Deep learning based classification of multi-label chest X-ray images via dual-weighted metric loss

Computers in Biology and Medicine, Journal Year: 2023, Volume and Issue: 157, P. 106683 - 106683

Published: Feb. 15, 2023

Language: Английский

---

Classification of Lung Diseases Using an Attention-Based Modified DenseNet Model

Deleted Journal, Journal Year: 2024, Volume and Issue: 37(4), P. 1625 - 1641

Published: March 11, 2024

Lung diseases represent a significant global health threat, impacting both well-being and mortality rates. Diagnostic procedures such as Computed Tomography (CT) scans X-ray imaging play pivotal role in identifying these conditions. X-rays, due to their easy accessibility affordability, serve convenient cost-effective option for diagnosing lung diseases. Our proposed method utilized the Contrast-Limited Adaptive Histogram Equalization (CLAHE) enhancement technique on images highlight key feature maps related using DenseNet201. We have augmented existing Densenet201 model with hybrid pooling channel attention mechanism. The experimental results demonstrate superiority of our over well-known pre-trained models, VGG16, VGG19, InceptionV3, Xception, ResNet50, ResNet152, ResNet50V2, ResNet152V2, MobileNetV2, DenseNet121, DenseNet169, achieves impressive accuracy, precision, recall, F1-scores 95.34%, 97%, 96%, respectively. also provide visual insights into model's decision-making process Gradient-weighted Class Activation Mapping (Grad-CAM) identify normal, pneumothorax, atelectasis cases. terms heatmap may help radiologists improve diagnostic abilities labelling processes.

Language: Английский

---

Lung Disease Classification in CXR Images Using Hybrid Inception-ResNet-v2 Model and Edge Computing

Journal of Healthcare Engineering, Journal Year: 2022, Volume and Issue: 2022, P. 1 - 15

Published: March 30, 2022

Chest X-ray (CXR) imaging is one of the most widely used and economical tests to diagnose a wide range diseases. However, even for expert radiologists, it challenge accurately diseases from CXR samples. Furthermore, there remains an acute shortage trained radiologists worldwide. In present study, machine learning (ML), deep (DL), transfer (TL) approaches have been evaluated classify in openly available image dataset. A combination synthetic minority over-sampling technique (SMOTE) weighted class balancing alleviate effects imbalance. hybrid Inception-ResNet-v2 model coupled with data augmentation enhancement gives best accuracy. The deployed edge environment using Amazon IoT Core automate task disease detection images three categories, namely pneumonia, COVID-19, normal. Comparative analysis has given various metrics such as precision, recall, accuracy, AUC-ROC score, etc. proposed average accuracy 98.66%. accuracies other TL models, SqueezeNet, VGG19, ResNet50, MobileNetV2 are 97.33%, 91.66%, 90.33%, 76.00%, respectively. Further, DL model, scratch, 92.43%. Two feature-based ML classification techniques, support vector local binary pattern (SVM + LBP) decision tree histogram oriented gradients (DT HOG) yield 87.98% 86.87%,

Language: Английский

---

Computer image analysis with artificial intelligence: a practical introduction to convolutional neural networks for medical professionals

Postgraduate Medical Journal, Journal Year: 2023, Volume and Issue: 99(1178), P. 1287 - 1294

Published: Oct. 4, 2023

Abstract Artificial intelligence tools, particularly convolutional neural networks (CNNs), are transforming healthcare by enhancing predictive, diagnostic, and decision-making capabilities. This review provides an accessible practical explanation of CNNs for clinicians highlights their relevance in medical image analysis. have shown themselves to be exceptionally useful computer vision, a field that enables machines ‘see’ interpret visual data. Understanding how these models work can help leverage full potential, especially as artificial continues evolve integrate into healthcare. already demonstrated efficacy diverse fields, including radiology, histopathology, photography. In been used automate the assessment conditions such pneumonia, pulmonary embolism, rectal cancer. assess classify colorectal polyps, gastric epithelial tumours, well assist multiple malignancies. photography, retinal diseases skin conditions, detect polyps during endoscopic procedures. surgical laparoscopy, they may provide intraoperative assistance surgeons, helping anatomy demonstrate safe dissection zones. The integration analysis promises enhance diagnostic accuracy, streamline workflow efficiency, expand access expert-level analysis, contributing ultimate goal delivering further improvements patient outcomes.

Language: Английский

---

Deep learning for understanding multilabel imbalanced Chest X-ray datasets

Future Generation Computer Systems, Journal Year: 2023, Volume and Issue: 144, P. 291 - 306

Published: March 6, 2023

Over the last few years, convolutional neural networks (CNNs) have dominated field of computer vision thanks to their ability extract features and outstanding performance in classification problems, for example automatic analysis X-rays. Unfortunately, these are considered black-box algorithms, i.e. it is impossible understand how algorithm has achieved final result. To apply algorithms different fields test methodology works, we need use eXplainable AI techniques. Most work medical focuses on binary or multiclass problems. However, many real-life situations, such as chest X-rays, radiological signs diseases can appear at same time. This gives rise what known "multilabel problems". A disadvantage tasks class imbalance, labels do not number samples. The main contribution this paper a Deep Learning imbalanced, multilabel X-ray datasets. It establishes baseline currently underutilised PadChest dataset new technique based heatmaps. also includes probabilities inter-model matching. results our system promising, especially considering used. Furthermore, heatmaps match expected areas, they mark areas that an expert would make decision.

Language: Английский

---

A quantization assisted U-Net study with ICA and deep features fusion for breast cancer identification using ultrasonic data

PeerJ Computer Science, Journal Year: 2021, Volume and Issue: 7, P. e805 - e805

Published: Dec. 16, 2021

Breast cancer is one of the leading causes death in women worldwide-the rapid increase breast has brought about more accessible diagnosis resources. The ultrasonic modality for relatively cost-effective and valuable. Lesion isolation images a challenging task due to its robustness intensity similarity. Accurate detection lesions using can reduce rates. In this research, quantization-assisted U-Net approach segmentation proposed. It contains two step segmentation: (1) (2) quantization. quantization assists U-Net-based order isolate exact lesion areas from sonography images. Independent Component Analysis (ICA) method then uses isolated extract features are fused with deep automatic features. Public ultrasonic-modality-based datasets such as Ultrasound Images Dataset (BUSI) Open Access Database Raw Ultrasonic Signals (OASBUD) used evaluation comparison. OASBUD data extracted same However, classification was done after feature regularization lasso method. obtained results allow us propose computer-aided design (CAD) system identification modalities.

Language: Английский

---

Lung Infection and Identification using Heatmap

Published: May 4, 2023

Lung disease identification using heatmap is an automated diagnosis system that utilizes the visualization of heatmaps to identify and classify lung diseases from chest X- Radiation images. The applies a deep learning-based approach automatically extract learn discriminative features input images, which are then used generate highlighting regions affected by disease. provide intuitive disease, can be aid radiologists in making accurate diagnoses. has potential increase efficiency accuracy clinical been proven achieve high categorization variety infection, including pneumonia Novel coronavirus. have become major health concern worldwide, causing significant morbidity mortality. Early timely treatment these significantly improve patient outcomes. This research paper, proposes novel analysis. CXR patients was collected with various images were pre-processed enhance reduce noise. A analysis technique applied highlight most learning model heatmaps. pictures categorized into several types infection groups convolutional neural network (CNN). CNN obtained good illness classification after being trained on huge dataset CXR. proposed evaluated 317 findings indicated our method classified overall 98.55%. suggested may precision diagnosing diseases. help clinicians planning. Furthermore, large datasets its robustness.

Language: Английский

---

Deep learning classification of active tuberculosis lung zones wise manifestations using chest X-rays: a multi label approach

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

Published: Jan. 17, 2023

Abstract Chest X-rays are the most economically viable diagnostic imaging test for active pulmonary tuberculosis screening despite high sensitivity and low specificity when interpreted by clinicians or radiologists. Computer aided detection (CAD) algorithms, especially convolution based deep learning architecture, have been proposed to facilitate automation of radiography modalities. Deep algorithms found success in classifying various abnormalities lung using chest X-ray. We fine-tuned, validated tested EfficientNetB4 architecture utilized transfer methodology multilabel approach detect zone wise image manifestations used Area Under Receiver Operating Characteristic (AUC), along with 95% confidence interval as model evaluation metrics. also visualisation capabilities convolutional neural networks (CNN), Gradient-weighted Class Activation Mapping (Grad-CAM) post-hoc attention method investigate Tuberculosis discuss them from radiological perspectives. trained network achieved remarkable AUC, intramural set external different geographical region. The grad-CAM visualisations their ability localize can aid at primary care settings triaging where resources constrained overburdened.

Language: Английский

---

A novel feature fusion based deep learning framework for white blood cell classification

Journal of Ambient Intelligence and Humanized Computing, Journal Year: 2022, Volume and Issue: 14(8), P. 9839 - 9851

Published: Jan. 15, 2022

Language: Английский

---

Performance Evaluation of the Deep Learning Based Convolutional Neural Network Approach for the Recognition of Chest X-Ray Images

Frontiers in Oncology, Journal Year: 2022, Volume and Issue: 12

Published: June 29, 2022

Recent advancement in the field of deep learning has provided promising performance for analysis medical images. Every year, pneumonia is leading cause death various children under age 5 years. Chest X-rays are first technique that used detection pneumonia. Various and computer vision techniques can be to determine virus which causes using X-ray These days, it possible use Convolutional Neural Networks (CNN) classification images due availability a large number datasets. In this work, CNN model implemented recognition Pneumonia. The trained on publicly available dataset having two classes: Normal chest Pneumonic images, where each class 5000 Samples. 80% collected data purpose train model, rest testing model. validated optimizers: Adam RMSprop. maximum accuracy 98% obtained validation dataset. results further compared with by other researchers biomedical

Language: Английский

---