Application of explainable artificial intelligence for healthcare: A systematic review of the last decade (2011–2022)

Computer Methods and Programs in Biomedicine, Journal Year: 2022, Volume and Issue: 226, P. 107161 - 107161

Published: Sept. 27, 2022

Language: Английский

---

History of Artificial Intelligence

Published: Jan. 1, 2024

Language: Английский

---

A review of advancements of artificial intelligence in dentistry

Dentistry Review, Journal Year: 2024, Volume and Issue: 4(2), P. 100081 - 100081

Published: March 13, 2024

Artificial intelligence (AI) has been used in healthcare for decades and the potential to revolutionize dentistry by solving multiple clinical problems making work of clinicians easier. In particular, study AI applications periodontal disease cariology is important because these are two major areas concern dental health. Periodontal disease, which affects gums bone surrounding teeth, a cause tooth loss adults. Cariology, decay, also an area focus research. algorithms can be analyze images detect early signs decay that may missed human dentists. The review first discusses history then highlights some ways technology improved describe basic models such as artificial neural networks (ANNs), convolutional (CNNs), random forest. article delves into how involved cariology, endodontics, prosthodontics, orthodontics including classifying different types identifying loss, determining severity analyzing images, detecting diseases. On other hand, application relatively uncommon implementing technologies presents several challenges need addressed successful implementation dentistry.

Language: Английский

---

A Hybrid Deep Transfer Learning of CNN-Based LR-PCA for Breast Lesion Diagnosis via Medical Breast Mammograms

Sensors, Journal Year: 2022, Volume and Issue: 22(13), P. 4938 - 4938

Published: June 30, 2022

One of the most promising research areas in healthcare industry and scientific community is focusing on AI-based applications for real medical challenges such as building computer-aided diagnosis (CAD) systems breast cancer. Transfer learning one recent emerging techniques that allow rapid progress improve imaging performance. Although deep classification cancer has been widely covered, certain obstacles still remain to investigate independency among extracted high-level features. This work tackles two exist when designing effective CAD lesion from mammograms. The first challenge enrich input information models by generating pseudo-colored images instead only using original grayscale images. To achieve this goal different image preprocessing are parallel used: contrast-limited adaptive histogram equalization (CLAHE) Pixel-wise intensity adjustment. preserved channel, while other channels receive processed images, respectively. generated three-channel fed directly into layer backbone CNNs generate more powerful second overcome multicollinearity problem occurs high correlated features models. A new hybrid processing technique based Logistic Regression (LR) well Principal Components Analysis (PCA) presented called LR-PCA. Such a process helps select significant principal components (PCs) further use them purpose. proposed system examined public benchmark datasets which INbreast mini-MAIS. could highest performance accuracies 98.60% 98.80% mini-MAIS datasets, seems be useful reliable diagnosis.

Language: Английский

---

Applying Deep Learning to Medical Imaging: A Review

Applied Sciences, Journal Year: 2023, Volume and Issue: 13(18), P. 10521 - 10521

Published: Sept. 21, 2023

Deep learning (DL) has made significant strides in medical imaging. This review article presents an in-depth analysis of DL applications imaging, focusing on the challenges, methods, and future perspectives. We discuss impact diagnosis treatment diseases how it revolutionized imaging field. Furthermore, we examine most recent techniques, such as convolutional neural networks (CNNs), recurrent (RNNs), generative adversarial (GANs), their Lastly, provide insights into highlighting its potential advancements challenges.

Language: Английский

---

VGG16 Feature Extractor with Extreme Gradient Boost Classifier for Pancreas Cancer Prediction

Journal of Imaging, Journal Year: 2023, Volume and Issue: 9(7), P. 138 - 138

Published: July 7, 2023

The prognosis of patients with pancreatic ductal adenocarcinoma (PDAC) is greatly improved by an early and accurate diagnosis. Several studies have created automated methods to forecast PDAC development utilising various medical imaging modalities. These papers give a general overview the classification, segmentation, or grading many cancer types conventional machine learning techniques hand-engineered characteristics, including cancer. This study uses cutting-edge deep identify computerised tomography (CT) work suggests that hybrid model VGG16-XGBoost (VGG16-backbone feature extractor Extreme Gradient Boosting-classifier) for images. According studies, proposed performs better, obtaining accuracy 0.97 weighted F1 score dataset under study. experimental validation Cancer Imaging Archive (TCIA) public access dataset, which has pancreas CT results this can be extremely helpful diagnosis from images, categorising them into five different tumours (T), node (N), metastases (M) (TNM) staging system class labels, are T0, T1, T2, T3, T4.

Language: Английский

---

An optimised YOLOv4 deep learning model for efficient malarial cell detection in thin blood smear images

Parasites & Vectors, Journal Year: 2024, Volume and Issue: 17(1)

Published: April 16, 2024

Abstract Background Malaria is a serious public health concern worldwide. Early and accurate diagnosis essential for controlling the disease’s spread avoiding severe complications. Manual examination of blood smear samples by skilled technicians time-consuming aspect conventional malaria toolbox. persists in many parts world, emphasising urgent need sophisticated automated diagnostic instruments to expedite identification infected cells, thereby facilitating timely treatment reducing risk disease transmission. This study aims introduce more lightweight quicker model—but with improved accuracy—for diagnosing using YOLOv4 (You Only Look Once v. 4) deep learning object detector. Methods The model modified direct layer pruning backbone replacement. primary objective removal individual analysis residual blocks within C3, C4 C5 (C3–C5) Res-block bodies architecture’s C3-C5 bodies. CSP-DarkNet53 simultaneously replaced enhanced feature extraction shallower ResNet50 network. performance metrics models are compared analysed. Results outperform original model. YOLOv4-RC3_4 pruned from C3 body achieves highest mean accuracy precision (mAP) 90.70%. mAP > 9% higher than that model, saving approximately 22% billion floating point operations (B-FLOPS) 23 MB size. findings indicate also performs better, an increase 9.27% detecting cells upon redundant layers CSP-DarkeNet53 backbone. Conclusions results this highlight use red cells. Pruning helps determine which contribute most least, respectively, model’s performance. Our method has potential revolutionise pave way novel learning-based bioinformatics solutions. Developing effective process will considerably global efforts combat debilitating disease. We have shown removing undesirable can reduce size its computational complexity without compromising precision. Graphical

Language: Английский

---

A Multidisciplinary Hyper-Modeling Scheme in Personalized In Silico Oncology: Coupling Cell Kinetics with Metabolism, Signaling Networks, and Biomechanics as Plug-In Component Models of a Cancer Digital Twin

Journal of Personalized Medicine, Journal Year: 2024, Volume and Issue: 14(5), P. 475 - 475

Published: April 29, 2024

The massive amount of human biological, imaging, and clinical data produced by multiple diverse sources necessitates integrative modeling approaches able to summarize all this information into answers specific questions. In paper, we present a hypermodeling scheme combine models cancer aspects regardless their underlying method or scale. Describing tissue-scale cell proliferation, biomechanical tumor growth, nutrient transport, genomic-scale aberrant metabolism, cell-signaling pathways that regulate the cellular response therapy, hypermodel integrates mutation, miRNA expression, data. constituting hypomodels, as well orchestration links, are described. Two types, Wilms (nephroblastoma) non-small lung cancer, addressed proof-of-concept study cases. Personalized simulations actual anatomy patient have been conducted. has also applied predict control after radiotherapy relationship between proliferative activity neoadjuvant chemotherapy. Our innovative holds promise digital twin-based decision support system core future in silico trial platforms, although additional retrospective adaptation validation necessary.

Language: Английский

---

Effect of neural network structure in accelerating performance and accuracy of a convolutional neural network with GPU/TPU for image analytics

PeerJ Computer Science, Journal Year: 2022, Volume and Issue: 8, P. e909 - e909

Published: March 3, 2022

In deep learning the most significant breakthrough in field of image recognition, object detection language processing was done by Convolutional Neural Network (CNN). Rapid growth data and neural networks performance DNN algorithms depends on computation power storage capacity devices.In this paper, convolutional network used for various applications studied its acceleration platforms like CPU, GPU, TPU done. The structure computing characteristics analyzed summarized, effect these accelerating tasks is also explained. Cross-platform comparison CNN using three face mask (object detection/Computer Vision), Virus Detection Plants (Image Classification: agriculture sector), Pneumonia from X-ray Images Classification/medical field).The implementation a comprehensive to identify performance, throughput, bottlenecks, training time. layer-wise execution GPU explained with analysis. impact fully connected layer analyzed. challenges faced during process were discussed future works are identified.

Language: Английский

---

Deep learning for image-based liver analysis — A comprehensive review focusing on malignant lesions

Artificial Intelligence in Medicine, Journal Year: 2022, Volume and Issue: 130, P. 102331 - 102331

Published: June 10, 2022

Deep learning-based methods, in particular, convolutional neural networks and fully are now widely used the medical image analysis domain. The scope of this review focuses on using deep learning focal liver lesions, with a special interest hepatocellular carcinoma metastatic cancer; structures like parenchyma or vascular system. Here, we address several network architectures for analyzing anatomical lesions from various imaging modalities such as computed tomography, magnetic resonance ultrasound. Image tasks segmentation, object detection classification liver, vessels discussed. Based qualitative search, 91 papers were filtered out survey, including journal publications conference proceedings. reviewed work grouped into eight categories based methodologies used. By comparing evaluation metrics, hybrid models performed better both lesion segmentation tasks, ensemble classifiers vessel combined approach tasks. performance was measured Dice score accuracy which most commonly metrics.

Language: Английский

---