A deep convolutional neural network model for medical data classification from computed tomography images DOI Open Access

S. Sreelakshmi,

V. S. Anoop

Expert Systems, Год журнала: 2023, Номер unknown

Опубликована: Авг. 21, 2023

Abstract Machine learning provides powerful techniques for several applications, including automated disease diagnosis through medical image classification. Recently, many studies reported that deep approaches have demonstrated significant performance and accuracy improvements over shallow techniques. The been used in problems related to diagnoses, such as thyroid diagnosis, diabetic retinopathy detection, foetal localization, breast cancer detection. Many methods the recent past uses images from various sources, healthcare providers open data initiatives, improvement terms of precision, recall, accuracy. This paper proposes a framework incorporating convolutional neural networks an enhanced feature extraction technique classifying data. To show real‐world usability proposed approach, it has classification COVID‐19 computed tomography scans. experimental results approach outperformed some chosen baselines obtained 98.91%, comparable with already accuracies.

Язык: Английский

Advancing Pulmonary Infection Diagnosis: A Comprehensive Review of Deep Learning Approaches in Radiological Data Analysis DOI
Sapna Yadav, Syed A. Rizvi, Pankaj Agarwal

и другие.

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

Опубликована: Март 13, 2025

Язык: Английский

Процитировано

1
Fusion of edge detection and graph neural networks to classifying electrocardiogram signals DOI Creative Commons
Linh T. Duong, Doan Thi Hoai Thu,

Cong Q. Chu

и другие.

Expert Systems with Applications, Год журнала: 2023, Номер 225, С. 120107 - 120107

Опубликована: Апрель 15, 2023

The analysis of electrocardiogram (ECG) signals are among the key factors in diagnosis cardiovascular diseases (CVDs). However, automatic processing ECG clinical practice is still restrained by accuracy existing algorithms. Deep learning methods have recently achieved striking success a variety task including predictive healthcare. Graph neural networks class machine algorithms which can learn directly extracting important information from graph-structured data, and perform prediction on unknown data. Such suitable for mining complex graph deducing useful predictions. In this work, we present Neural Network (GNN) model trained two datasets with more than 107,000 single-lead signal images extracted laboratories Boston's Beth Israel Hospital Massachusetts Institute Technology (MITBIH), 1.5 million labeled exams analyzed Physikalisch-Technische Bundesanstalt (PTB). Our proposed GNN achieves promising performance, i.e., results show that classification based GNNs using either or 12-lead setup closer to human-level standard practice. By several testing instances, approach obtains an 1.0, thereby outperforming various state-of-the-art baselines both databases respect effectiveness timing efficiency. We anticipate be deployed as non-invasive pre-screening tool assist doctors real-time monitoring performing their activities.

Язык: Английский

Процитировано

16
Emb-trattunet: a novel edge loss function and transformer-CNN architecture for multi-classes pneumonia infection segmentation in low annotation regimes DOI Creative Commons
Fares Bougourzi, Fadi Dornaika, Amir Nakib

и другие.

Artificial Intelligence Review, Год журнала: 2024, Номер 57(4)

Опубликована: Март 14, 2024

Abstract One of the primary challenges in applying deep learning approaches to medical imaging is limited availability data due various factors. These factors include concerns about privacy and requirement for expert radiologists perform time-consuming labor-intensive task labeling data, particularly tasks such as segmentation. Consequently, there a critical need develop novel few-shot this domain. In work, we propose Novel CNN-Transformer Fusion scheme segment Multi-classes pneumonia infection from CT-scans data. total, are three main contributions: (i) encoders fusion, which allows extract fuse richer features encoding phase, contains: local, global long-range dependencies features, (ii) Multi-Branches Skip Connection (MBSC) proposed encoder then integrate them into decoder layers, where MBSC blocks higher-level related finer details different types, (iii) Boundary Aware Cross-Entropy (MBA-CE) Loss function deal with fuzzy boundaries, enhance separability between classes give more attention minority classes. The performance approach evaluated using two evaluation scenarios compared baseline state-of-the-art segmentation architectures Covid-19 obtained results show that our outperforms comparison methods both Ground-Glass Opacity (GGO) Consolidation On other hand, shows consistent when training reduced half, proves efficiency learning. contrast, drops scenario. Moreover, able imbalanced advantages prove effectiveness EMB-TrAttUnet pandemic scenario time save patient lives.

Язык: Английский

Процитировано

6
Ensemble Federated Learning: An approach for collaborative pneumonia diagnosis DOI Creative Commons
Alhassan Mabrouk, Rebeca P. Dı́az Redondo, Mohamed Abd Elaziz

и другие.

Applied Soft Computing, Год журнала: 2023, Номер 144, С. 110500 - 110500

Опубликована: Июнь 19, 2023

Federated learning is a very convenient approach for scenarios where (i) the exchange of data implies privacy concerns and/or (ii) quick reaction needed. In smart healthcare systems, both aspects are usually required. this paper, we work on first scenario, preserving key and, consequently, building unique and massive medical image set by fusing different sets from institutions or research centers (computation nodes) not an option. We propose ensemble federated (EFL) that based following characteristics: First, each computation node works with (but same type). They locally apply combining eight well-known CNN models (densenet169, mobilenetv2, xception, inceptionv3, vgg16, resnet50, densenet121, resnet152v2) Chest X-ray images. Second, best two local used to create model shared central node. Third, aggregated obtain global model, which nodes continue new iteration. This procedure continues until there no changes in models. have performed experiments compare our centralized ones (with without approach)\color{black}. The results conclude proposal outperforms these images (achieving accuracy 96.63\%) offers competitive compared other proposals literature.

Язык: Английский

Процитировано

13
COVID-19 Image Classification: A Comparative Performance Analysis of Hand-Crafted vs. Deep Features DOI Creative Commons
Sadiq Alinsaif

Computation, Год журнала: 2024, Номер 12(4), С. 66 - 66

Опубликована: Март 30, 2024

This study investigates techniques for medical image classification, specifically focusing on COVID-19 scans obtained through computer tomography (CT). Firstly, handcrafted methods based feature engineering are explored due to their suitability training traditional machine learning (TML) classifiers (e.g., Support Vector Machine (SVM)) when faced with limited datasets. In this context, I comprehensively evaluate and compare 27 descriptor sets. More recently, deep (DL) models have successfully analyzed classified natural images. However, the scarcity of well-annotated images, particularly those related COVID-19, presents challenges DL from scratch. Consequently, leverage features extracted 12 pre-trained classification tasks. work a comprehensive comparative analysis between TML approaches in classification.

Язык: Английский

Процитировано

4
A tree-based explainable AI model for early detection of Covid-19 using physiological data DOI Creative Commons

Manar Abu Talib,

Yaman Afadar,

Qassim Nasir

и другие.

BMC Medical Informatics and Decision Making, Год журнала: 2024, Номер 24(1)

Опубликована: Июнь 24, 2024

Abstract With the outbreak of COVID-19 in 2020, countries worldwide faced significant concerns and challenges. Various studies have emerged utilizing Artificial Intelligence (AI) Data Science techniques for disease detection. Although cases declined, there are still deaths around world. Therefore, early detection before onset symptoms has become crucial reducing its extensive impact. Fortunately, wearable devices such as smartwatches proven to be valuable sources physiological data, including Heart Rate (HR) sleep quality, enabling inflammatory diseases. In this study, we utilize an already-existing dataset that includes individual step counts heart rate data predict probability infection symptoms. We train three main model architectures: Gradient Boosting classifier (GB), CatBoost trees, TabNet analyze compare their respective performances. also add interpretability layer our best-performing model, which clarifies prediction results allows a detailed assessment effectiveness. Moreover, created private by gathering from Fitbit guarantee reliability avoid bias. The identical set models was then applied using same pre-trained models, were documented. Using tree-based method, outperformed previous with accuracy 85% on publicly available dataset. Furthermore, produced 81% when You will find source code link: https://github.com/OpenUAE-LAB/Covid-19-detection-using-Wearable-data.git .

Язык: Английский

Процитировано

4
Identification and Diagnosis of Injuries to the Long Head of the Biceps Tendon-Superior Labrum Complex Using Convolutional Neural Networks in Artificial Intelligence: A Validation Study DOI
Zhewei Zhang, Xiang Li

Опубликована: Янв. 1, 2025

Download This Paper Open PDF in Browser Add to My Library Share: Permalink Using these links will ensure access this page indefinitely Copy URL DOI

Язык: Английский

Процитировано

0
Automated Visual Insight via AI-Based Processing DOI Open Access

Renuka Deshpande,

Pratik Golatkar,

Aniket Lohkare

и другие.

International Journal of Advanced Research in Science Communication and Technology, Год журнала: 2025, Номер unknown, С. 689 - 698

Опубликована: Апрель 14, 2025

Artificial Intelligence (AI) has revolutionized image processing by enhancing automation, accuracy, and efficiency across various domains such as medical diagnostics, autonomous vehicles, security, agriculture, entertainment. Traditional techniques relied on rule-based algorithms, which had limitations in complex scenarios. AI-powered leverages deep learning models,neural networks, computer vision to analyze manipulate images with human-like intelligence. This paper provides an in-depth analysis of AI-driven processing, covering fundamental techniques, applications, challenges, future trends

Язык: Английский

Процитировано

0
Multiple lung diseases detection using advanced deep learning model with attention mechanisms and upsampling features DOI

Jie Zhu,

Mohammed A. A. Al‐qaness, Dalal AL-Alimi

и другие.

Engineering Applications of Artificial Intelligence, Год журнала: 2025, Номер 156, С. 111038 - 111038

Опубликована: Май 17, 2025

Язык: Английский

Процитировано

0
Automatic detection of COVID-19 and pneumonia from chest X-ray images using texture features DOI
Farnaz Sheikhi,

Aliakbar Taghdiri,

Danial Moradisabzevar

и другие.

The Journal of Supercomputing, Год журнала: 2023, Номер 79(18), С. 21449 - 21473

Опубликована: Июнь 21, 2023

Язык: Английский

Процитировано

6