. Traditional diagnostic methods AD, primarily reliant on cognitive assessments, often lead late-stage detection. scans offer non-invasive means examine brain structure detect pathological changes associated with AD. However, manual interpretation these is labor-intensive subject variability. METHODS: Various models, including Convolutional Neural Networks (CNNs) advanced architectures like DenseNet, VGG16, ResNet50, MobileNet, AlexNet, Xception, are explored scan analysis. The performance assessed compared. Deep autonomously learn hierarchical features data, potentially recognizing intricate patterns different stages that may be overlooked analysis. RESULTS: evaluates scans. results highlight capturing subtle indicative progression. Moreover, comparison underscores strengths limitations each model, aiding selection appropriate prognosis. CONCLUSION: contributes growing field AI-driven healthcare by showcasing revolutionizing diagnosis prognosis. findings emphasize importance leveraging technologies, such as learning, enhance accuracy timeliness remain, need large annotated datasets, model interpretability, integration into clinical workflows. Continued efforts this area hold promise improving management ultimately enhancing outcomes.

Language: Английский

---