A Machine Learning-based Method for COVID-19 and Pneumonia Detection

IgMin Research, Год журнала: 2024, Номер 2(7), С. 518 - 523

Опубликована: Июль 5, 2024

Pneumonia is described as an acute infection of lung tissue produced by one or more bacteria, and Coronavirus Disease (COVID-19) a deadly virus that affects the lungs human body. The symptoms COVID-19 disease are closely related to pneumonia. In this work, we identify patients pneumonia coronavirus from chest X-ray images. We used convolutional neural network for spatial feature learning experimented with images in Kaggle dataset. corona classified using feed-forward hybrid models (CNN+SVM, CNN+RF, CNN+Xgboost). experimental findings on dataset demonstrate CNN detects 99.47% recall. overall experiments x-ray show detected 95.45% accuracy.

Язык: Английский

---

Attention-enhanced dilated convolution for Parkinson’s disease detection using transcranial sonography

BioMedical Engineering OnLine, Год журнала: 2024, Номер 23(1)

Опубликована: Июль 31, 2024

Abstract Background Transcranial sonography (TCS) plays a crucial role in diagnosing Parkinson's disease. However, the intricate nature of TCS pathological features, lack consistent diagnostic criteria, and dependence on physicians' expertise can hinder accurate diagnosis. Current TCS-based methods, which rely machine learning, often involve complex feature engineering may struggle to capture deep image features. While learning offers advantages processing, it has not been tailored address specific movement disorder considerations. Consequently, there is scarcity research algorithms for PD Methods This study introduces residual network model, augmented with attention mechanisms multi-scale extraction, termed AMSNet, assist Initially, extraction module implemented robustly handle irregular morphological features significant area information present images. effectively mitigates effects artifacts noise. When combined convolutional module, enhances model's ability learn lesion areas. Subsequently, architecture, integrated channel attention, utilized hierarchical detailed textures within images, further enhancing representation capabilities. Results The compiled images personal data from 1109 participants. Experiments conducted this dataset demonstrated that AMSNet achieved remarkable classification accuracy (92.79%), precision (95.42%), specificity (93.1%). It surpassed performance previously employed domain, as well current general-purpose models. Conclusion proposed deviates traditional approaches necessitate engineering. capable automatically extracting capacity comprehend articulate data. underscores substantial potential methods application diagnosis disorders. Graphical

Язык: Английский

---

Challenges issues and future recommendations of deep learning techniques for SARS-CoV-2 detection utilising X-ray and CT images: a comprehensive review

PeerJ Computer Science, Год журнала: 2024, Номер 10, С. e2517 - e2517

Опубликована: Дек. 24, 2024

The global spread of SARS-CoV-2 has prompted a crucial need for accurate medical diagnosis, particularly in the respiratory system. Current diagnostic methods heavily rely on imaging techniques like CT scans and X-rays, but identifying these images proves to be challenging time-consuming. In this context, artificial intelligence (AI) models, specifically deep learning (DL) networks, emerge as promising solution image analysis. This article provides meticulous comprehensive review imaging-based diagnosis using up May 2024. starts with an overview covering basic steps learning-based data sources, pre-processing methods, taxonomy techniques, findings, research gaps performance evaluation. We also focus addressing current privacy issues, limitations, challenges realm diagnosis. According taxonomy, each model is discussed, encompassing its core functionality critical assessment suitability detection. A comparative analysis included by summarizing all relevant studies provide overall visualization. Considering best deep-learning detection, conducts experiment twelve contemporary techniques. experimental result shows that MobileNetV3 outperforms other models accuracy 98.11%. Finally, elaborates explores potential future directions methodological recommendations advancement.

Язык: Английский

---

Brain CT image classification based on mask RCNN and attention mechanism

Scientific Reports, Год журнала: 2024, Номер 14(1)

Опубликована: Ноя. 26, 2024

Язык: Английский

---

Perturbing BEAMs: EEG adversarial attack to deep learning models for epilepsy diagnosing

BMC Medical Informatics and Decision Making, Год журнала: 2023, Номер 23(1)

Опубликована: Июль 6, 2023

Abstract Deep learning models have been widely used in electroencephalogram (EEG) analysis and obtained excellent performance. But the adversarial attack defense for them should be thoroughly studied before putting into safety-sensitive use. This work exposes an important safety issue deep-learning-based brain disease diagnostic systems by examining vulnerability of deep diagnosing epilepsy with electrical activity mappings (BEAMs) to white-box attacks. It proposes two methods, Gradient Perturbations BEAMs (GPBEAM), Differential Evolution (GPBEAM-DE), which generate EEG samples, first time perturbing densely sparsely respectively, find that these BEAMs-based samples can easily mislead models. The experiments use data from CHB-MIT dataset types victim each has four different neural network (DNN) architectures. is shown that: (1) BEAM-based produced proposed methods this paper are aggressive BEAM-related as input internal DNN architectures, but unaggressive EEG-related raw top success rate attacking up 0.8 while only 0.01; (2) GPBEAM-DE outperforms GPBEAM when they same model under a distortion constraint, former 0.59 latter; (3) simple modification GPBEAM/GPBEAM-DE will make it aggressiveness both BEAMs-related (with 0.64), capacity enhancement done without any cost increment. goal study not medical systems, raise concerns about hope lead safer design.

Язык: Английский

---