Fetal cardiac ultrasound standard section detection model based on multitask learning and mixed attention mechanism

Neurocomputing, Journal Year: 2024, Volume and Issue: 579, P. 127443 - 127443

Published: Feb. 21, 2024

Language: Английский

---

Integrating Multimodal Data Fusion for Advanced Biomedical Analysis

Published: Jan. 13, 2025

The multimodality of biomedical data is on the rise, enabling recording intricate interactions between biological functions. Data fusion algorithms that utilize deep learning (DL) are often used to simulate these dynamic connections. Hence, we conduct a comprehensive examination existing cutting-edge techniques and put forth an elaborate classification system enables more knowledgeable selection approaches for purposes, along with exploration innovative methodologies. Through this process, see frequently surpass unimodal in nature shallow terms performance. Furthermore, suggested subcategories exhibit distinct benefits limitations. methodologies has revealed combined representation remains favored strategy, particularly intermediate merging methods, due its ability accurately capture relationships among many levels organization. Ultimately, acknowledge use progressive fusion, which relies understanding search methodologies, holds great potential as future avenue study. Moreover, application transfer might potentially address constraints imposed by restricted number samples multimodal collections. With growing availability large sets, offer chance train models capable regulatory dynamics associated health illness.

Language: Английский

---

Multi-Modal Fusion of Routine Care Electronic Health Records (EHR): A Scoping Review

Information, Journal Year: 2025, Volume and Issue: 16(1), P. 54 - 54

Published: Jan. 15, 2025

Background: Electronic health records (EHR) are now widely available in healthcare institutions to document the medical history of patients as they interact with services. In particular, routine care EHR data collected for a large number patients.These span multiple heterogeneous elements (i.e., demographics, diagnosis, medications, clinical notes, vital signs, and laboratory results) which contain semantic, concept, temporal information. Recent advances generative learning techniques were able leverage fusion enhance decision support. Objective: A scoping review proposed including architectures, input elements, application areas is needed synthesize variances identify research gaps that can promote re-use these new outcomes. Design: comprehensive literature search was conducted using Google Scholar high impact architectures over multi-modal during period 2018 2023. The guidelines from PRISMA (Preferred Reporting Items Systematic Reviews Meta-Analyses) extension followed. findings derived selected studies thematic comparative analysis. Results: revealed lack standard definition transformed into modalities. These definitions ignore one or more key characteristics source, encoding scheme, concept level. Moreover, order adapt emergent techniques, classification should distinguish take consideration concurrently happen all three layers encoding, representation, decision). aspects constitute first step towards streamlined approach design data. addition, current pretrained models inconsistent their handling semantic information thereby hindering different applications settings. Conclusions: Current mostly follow design-by-example methodology. Guidelines efficient broad range applications. addition promoting re-use, need outline best practices combining modalities while leveraging transfer co-learning well encoding.

Language: Английский

---

Integrating language into medical visual recognition and reasoning: A survey

Medical Image Analysis, Journal Year: 2025, Volume and Issue: 102, P. 103514 - 103514

Published: Feb. 27, 2025

Language: Английский

---

Transformer and Attention-Based Architectures for Segmentation of Coronary Arterial Walls in Intravascular Ultrasound: A Narrative Review

Diagnostics, Journal Year: 2025, Volume and Issue: 15(7), P. 848 - 848

Published: March 26, 2025

Background: The leading global cause of death is coronary artery disease (CAD), necessitating early and precise diagnosis. Intravascular ultrasound (IVUS) a sophisticated imaging technique that provides detailed visualization arteries. However, the methods for segmenting walls in IVUS scan into internal wall structures quantifying plaque are still evolving. This study explores use transformers attention-based models to improve diagnostic accuracy segmentation scans. Thus, objective explore application transformer scans assess their inherent biases artificial intelligence systems improving accuracy. Methods: By employing Preferred Reporting Items Systematic Reviews Meta-Analyses (PRISMA) framework, we pinpointed examined top strategies using transformer-based techniques, assessing traits, scientific soundness, clinical relevancy. Coronary thickness determined by boundaries (inner: lumen-intima outer: media-adventitia) through cross-sectional Additionally, it first investigate deep learning (DL) associated with segmentation. Finally, incorporates explainable AI (XAI) concepts DL structure Findings: Because its capacity automatically extract features at numerous scales encoders, rebuild segmented pictures via decoders, fuse variations skip connections, UNet model stands out as an efficient Conclusions: investigation underscores deficiency incentives embracing XAI pruned (PAI) models, no attaining bias-free configuration. Shifting from theoretical practical usage crucial bolstering evaluation deployment.

Language: Английский

---

Leveraging Transparent Ontology Learning to Refine Constructs in Neuroscience

Neuroscience Informatics, Journal Year: 2025, Volume and Issue: unknown, P. 100199 - 100199

Published: March 1, 2025

Language: Английский

---

A vision foundation model-based method for large-scale forest disturbance mapping using time series Sentinel-1 SAR data

Remote Sensing of Environment, Journal Year: 2025, Volume and Issue: 325, P. 114775 - 114775

Published: April 29, 2025

Language: Английский

---

Alzheimer’s Disease: Exploring Pathophysiological Hypotheses and the Role of Machine Learning in Drug Discovery

International Journal of Molecular Sciences, Journal Year: 2025, Volume and Issue: 26(3), P. 1004 - 1004

Published: Jan. 24, 2025

Alzheimer’s disease (AD) is a major neurodegenerative dementia, with its complex pathophysiology challenging current treatments. Recent advancements have shifted the focus from traditionally dominant amyloid hypothesis toward multifactorial understanding of disease. Emerging evidence suggests that while amyloid-beta (Aβ) accumulation central to AD, it may not be primary driver but rather part broader pathogenic process. Novel hypotheses been proposed, including role tau protein abnormalities, mitochondrial dysfunction, and chronic neuroinflammation. Additionally, gut–brain axis epigenetic modifications gained attention as potential contributors AD progression. The limitations existing therapies underscore need for innovative strategies. This study explores integration machine learning (ML) in drug discovery accelerate identification novel targets candidates. ML offers ability navigate AD’s complexity, enabling rapid analysis extensive datasets optimizing clinical trial design. synergy between these themes presents promising future more effective

Language: Английский

---

Special Issue: Artificial Intelligence Technology in Medical Image Analysis

Applied Sciences, Journal Year: 2024, Volume and Issue: 14(5), P. 2180 - 2180

Published: March 5, 2024

Artificial intelligence (AI) technologies have significantly advanced the field of medical imaging, revolutionizing diagnostic and therapeutic processes [...]

Language: Английский

---

Optimizing Image Enhancement: Feature Engineering for Improved Classification in AI-Assisted Artificial Retinas

Sensors, Journal Year: 2024, Volume and Issue: 24(9), P. 2678 - 2678

Published: April 23, 2024

Artificial retinas have revolutionized the lives of many blind people by enabling their ability to perceive vision via an implanted chip. Despite significant advancements, there are some limitations that cannot be ignored. Presenting all objects captured in a scene makes identification difficult. Addressing this limitation is necessary because artificial retina can utilize very limited number pixels represent information. This problem multi-object scenario mitigated enhancing images such only major considered shown vision. Although simple techniques like edge detection used, they fall short representing identifiable complex scenarios, suggesting idea integrating primary object edges. To support idea, proposed classification model aims at identifying based on suggested set selective features. The then equipped into system for filtering multiple enhance suitability handling multi-objects enables cope with real-world scenarios. multi-label deep neural network, specifically designed leverage from feature set. Initially, enhanced research compared ones technique single, dual, and images. These enhancements also verified through intensity profile analysis. Subsequently, model's performance evaluated show significance utilizing includes evaluating correctly classify top five, four, three, two, one object(s), respective accuracies up 84.8%, 85.2%, 86.8%, 91.8%, 96.4%. Several comparisons as training/validation loss accuracies, precision, recall, specificity, area under curve indicate reliable results. Based overall evaluation study, it concluded using features not improves performance, but aligns specific address challenge Therefore, basis useful tool supporting optimizing image enhancement.

Language: Английский

---