DeepBreastCancerNet: A Novel Deep Learning Model for Breast Cancer Detection Using Ultrasound Images

Applied Sciences, Год журнала: 2023, Номер 13(4), С. 2082 - 2082

Опубликована: Фев. 6, 2023

Breast cancer causes hundreds of women’s deaths each year. The manual detection breast is time-consuming, complicated, and prone to inaccuracy. For Cancer (BC) detection, several imaging methods are explored. However, sometimes misidentification leads unnecessary treatment diagnosis. Therefore, accurate BC can save many people from surgery biopsy. Due recent developments in the industry, deep learning’s (DL) performance processing medical images has significantly improved. Deep Learning techniques successfully identify ultrasound due their superior prediction ability. Transfer learning reuses knowledge representations public models built on large-scale datasets. problem overfitting. key idea this research propose an efficient robust deep-learning model for classification. paper presents a novel DeepBraestCancerNet DL proposed framework 24 layers, including six convolutional nine inception modules, one fully connected layer. Also, architecture uses clipped ReLu activation function, leaky batch normalization cross-channel as its two operations. We observed that reached highest classification accuracy 99.35%. also compared approach with existing models, experiment results showed outperformed state-of-the-art. Furthermore, we validated using another standard, publicaly available dataset. 99.63%.

Язык: Английский

---

A robust MRI-based brain tumor classification via a hybrid deep learning technique

The Journal of Supercomputing, Год журнала: 2023, Номер 80(2), С. 2403 - 2427

Опубликована: Авг. 8, 2023

Abstract The brain is the most vital component of neurological system. Therefore, tumor classification a very challenging task in field medical image analysis. There has been qualitative leap artificial intelligence, deep learning, and their imaging applications last decade. importance this remarkable development emerged biomedical engineering due to sensitivity seriousness issues related it. use learning detecting classifying tumors general particular using magnetic resonance (MRI) crucial factor accuracy speed diagnosis. This its great ability deal with huge amounts data avoid errors resulting from human intervention. aim research develop an efficient automated approach for assist radiologists instead consuming time looking at several images precise proposed based on 3064 T1-weighted contrast-enhanced MR (T1W-CE MRI) 233 patients. In study, system results five different models combined potential multiple models, trying achieve promising results. led significant improvement results, overall 99.31%.

Язык: Английский

---

YOLOv8-CAB: Improved YOLOv8 for Real-time object detection

Karbala International Journal of Modern Science, Год журнала: 2024, Номер 10(1)

Опубликована: Янв. 24, 2024

This study presents a groundbreaking approach to enhance the accuracy of YOLOv8 model in object detection, focusing mainly on addressing limitations detecting objects varied image types, particularly for small objects. The proposed strategy this work incorporates Context Attention Block (CAB) effectively locate and identify images. Furthermore, improves feature extraction capability without increasing complexity by thickness Coarse-to-Fine(C2F) block. In addition, Spatial (SA) has been modified accelerate detection performance. enhanced (Namely YOLOv8-CAB) strongly emphasizes performance smaller leveraging CAB block exploit multi-scale maps iterative feedback, thereby optimizing mechanisms. As result, innovative design facilitates superior extraction, “especially weak features,” contextual information preservation, efficient fusion. Rigorous testing Common Objects (COCO) dataset was performed demonstrate efficacy technique. It is resulting remarkable improvement over standard YOLO models. YOLOv8-CAB achieved mean average precision 97% rate, indicating 1% increase compared conventional highlights capabilities our improved method objects, representing breakthrough that sets stage advancements real-time techniques.

Язык: Английский

---

Hybrid framework for respiratory lung diseases detection based on classical CNN and quantum classifiers from chest X-rays

Biomedical Signal Processing and Control, Год журнала: 2023, Номер 88, С. 105567 - 105567

Опубликована: Окт. 18, 2023

Язык: Английский

---

Automated Uterine Fibroids Detection in Ultrasound Images Using Deep Convolutional Neural Networks

Healthcare, Год журнала: 2023, Номер 11(10), С. 1493 - 1493

Опубликована: Май 20, 2023

Fibroids of the uterus are a common benign tumor affecting women childbearing age. Uterine fibroids (UF) can be effectively treated with earlier identification and diagnosis. Its automated diagnosis from medical images is an area where deep learning (DL)-based algorithms have demonstrated promising results. In this research, we evaluated state-of-the-art DL architectures VGG16, ResNet50, InceptionV3, our proposed innovative dual-path convolutional neural network (DPCNN) architecture for UF detection tasks. Using preprocessing methods including scaling, normalization, data augmentation, ultrasound image dataset Kaggle prepared use. After used to train validate models, model performance using different measures. When compared existing suggested DPCNN achieved highest accuracy 99.8 percent. Findings show that pre-trained deep-learning may significantly improve application fine-tuning strategies. particular, InceptionV3 90% accuracy, ResNet50 achieving 89% accuracy. It should noted VGG16 was found lower level 85%. Our findings DL-based utilized facilitate images. Further research in holds great potential could lead creation cutting-edge computer-aided systems. To further advance imaging analysis, community invited investigate these lines research. Although performed best, fine-tuned versions models like also delivered strong This work lays foundation future studies has enhance precision suitability which detected.

Язык: Английский

---

Short- and Long-Term Chest-CT Findings after Recovery from COVID-19: A Systematic Review and Meta-Analysis

Diagnostics, Год журнала: 2024, Номер 14(6), С. 621 - 621

Опубликована: Март 14, 2024

While ground-glass opacity, consolidation, and fibrosis in the lungs are some of hallmarks acute SAR-CoV-2 infection, it remains unclear whether these pulmonary radiological findings would resolve after symptoms have subsided. We conducted a systematic review meta-analysis to evaluate chest computed tomography (CT) abnormalities stratified by COVID-19 disease severity multiple timepoints post-infection. PubMed/MEDLINE was searched for relevant articles until 23 May 2023. Studies with COVID-19-recovered patients follow-up CT at least 12 months post-infection were included. evaluated short-term (1–6 months) long-term (12–24 follow-ups (severe non-severe). A generalized linear mixed-effects model random effects used estimate event rates findings. total 2517 studies identified, which 43 met inclusion (N = 8858 patients). Fibrotic-like changes had highest rate (0.44 [0.3–0.59]) (0.38 [0.23–0.56]) follow-ups. meta-regression showed that over time decreased any abnormality (β −0.137, p 0.002), opacities −0.169, < 0.001), increased honeycombing 0.075, 0.03), did not change fibrotic-like changes, bronchiectasis, reticulation, interlobular septal thickening (p > 0.05 all). The severe subgroup significantly higher bronchiectasis 0.02), reticulation 0.001) when compared non-severe subgroup. In conclusion, significant remained up 2 years post-COVID-19, especially disease. Long-lasting post-SARS-CoV-2 infection signal future public health concern, necessitating extended monitoring, rehabilitation, survivor support, vaccination, ongoing research targeted therapies.

Язык: Английский

---

COVID-19 detection from chest CT images using optimized deep features and ensemble classification

Systems and Soft Computing, Год журнала: 2024, Номер 6, С. 200077 - 200077

Опубликована: Фев. 4, 2024

Diagnosis of COVID-19 positive patients is the eventual move to impede expansion coronavirus. Variations coronavirus make it tough recognize through symptoms. Hence, this research aims at a faster and automatic detection approach disease from chest Computed tomography (CT) scan images. For composition system, constructs feature vector CT images features fusion two Convolutional neural network (CNN) models namely VGG-19 ResNet-50. Before fusion, preprocessing techniques are applied gain more accurate outcomes. Moreover, pertinent identified by using several optimization methods Recursive elimination (RFE), Principal component analysis (PCA), Linear discriminant (LDA), among them, we have observed PCA as best preference. Classification performed on optimized utilizing Max voting ensemble classification (MVEC). The fused ResNet-50, processed with MVEC, provide outcomes accuracy, specificity, sensitivity, precision 98.51%, 97.58%, 99.49%, 97.47%, respectively, after 5-fold cross-validation for proposed method.

Язык: Английский

---

Enhancing explainability in brain tumor detection: A novel DeepEBTDNet model with LIME on MRI images

International Journal of Imaging Systems and Technology, Год журнала: 2023, Номер 34(1)

Опубликована: Дек. 22, 2023

Abstract Early detection of brain tumors is vital for improving patient survival rates, yet the manual analysis extensive 3D MRI images can be error‐prone and time‐consuming. This study introduces Deep Explainable Brain Tumor Network (DeepEBTDNet), a novel deep learning model binary classification MRIs as tumorous or normal. Employing sub‐image dualistic histogram equalization (DSIHE) enhanced image quality, DeepEBTDNet utilizes 12 convolutional layers with leaky ReLU (LReLU) activation feature extraction, followed by fully connected layer. Transparency interpretability are emphasized through application Local Interpretable Model‐Agnostic Explanations (LIME) method to explain predictions. Results demonstrate DeepEBTDNet's efficacy in tumor detection, even across datasets, achieving validation accuracy 98.96% testing 94.0%. underscores importance explainable AI healthcare, facilitating precise diagnoses transparent decision‐making early identification improved outcomes.

Язык: Английский

---

TumorDetNet: A unified deep learning model for brain tumor detection and classification

PLoS ONE, Год журнала: 2023, Номер 18(9), С. e0291200 - e0291200

Опубликована: Сен. 27, 2023

Accurate diagnosis of the brain tumor type at an earlier stage is crucial for treatment process and helps to save lives a large number people worldwide. Because they are non-invasive spare patients from having unpleasant biopsy, magnetic resonance imaging (MRI) scans frequently employed identify tumors. The manual identification tumors difficult requires considerable time due three-dimensional images that MRI scan one patient’s produces various angles. Moreover, variations in location, size, shape also make it challenging detect classify different types Thus, computer-aided diagnostics (CAD) systems have been proposed detection In this paper, we novel unified end-to-end deep learning model named TumorDetNet classification. Our framework employs 48 convolution layers with leaky ReLU (LReLU) activation functions compute most distinctive feature maps. average pooling dropout layer used learn patterns reduce overfitting. Finally, fully connected softmax into multiple types. We assessed performance our method on six standard Kaggle datasets classification (malignant benign), (glioma, pituitary, meningioma). successfully identified remarkable accuracy 99.83%, classified benign malignant ideal 100%, meningiomas, gliomas 99.27%. These outcomes demonstrate potency suggested methodology reliable categorization

Язык: Английский

---

An effective approach for plant leaf diseases classification based on a novel DeepPlantNet deep learning model

Frontiers in Plant Science, Год журнала: 2023, Номер 14

Опубликована: Окт. 11, 2023

Introduction Recently, plant disease detection and diagnosis procedures have become a primary agricultural concern. Early of diseases enables farmers to take preventative action, stopping the disease's transmission other sections. Plant are severe hazard food safety, but because essential infrastructure is missing in various places around globe, quick still difficult. The may experience variety attacks, from minor damage total devastation, depending on how infections are. Thus, early necessary optimize output prevent such destruction. physical examination produced low accuracy, required lot time, could not accurately anticipate disease. Creating an automated method capable classifying deal with these issues vital. Method This research proposes efficient, novel, lightweight DeepPlantNet deep learning (DL)-based architecture for predicting categorizing leaf diseases. proposed model comprises 28 learned layers, i.e., 25 convolutional layers (ConV) three fully connected (FC) layers. framework employed Leaky RelU (LReLU), batch normalization (BN), fire modules, mix 3×3 1×1 filters, making it novel classification framework. Proposed can categorize images into many classifications. Results approach categorizes following ten groups: Apple_Black_rot (ABR), Cherry_(including_sour)_Powdery_mildew (CPM), Grape_Leaf_blight_(Isariopsis_Leaf_Spot) (GLB), Peach_Bacterial_spot (PBS), Pepper_bell_Bacterial_spot (PBBS), Potato_Early_blight (PEB), Squash_Powdery_mildew (SPM), Strawberry_Leaf_scorch (SLS), bacterial tomato spot (TBS), maize common rust (MCR). achieved average accuracy 98.49 99.85in case eight-class three-class schemes, respectively. Discussion experimental findings demonstrated model's superiority alternatives. technique reduce financial losses by quickly effectively assisting professionals identifying

Язык: Английский

---