Automated detection and forecasting of COVID-19 using deep learning techniques: A review

Neurocomputing, Journal Year: 2024, Volume and Issue: 577, P. 127317 - 127317

Published: Jan. 26, 2024

Language: Английский

---

Segmentation-Based Classification Deep Learning Model Embedded with Explainable AI for COVID-19 Detection in Chest X-ray Scans

Diagnostics, Journal Year: 2022, Volume and Issue: 12(9), P. 2132 - 2132

Published: Sept. 2, 2022

Background and Motivation: COVID-19 has resulted in a massive loss of life during the last two years. The current imaging-based diagnostic methods for detection multiclass pneumonia-type chest X-rays are not so successful clinical practice due to high error rates. Our hypothesis states that if we can have segmentation-based classification rate <5%, typically adopted 510 (K) regulatory purposes, system be adapted settings. Method: This study proposes 16 types deep learning-based systems automatic, rapid, precise COVID-19. segmentation networks, namely UNet UNet+, along with eight models, VGG16, VGG19, Xception, InceptionV3, Densenet201, NASNetMobile, Resnet50, MobileNet, were applied select best-suited combination networks. Using cross-entropy function, performance was evaluated by Dice, Jaccard, area-under-the-curve (AUC), receiver operating characteristics (ROC) validated using Grad-CAM explainable AI framework. Results: best performing model UNet, which exhibited accuracy, loss, AUC 96.35%, 0.15%, 94.88%, 90.38%, 0.99 (p-value <0.0001), respectively. UNet+Xception, precision, recall, F1-score, 97.45%, 97.46%, 97.43%, 0.998 outperformed existing models. mean improvement UNet+Xception over all remaining studies 8.27%. Conclusion: is viable option as (error <5%) holds true thus adaptable practice.

Language: Английский

---

Brain Tumor Characterization Using Radiogenomics in Artificial Intelligence Framework

Cancers, Journal Year: 2022, Volume and Issue: 14(16), P. 4052 - 4052

Published: Aug. 22, 2022

Brain tumor characterization (BTC) is the process of knowing underlying cause brain tumors and their characteristics through various approaches such as segmentation, classification, detection, risk analysis. The substantial includes identification molecular signature useful genomes whose alteration causes tumor. radiomics approach uses radiological image for disease by extracting quantitative features in artificial intelligence (AI) environment. However, when considering a higher level genetic information mutation status, combined study “radiomics genomics” has been considered under umbrella “radiogenomics”. Furthermore, AI radiogenomics’ environment offers benefits/advantages finalized outcome personalized treatment individualized medicine. proposed summarizes tumor’s prospect an emerging field research, i.e., radiogenomics environment, with help statistical observation risk-of-bias (RoB) PRISMA search was used to find 121 relevant studies review using IEEE, Google Scholar, PubMed, MDPI, Scopus. Our findings indicate that both have successfully applied aggressively several oncology applications numerous advantages. paradigm, conventional deep made impact on favorable outcomes BTC. analysis better understanding architectures stronger benefits providing bias involved them.

Language: Английский

---

UNet Deep Learning Architecture for Segmentation of Vascular and Non-Vascular Images: A Microscopic Look at UNet Components Buffered With Pruning, Explainable Artificial Intelligence, and Bias

IEEE Access, Journal Year: 2022, Volume and Issue: 11, P. 595 - 645

Published: Dec. 26, 2022

Biomedical image segmentation (BIS) task is challenging due to the variations in organ types, position, shape, size, scale, orientation, and contrast. Conventional methods lack accurate automated designs. Artificial intelligence (AI)-based UNet has recently dominated BIS. This first review of its kind that microscopically addressed types by complexity, stratification components, addressing vascular vs. non-vascular framework, key challenge UNet-based architecture, finally interfacing three facets AI, pruning, explainable AI (XAI), AI-bias. PRISMA was used select 267 studies. Five classes were identified labeled as conventional UNet, superior attention-channel hybrid ensemble UNet. We discovered 81 considering six kinds namely encoder, decoder, skip connection, bridge network, loss function, their combination. Vascular architecture compared. AP(ai)Bias 2.0-UNet these based on (i) attributes performance, (ii) and, (iii) pruning (compression). bias such ranking, radial, regional area, (iv) PROBAST, (v) ROBINS-I applied compared using a Venn diagram. systems with sUNet attention. Most studies suffered from low interest XAI strategies. None models qualified be bias-free. There need move paper-to-practice paradigms for clinical evaluation settings.

Language: Английский

---

Attention-Based UNet Deep Learning Model for Plaque Segmentation in Carotid Ultrasound for Stroke Risk Stratification: An Artificial Intelligence Paradigm

Journal of Cardiovascular Development and Disease, Journal Year: 2022, Volume and Issue: 9(10), P. 326 - 326

Published: Sept. 27, 2022

Stroke and cardiovascular diseases (CVD) significantly affect the world population. The early detection of such events may prevent burden death costly surgery. Conventional methods are neither automated nor clinically accurate. Artificial Intelligence-based automatically detecting predicting severity CVD stroke in their stages prime importance. This study proposes an attention-channel-based UNet deep learning (DL) model that identifies carotid plaques internal artery (ICA) common (CCA) images. Our experiments consist 970 ICA images from UK, 379 CCA diabetic Japanese patients, 300 post-menopausal women Hong Kong. We combined both to form integrated database 679 A rotation transformation technique was applied images, doubling for experiments. cross-validation K5 (80% training: 20% testing) protocol accuracy determination. results Attention-UNet benchmarked against UNet, UNet++, UNet3P models. Visual plaque segmentation showed improvement compared other three correlation coefficient (CC) value is 0.96, 0.93, 0.92 Similarly, AUC 0.97, 0.964, 0.966, 0.965 Conclusively, beneficial segmenting very bright fuzzy hard diagnose using methods. Further, we present a multi-ethnic, multi-center, racial bias-free risk assessment.

Language: Английский

---

Explainable artificial intelligence (XAI) in radiology and nuclear medicine: a literature review

Frontiers in Medicine, Journal Year: 2023, Volume and Issue: 10

Published: May 12, 2023

Rational Deep learning (DL) has demonstrated a remarkable performance in diagnostic imaging for various diseases and modalities therefore high potential to be used as clinical tool. However, current practice shows low deployment of these algorithms practice, because DL lack transparency trust due their underlying black-box mechanism. For successful employment, explainable artificial intelligence (XAI) could introduced close the gap between medical professionals algorithms. In this literature review, XAI methods available magnetic resonance (MR), computed tomography (CT), positron emission (PET) are discussed future suggestions made. Methods PubMed, Embase.com Clarivate Analytics/Web Science Core Collection were screened. Articles considered eligible inclusion if was (and well described) describe behavior model MR, CT PET imaging. Results A total 75 articles included which 54 17 described post ad hoc methods, respectively, 4 both methods. Major variations is seen Overall, lacks ability provide class-discriminative target-specific explanation. Ad seems tackle its intrinsic explain. quality control rarely applied systematic comparison difficult. Conclusion There currently no clear consensus on how should deployed order implementation. We advocate technical assessment Also, ensure end-to-end unbiased safe integration workflow, (anatomical) data minimization included.

Language: Английский

---

Fused deep learning paradigm for the prediction of o6-methylguanine-DNA methyltransferase genotype in glioblastoma patients: A neuro-oncological investigation

Computers in Biology and Medicine, Journal Year: 2023, Volume and Issue: 153, P. 106492 - 106492

Published: Jan. 4, 2023

Language: Английский

---

Ensemble Deep Learning Derived from Transfer Learning for Classification of COVID-19 Patients on Hybrid Deep-Learning-Based Lung Segmentation: A Data Augmentation and Balancing Framework

Diagnostics, Journal Year: 2023, Volume and Issue: 13(11), P. 1954 - 1954

Published: June 2, 2023

Lung computed tomography (CT) techniques are high-resolution and well adopted in the intensive care unit (ICU) for COVID-19 disease control classification. Most artificial intelligence (AI) systems do not undergo generalization typically overfitted. Such trained AI practical clinical settings therefore give accurate results when executed on unseen data sets. We hypothesize that ensemble deep learning (EDL) is superior to transfer (TL) both non-augmented augmented frameworks.

Language: Английский

---

Automated machine learning with interpretation: A systematic review of methodologies and applications in healthcare

Medicine Advances, Journal Year: 2024, Volume and Issue: 2(3), P. 205 - 237

Published: Aug. 27, 2024

Abstract Machine learning (ML) has achieved substantial success in performing healthcare tasks which the configuration of every part ML pipeline relies heavily on technical knowledge. To help professionals with borderline expertise to better use techniques, Automated (AutoML) emerged as a prospective solution. However, most models generated by AutoML are black boxes that challenging comprehend and deploy settings. We conducted systematic review examine interpretation systems for healthcare. searched four databases (MEDLINE, EMBASE, Web Science, Scopus) complemented seven prestigious conferences (AAAI, ACL, ICLR, ICML, IJCAI, KDD, NeurIPS) reported before September 1, 2023. included 118 articles related First, we illustrated techniques used publications, including automated data preparation, feature engineering, model development, accompanied real‐world case study demonstrate advantages over classic ML. Then, summarized methods: interaction importance, dimensionality reduction, intrinsically interpretable models, knowledge distillation rule extraction. Finally, detailed how been six major types: image, free text, tabular data, signal, genomic sequences, multi‐modality. some extent, provides effortless development improves users' trust In future studies, researchers should explore seamless integration automation interpretation, compatibility multi‐modality, utilization foundation models.

Language: Английский

---

Unlocking the black box: an in-depth review on interpretability, explainability, and reliability in deep learning

Neural Computing and Applications, Journal Year: 2024, Volume and Issue: unknown

Published: Nov. 18, 2024

Language: Английский

---