Generating Medical Reports With a Novel Deep Learning Architecture

International Journal of Imaging Systems and Technology, Год журнала: 2025, Номер 35(2)

Опубликована: Март 1, 2025

ABSTRACT The writing of medical reports by doctors in hospitals is a critical and sensitive process that time‐consuming, prone to human error, requires experts on site. Existing work autonomous report generation using images as input has not achieved sufficiently high success. goal this paper present new, fast, high‐performance method. For the paragraph‐level reports. A deep learning‐based hybrid encoder–decoder architecture called G‐CNX developed generate meaningful ConvNeXtBase used encoder side, GRU‐based RNN decoder side. Images from Indiana University Chest X‐ray ROCOv2 data sets were training, validation, testing processes study. results experiments showed autonomously generated had highest performance compared other studies literature. In set, success rates 0.6544, 0.5035, 0.3682, 0.2766, 0.4277 obtained Bleu‐1, Bleu‐2, Bleu‐3, Bleu‐4, Rouge evaluation metrics, respectively. scores 0.5593 0.3990 Bleu‐1 addition numerical quantifiable analysis, study also analyzed observationally based density plots. Statistical significance tests conducted prove reliability results. show test have semantic properties similar those written real produced are consistent reliable. proposed method can improve efficiency reporting, reduce workload specialized doctors, quality diagnosis treatment processes.

Язык: Английский

---

Multi-stage glaucoma classification using pre-trained convolutional neural networks and voting-based classifier fusion

Frontiers in Physiology, Год журнала: 2023, Номер 14

Опубликована: Июнь 13, 2023

Aim: To design an automated glaucoma detection system for early of using fundus images. Background: Glaucoma is a serious eye problem that can cause vision loss and even permanent blindness. Early prevention are crucial effective treatment. Traditional diagnostic approaches time consuming, manual, often inaccurate, thus making diagnosis necessary. Objective: propose stage classification model pre-trained deep convolutional neural network (CNN) models classifier fusion. Methods: The proposed utilized five CNN models: ResNet50, AlexNet, VGG19, DenseNet-201, Inception-ResNet-v2. was tested four public datasets: ACRIMA, RIM-ONE, Harvard Dataverse (HVD), Drishti. Classifier fusion created to merge the decisions all maximum voting-based approach. Results: achieved area under curve 1 accuracy 99.57% ACRIMA dataset. HVD dataset had 0.97 85.43%. rates Drishti RIM-ONE were 90.55 94.95%, respectively. experimental results showed performed better than state-of-the-art methods in classifying its stages. Understanding output includes both attribution-based such as activations gradient class activation map perturbation-based locally interpretable model-agnostic explanations occlusion sensitivity, which generate heatmaps various sections image prediction. Conclusion: method glaucoma. indicate high superior performance compared existing methods.

Язык: Английский

---

Chest X-ray Images for Lung Disease Detection Using Deep Learning Techniques: A Comprehensive Survey

Archives of Computational Methods in Engineering, Год журнала: 2024, Номер 31(6), С. 3267 - 3301

Опубликована: Фев. 19, 2024

Язык: Английский

---

Deep Learning for Pneumonia Detection in Chest X-ray Images: A Comprehensive Survey

Journal of Imaging, Год журнала: 2024, Номер 10(8), С. 176 - 176

Опубликована: Июль 23, 2024

This paper addresses the significant problem of identifying relevant background and contextual literature related to deep learning (DL) as an evolving technology in order provide a comprehensive analysis application DL specific pneumonia detection via chest X-ray (CXR) imaging, which is most common cost-effective imaging technique available worldwide for diagnosis. particular key period associated with COVID-19, 2020–2023, explain, analyze, systematically evaluate limitations approaches determine their relative levels effectiveness. The context applied both aid automated substitute existing expert radiography professionals, who often have limited availability, elaborated detail. rationale undertaken research provided, along justification resources adopted relevance. explanatory text subsequent analyses are intended sufficient detail being addressed, solutions, these, ranging from more general. Indeed, our evaluation agree generally held view that use transformers, specifically, vision transformers (ViTs), promising obtaining further effective results area using CXR images. However, ViTs require extensive address several limitations, specifically following: biased datasets, data code ease model can be explained, systematic methods accurate comparison, notion class imbalance possibility adversarial attacks, latter remains fundamental research.

Язык: Английский

---

ResNet-50 vs. EfficientNet-B0: Multi-Centric Classification of Various Lung Abnormalities Using Deep Learning "Session id: ICMLDsE.004"

Procedia Computer Science, Год журнала: 2024, Номер 235, С. 70 - 80

Опубликована: Янв. 1, 2024

Lung abnormalities are among the significant contributors to morbidity and mortality worldwide. It induces symptoms like coughing, sneezing, fever, breathlessness, etc., which, if left untreated, may lead death. In current clinical practice, chest X-ray (CXR) images widely preferred diagnose different lung abnormalities. However, pathological tests time-consuming, expensive require domain experts. On other hand, diagnosis through CXR is manual subject inter-observer intra-observer variability. The recent advancement in deep learning (DL) algorithms be employed address these challenges. selection of correct along with finetuned parameters challenging. this study, we comprehensively compared performance two state-of-the-art DL algorithms, Resnet-50 Efficient-B0. These models pervasively used literature have shown promising classification performance. validated using multi-centric dataset from Kaggle (having Covid-19, Normal, Pneumonia classes) Mendeley Pneumonia-Bacterial, Pneumonia-Viral, Covid-19 classes). Upon training a mini-batch size 32 maximum epoch 40 data, achieved 0.9807 0.9874 accuracy for Dataset Dataset, respectively. Similarly, 0.9962 9.9978 test From result, it evident that EfficientNet-B0 model outperformed datasets.

Язык: Английский

---

ZooCNN: A Zero-Order Optimized Convolutional Neural Network for Pneumonia Classification Using Chest Radiographs

Journal of Imaging, Год журнала: 2025, Номер 11(1), С. 22 - 22

Опубликована: Янв. 13, 2025

Pneumonia, a leading cause of mortality in children under five, is usually diagnosed through chest X-ray (CXR) images due to its efficiency and cost-effectiveness. However, the shortage radiologists Least Developed Countries (LDCs) emphasizes need for automated pneumonia diagnostic systems. This article presents Deep Learning model, Zero-Order Optimized Convolutional Neural Network (ZooCNN), Optimization (Zoo)-based CNN model classifying CXR into three classes, Normal Lungs (NL), Bacterial Pneumonia (BP), Viral (VP); this utilizes Adaptive Synthetic Sampling (ADASYN) approach ensure class balance Kaggle Images (Pneumonia) dataset. Conventional models, though promising, face challenges such as overfitting have high computational costs. The use ZooPlatform (ZooPT), hyperparameter finetuning strategy, on baseline finetunes hyperparameters provides modified architecture, ZooCNN, with 72% reduction weights. was trained, tested, validated ZooCNN achieved an accuracy 97.27%, sensitivity 97.00%, specificity 98.60%, F1 score 97.03%. results were compared contemporary models highlight efficacy classification (PC), offering potential tool aid physicians clinical settings.

Язык: Английский

---

Dung Beetle Optimization with Deep Feature Fusion Model for Lung Cancer Detection and Classification

Cancers, Год журнала: 2023, Номер 15(15), С. 3982 - 3982

Опубликована: Авг. 5, 2023

Lung cancer is the main cause of deaths all over world. An important reason for these was late analysis and worse prediction. With accelerated improvement deep learning (DL) approaches, DL can be effectively widely executed several real-world applications in healthcare systems, like medical image interpretation disease analysis. Medical imaging devices vital primary-stage lung tumor observation tumors from treatment. Many modalities computed tomography (CT), chest X-ray (CXR), molecular imaging, magnetic resonance (MRI), positron emission (PET) systems are analyzed detection. This article presents a new dung beetle optimization modified feature fusion model detection classification (DBOMDFF-LCC) technique. The presented DBOMDFF-LCC technique mainly depends upon hyperparameter tuning process. To accomplish this, uses process comprising three models, namely residual network (ResNet), densely connected (DenseNet), Inception-ResNet-v2. Furthermore, DBO approach employed optimum selection approaches. For purposes, system utilizes long short-term memory (LSTM) approach. simulation result dataset investigated using different evaluation metrics. extensive comparative results highlighted betterment classification.

Язык: Английский

---

Lung-DT: An AI-Powered Digital Twin Framework for Thoracic Health Monitoring and Diagnosis

Sensors, Год журнала: 2024, Номер 24(3), С. 958 - 958

Опубликована: Фев. 1, 2024

The integration of artificial intelligence (AI) with Digital Twins (DTs) has emerged as a promising approach to revolutionize healthcare, particularly in terms diagnosis and management thoracic disorders. This study proposes comprehensive framework, named Lung-DT, which leverages IoT sensors AI algorithms establish the digital representation patient’s respiratory health. Using YOLOv8 neural network, Lung-DT system accurately classifies chest X-rays into five distinct categories lung diseases, including “normal”, “covid”, “lung_opacity”, “pneumonia”, “tuberculosis”. performance was evaluated employing X-ray dataset available literature, demonstrating average accuracy 96.8%, precision 92%, recall 97%, F1-score 94%. proposed framework offers several advantages over conventional diagnostic methods. Firstly, it enables real-time monitoring health through continuous data acquisition from sensors, facilitating early intervention. Secondly, AI-powered classification module provides automated objective assessments X-rays, reducing dependence on subjective human interpretation. Thirdly, twin allows for analysis correlation multiple streams, providing valuable insights personalized treatment plans. algorithms, DT technology within demonstrates significant step towards improving healthcare. By enabling monitoring, diagnosis, analysis, enormous potential enhance patient outcomes, reduce healthcare costs, optimize resource allocation.

Язык: Английский

---

Glaucoma detection with explainable AI using convolutional neural networks based feature extraction and machine learning classifiers

IET Image Processing, Год журнала: 2024, Номер 18(13), С. 3827 - 3853

Опубликована: Авг. 19, 2024

Abstract Glaucoma is an eye disease that damages the optic nerve as a result of vision loss, it leading cause blindness worldwide. Due to time‐consuming, inaccurate, and manual nature traditional methods, automation in glaucoma detection important. This paper proposes explainable artificial intelligence (XAI) based model for automatic using pre‐trained convolutional neural networks (PCNNs) machine learning classifiers (MLCs). PCNNs are used feature extractors obtain deep features can capture important visual patterns characteristics from fundus images. Using extracted MLCs then classify healthy An empirical selection CNN MLC parameters has been made performance evaluation. In this work, total 1,865 1,590 images different datasets were used. The results on ACRIMA dataset show accuracy, precision, recall 98.03%, 97.61%, 99%, respectively. Explainable aims create increase user's trust model's decision‐making process transparent interpretable manner. assessment image misclassification carried out facilitate future investigations.

Язык: Английский

---

DeepLungNet: An Effective DL-Based Approach for Lung Disease Classification Using CRIs

Electronics, Год журнала: 2023, Номер 12(8), С. 1860 - 1860

Опубликована: Апрель 14, 2023

Infectious disease-related illness has always posed a concern on global scale. Each year, pneumonia (viral and bacterial pneumonia), tuberculosis (TB), COVID-19, lung opacity (LO) cause millions of deaths because they all affect the lungs. Early detection diagnosis can help create chances for better care in circumstances. Numerous tests, including molecular tests (RT-PCR), complete blood count (CBC) Monteux tuberculin skin (TST), ultrasounds, are used to detect classify these diseases. However, take lot time, have 20% mistake rate, 80% sensitive. So, with aid doctor, radiographic such as computed tomography (CT) chest radiograph images (CRIs) disorders. With CRIs or CT-scan images, there is danger that features various diseases’ diagnoses will overlap. The automation method necessary correctly diseases using CRIs. key motivation behind study was no identifying classifying (LO, pneumonia, VP, BP, TB, COVID-19) In this paper, DeepLungNet deep learning (DL) model proposed, which comprises 20 learnable layers, i.e., 18 convolution (ConV) layers 2 fully connected (FC) layers. architecture uses Leaky ReLU (LReLU) activation function, fire module, maximum pooling layer, shortcut connections, batch normalization (BN) operation, group making it novel classification framework. This useful DL-based disorders, we tested effectiveness suggested framework two datasets variety from different datasets. We performed experiments: five-class (TB, LO, normal) six-class (VP, normal, LO). framework’s average accuracy into normal an impressive 97.47%. verified performance our publicly accessible database agriculture sector order further assess its validate generalizability. offers efficient automated aids early disease. strategy significantly improves patient survival, possible treatments, limits transmission infectious illnesses throughout society.

Язык: Английский

---