Search Results (1,227)

Automatic radiology report generation (ARRG) has emerged as a promising application of deep learning (DL) with the potential to alleviate reporting workload and improve diagnostic consistency. However, despite rapid methodological advances, the field remains technically fragmented and not yet mature for routine clinical adoption. This scoping review maps the current ARRG research landscape by examining DL architectures, multimodal integration strategies, and evaluation practices from 2015 to April 2025. Following the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) guidelines, a comprehensive literature search identified 89 eligible studies, revealing a marked predominance of chest radiography datasets (87.6%), primarily driven by their public availability and the accelerated development of automated tools during the COVID-19 pandemic. Most models employed hybrid architectures (73%), particularly CNN–Transformer pairings, reflecting a shift toward systems that combine local feature extraction with global contextual reasoning. Although these approaches have achieved measurable gains in textual and semantic coherence, several challenges persist, including limited anatomical diversity, weak alignment with radiological rationale, and evaluation metrics that insufficiently reflect diagnostic adequacy or clinical impact. Overall, the findings indicate a rapidly evolving but clinically immature field, underscoring the need for validation frameworks that more closely reflect radiological practice and support future deployment in real-world settings. Full article

This study investigates how large language models (LLMs) handle Chinese zero anaphora under symmetric minimal-pair conditions designed to neutralize shallow syntactic cues. We construct a Winograd-style benchmark of carefully controlled sentence pairs that require semantic interpretation, pragmatic inference, discourse tracking, and commonsense reasoning rather than structural heuristics. Using GPT-4, ChatGLM-4, and LLaMA-3 under zero-shot, one-shot, and few-shot prompting, we assess both accuracy and the reasoning traces generated through a standardized Chain-of-Thought diagnostic. Results show that all models perform consistently on items solvable through local cues but display systematic asymmetric errors on 19 universally misinterpreted sentences that demand deeper discourse reasoning. Analysis of these failures reveals weaknesses in semantic role differentiation, topic-chain maintenance, logical-relation interpretation, pragmatic inference, and long-distance dependency tracking. These findings suggest that while LLMs perform well on simpler tasks, they still face challenges in interpreting contextually omitted arguments in Chinese. The study provides a new controlled evaluation resource, an interpretable error analysis framework, and evidence of differences in symmetric versus asymmetric reasoning behaviors in LLMs. Future research could expand the current benchmark to longer discourse contexts, incorporate multi-modal or knowledge-grounded cues, and explore fine-tuning LLMs on discourse data, helping clarify whether asymmetric patterns stem from deeper reasoning challenges or from interactions between models and the evaluation format. Full article

This study was conducted to address the demand for interpretable intelligent recognition of fruit tree diseases in smart horticultural environments. A KAD-Former framework integrating an agricultural knowledge graph with a visual Transformer was proposed and systematically validated through extensive cross-regional, multi-variety, and multi-disease experiments. The primary objective of this work was to overcome the limitations of conventional deep models, including insufficient interpretability, unstable recognition of weak disease features, and poor cross-regional generalization. In the experimental evaluation, the model achieved significant advantages across multiple representative tasks: in the overall performance comparison, KAD-Former reached an accuracy of $0.946$, an F1-score of $0.933$, and a mAP of $0.938$, outperforming classical models such as ResNet50, EfficientNet, and Swin-T. In the cross-regional generalization assessment, a DGS of $0.933$ was obtained, notably surpassing competing models. In terms of explainability consistency, a Consistency@5 score of $0.826$ indicated strong alignment between the model’s attention regions and expert annotations. The ablation experiments further demonstrated that the three core modules—AKG (agricultural knowledge graph), SAM (semantic alignment module), and KGA (knowledge-guided attention)—each contributed substantially to final performance, with the complete model exhibiting the best results. These findings collectively demonstrate the comprehensive advantages of KAD-Former in disease classification, symptom localization, model interpretability, and cross-domain transfer. The proposed method not only achieved state-of-the-art performance in pure visual tasks but also advanced knowledge-enhanced and interpretable reasoning by emulating the diagnostic logic employed by agricultural experts in real orchard scenarios. Through the integration of the agricultural knowledge graph, semantic alignment, and knowledge-guided attention, the model maintained stable performance under challenging conditions such as complex illumination, background noise, and weak lesion features, while exhibiting strong robustness in cross-region and cross-variety transfer tests. Furthermore, the experimental results indicated that the approach enhanced fine-grained recognition capabilities for various fruit tree diseases, including apple ring rot, brown spot, powdery mildew, and downy mildew. Full article

Abductive reasoning, or abduction, is a key process in scientific discovery and medical diagnosis. In everyday dermatology and dermatopathology, however, it functions as the practical engine behind differential diagnosis, clinicopathologic correlation, and disciplined pattern recognition. In this paper, we retain the epistemological foundation of abduction but translate it into usable steps for clinicians and dermatopathologists. We distinguish abduction from deduction and induction; separate creative abduction (which generates new concepts) from selective abduction (daily diagnostic choice); and show how both operate within a simple Select-and-Test (ST) Model: select a hypothesis, deduce what else should be true, test against data, and then update. We then reinterpret Ackerman’s algorithmic method of pattern analysis as an operationalization of the ST-Model. Through a couple of concise case vignettes, we illustrate visual and manipulative abduction, nonmonotonic updates, and the role of artifacts (dermoscopy, DIF, stains) as so-called epistemic mediators. Finally, we map contemporary AI tools to selective abduction and propose practical guardrails for fairness, transparency, and accountability. The result is a pragmatic framework that preserves philosophical depth while addressing the daily needs of dermatologists and dermatopathologists in the clinic and at the microscope. Full article

Dizziness is one of the most common reasons for presentation to emergency departments (EDs). While excellent reviews exist on distinguishing central from peripheral vestibular causes, the aim of this article is different. Instead of focusing on this classic differentiation, dizziness as a nonspecific symptom will be addressed. Practical considerations will be outlined to help avoid missing red flags. Emphasis is placed on history-taking and clinical examination, supported by statistical reasoning. It will be argued that this approach does not require additional time—an especially valuable resource in the ED—and may even reduce overall time expenditure without compromising diagnostic accuracy. Full article

Construction safety remains a critical concern, with frequent accidents leading to fatalities, severe injuries, and significant economic losses. To address these challenges and enhance accident prevention, this study adopts a systems thinking approach to investigate the causal factors of construction safety accidents. First, drawing on Rasmussen’s risk management framework, this study developed a Construction Accident Causation System (CACS) model that comprises six hierarchical levels and 23 influencing factors. Through the analysis of 331 investigation reports of construction accidents in China, causal factor correlations were refined, and the topological structure and network parameters of the model were determined. This study integrates diagnostic reasoning, sensitivity analysis, and fuzzy mathematics within a Bayesian Network (BN) framework. Through this approach, it identifies the most probable accident pathways and highlights seven critical and three sensitive factors that jointly exacerbate construction safety risks. A real-world case of a formwork collapse in Baotou City is further analyzed to verify the model’s reliability and practical relevance. The results confirm that the integrated CACS and BN framework effectively captures the multi-level interactions among managerial, behavioral, and technical factors, providing a scientific basis for proactive safety management and accident prevention in the construction industry. Full article

Bridge health diagnosis plays a vital role in ensuring structural safety and extending service life while reducing maintenance costs. Traditional structural health monitoring approaches rely on sensor-based measurements, which are costly, labor-intensive, and limited in coverage. To address these challenges, we propose a three-phase solution that integrates the Dynamic Lightweight Vision-Language Model (DL-VLM), domain adaptation, and knowledge-enhanced reasoning. First, as the core of the framework, the DL-VLM consists of three components: a visual information encoder with multi-scale feature selection, a text encoder for processing inspection-related language, and a multimodal alignment module. Second, to enhance practical applicability, we further introduce domain-specific fine-tuning on the Bridge-SHM dataset, enabling the model to acquire specialized knowledge of bridge construction, defects, and structural components. Third, a knowledge retrieval augmentation module is incorporated, leveraging external knowledge graphs and vector-based retrieval to provide contextually relevant information and improve diagnostic reasoning. Experiments on high-resolution bridge inspection datasets demonstrate that DL-VLM achieves competitive diagnostic accuracy while substantially reducing computational cost. The combination of domain-specific fine-tuning and knowledge augmentation significantly improves performance on specialized tasks, supporting efficient and practical deployment in real-world structural health monitoring scenarios. Full article

Background/Objectives: Immunohistochemistry (IHC) is a critical diagnostic tool in forensic pathology, enabling molecular-level assessment of wound vitality, post-mortem interval, and cause of death. However, IHC interpretation is subject to variability due to its reliance on human expertise. This study investigates whether artificial intelligence (AI), specifically a generative model, can assist in the diagnostic evaluation of IHC slides and replicate expert-level scoring, thereby improving consistency and reproducibility. Methods: A total of 225 high-resolution IHC images were classified into five immunoreactivity categories. The AI model (ChatGPT-4V) was trained on 150 labeled images and tested blindly on 75 unseen slides. Performance was assessed using confusion matrices, per-class precision/recall/F1, overall accuracy, Cohen’s κ (unweighted and weighted), and binary metrics (sensitivity, specificity, MCC). Results: Overall accuracy was 81.3% (95% CI: 71.1–88.5%), with substantial agreement (κ = 0.767 unweighted; 0.805 linear-weighted; 0.848 quadratic-weighted). Binary classification achieved a sensitivity of 98.3%, specificity of 93.3%, MCC of 0.92. Accuracy was highest in extreme categories (− and +++, 93.3%), while intermediate classes (+ and ++) showed reduced performance (error rates up to 33%). Evaluation was rapid and consistent but lacked interpretative reasoning and struggled with borderline cases. Conclusions: AI-assisted diagnostic evaluation of IHC slides demonstrates promising accuracy and consistency, particularly in well-defined staining patterns. While not a replacement for human expertise, AI can serve as a valuable adjunct in forensic pathology, supporting rapid and standardized assessments. Ethical and legal considerations must guide its implementation in medico-legal contexts. Full article

In this study, a structured and methodological evaluation approach for eXplainable Artificial Intelligence (XAI) methods in medical image classification is proposed and implemented using LIME and SHAP explanations for chest X-ray interpretations. The evaluation framework integrates two critical perspectives: predictive model-centered and human-centered evaluations. Predictive model-centered evaluations examine the explanations’ ability to reflect changes in input and output data and the internal model structure. Human-centered evaluations, conducted with 97 medical experts, assess trust, confidence, and agreements with AI’s indicative and contra-indicative reasoning as well as their changes before and after provision of explainability. Key findings of our study include explanation of sensitivity of LIME and SHAP to model changes, their effectiveness in identifying critical features, and SHAP’s significant impact on diagnosis changes. Our results show that both LIME and SHAP negatively affected contra-indicative agreement. Case-based analysis revealed AI explanations reinforce trust and agreement when participant’s initial diagnoses are correct. In these cases, SHAP effectively facilitated correct diagnostic changes. This study establishes a benchmark for future research in XAI for medical image analysis, providing a robust foundation for evaluating and comparing different XAI methods. Full article

Background: Vision-language models show promise in medical image interpretation, but their performance in musculoskeletal tumor diagnostics remains underexplored. Objective: To evaluate the diagnostic accuracy of six large language models on orthopedic radiographs for tumor detection, classification, anatomical localization, and X-ray view interpretation, and to assess the impact of demographic context and self-reported certainty. Methods: We retrospectively evaluated six VLMs on 3746 expert-annotated orthopedic radiographs from the Bone Tumor X-ray Radiograph dataset. Each image was analyzed by all models with and without patient age and sex using a standardized prompting scheme across four predefined tasks. Results: Over 48,000 predictions were analyzed. Tumor detection accuracy ranged from 59.9–73.5%, with the Gemini Ensemble achieving the highest F1 score (0.723) and recall (0.822). Benign/malignant classification reached up to 85.2% accuracy; tumor type identification 24.6–55.7%; body region identification 97.4%; and view classification 82.8%. Demographic data improved tumor detection accuracy (+1.8%, p < 0.001) but had no significant effect on other tasks. Certainty scores were weakly correlated with correctness, with Gemini Pro highest (r = 0.089). Conclusion: VLMs show strong potential for basic musculoskeletal radiograph interpretation without task-specific training but remain less accurate than specialized deep learning models for complex classification. Limited calibration, interpretability, and contextual reasoning must be addressed before clinical use. This is the first systematic assessment of image-based diagnosis and self-assessment in LLMs using a real-world radiology dataset. Full article

Background/Objectives: Genital discomfort, manifested by vulvar itching and burning, is a frequent complaint among women of all ages and has multifactorial origins—including dermatoses, infections, allergies, and hormonal disorders. The study aimed to determine whether selected medical history factors—age, obstetric history, and body mass index (BMI)—influence the frequency of genital discomfort as a reason for gynecological consultation. Methods: A pilot study included 288 female patients aged 11–91 years who presented to outpatient gynecological clinics between September 2018 and February 2025 with symptoms of vulvar itching and genital discomfort. Qualitative data were expressed as numbers and percentages, and age was described using mean, median, quartiles, and range. Associations between categorical variables were assessed using Pearson’s chi-square test, with statistical significance set at p < 0.05. Results: The mean age of patients was 47.4 ± 20.3 years. Most were diagnosed with ICD-10 code N90 (82.6%), while 17.4% had N76. Genital discomfort was most frequently reported by women aged 41–50 years (p < 0.0001). Comorbidities (p < 0.0001) and obstetric history (p < 0.0001) significantly influenced the occurrence of genital discomfort, which was more prevalent among women with chronic conditions and those who had been pregnant. No significant associations were found with BMI (p = 0.2353) or menopausal status (p = 0.3458). Conclusions: Genital discomfort is a common and multifactorial condition requiring an interdisciplinary diagnostic and therapeutic approach. Collaboration among gynecologists, dermatologists, endocrinologists, and dietitians is crucial for effective management and prevention. Full article

This paper provides a comprehensive overview of the application of various radiological modalities, with a critical comparison between human and veterinary medicine. The modalities discussed include conventional radiography, dual-energy X-ray absorptiometry (DXA), computed tomography (CT), magnetic resonance imaging (MRI), ultrasound (US), quantitative ultrasound (QUS), positron emission tomography-computed tomography (PET-CT) and micro and nano computed tomography (micro-CT, nano-CT) in clinical practice and basic research of skeletal system. Radiological imaging plays a crucial role in the diagnosis, monitoring and research of skeletal system disorders in both human and veterinary medicine. In preclinical research, advanced diagnostic imaging modalities such as micro-CT and nano-CT allow for 3D quantification of trabecular and cortical bone microarchitecture for studies in bone biology, regenerative medicine and pharmacological research. Furthermore, the integration of artificial intelligence is advancing image interpretation, precision diagnostics and disease tracking. Despite their broad utility, imaging modalities must be selected based on clinical indication, species, age and anatomical region with consideration of radiation dose, cost and availability, especially in remote regions. For this reason, clinicians and radiologists remain an irreplaceable part of diagnostic imaging. Full article

Background: Acute Respiratory Infections are a common reason for Pediatric Emergency Department (PED) visits. Differentiating bacterial and viral infections may be challenging and might result in incorrect antibiotic prescriptions and exacerbation of antimicrobial resistance. This study evaluated the impact of new diagnostic tests in PED. Methods: A retrospective cohort of 4882 acute febrile respiratory infection cases presenting to the PED was analyzed, comparing two periods: Period 1 (October 2016–March 2017, n = 2181) and Period 2 (October 2023–March 2024, n = 2701). During Period 1, Group A Streptococcus and Respiratory Syncytial Virus rapid antigen detection tests were available. During Period 2, new point-of-care tests (POCTs), including rapid C-reactive protein and rapid antigen detection for Influenza A, Influenza B, and SARS-CoV-2, and a multiplex PCR nasal swab, were introduced. Results: In Period 2, antibiotic prescriptions decreased by 28.4%, along with a reduction in broad-spectrum antibiotic use. A significant correlation was observed between reduced antibiotic prescription and the use of new POCTs and multiplex PCR tests. Performance of blood tests and chest radiographs also decreased. Conclusions: Implementing novel diagnostic tests in PED helps clinicians select more appropriate management options with an impact on reduced stress and radiation exposure and antibiotic prescription. Full article

Background: The management of hard-to-heal wounds poses a major clinical challenge due to heterogeneous etiology and significant global healthcare costs (estimated at USD 148.64 billion in 2022). Large Language Models (LLMs), such as ChatGPT and Gemini, are emerging as potential decision-support tools. This study aimed to rigorously assess the accuracy and reliability of ChatGPT and Gemini in the visual description and initial therapeutic management of complex wounds based solely on clinical images. Methods: Twenty clinical images of complex wounds from diverse etiologies were independently analyzed by ChatGPT (version dated 15 October 2025) and Gemini (version dated 15 October 2025). The models were queried using two standardized, concise prompts. The AI responses were compared against a clinical gold standard established by the unanimous consensus of an expert panel of three plastic surgeons. Results: Statistical analysis showed no significant difference in overall performance between the two models and the expert consensus. Gemini achieved a slightly higher percentage of perfect agreement in management recommendations (75.0% vs. 60.0% for ChatGPT). Both LLMs demonstrated high proficiency in identifying the etiology of vascular lesions and recognizing critical “red flags,” such as signs of ischemia requiring urgent vascular assessment. Noted divergences included Gemini’s greater suspicion of potential neoplastic etiology and the models’ shared error in suggesting Negative Pressure Wound Therapy (NPWT) in a case potentially contraindicated by severe infection. Conclusions: LLMs, particularly ChatGPT and Gemini, demonstrate significant potential as decision-support systems and educational tools in wound care, offering rapid diagnosis and standardized initial management, especially in non-specialist settings. Instances of divergence in systemic treatments or in atypical presentations highlight the limitations of relying on image-based reasoning alone. Ultimately, LLMs serve as powerful, scalable assets that, under professional supervision, can enhance diagnostic speed and improve care pathways. Full article

Population aging is increasing dementia care demand. We present an audio-driven monitoring pipeline that operates either on mobile phones, microcontroller nodes, or smart television sets. The system combines audio signal processing with AI tools for structured interpretation. Preprocessing includes voice activity detection, speaker diarization, automatic speech recognition for dialogs, and speech-emotion recognition. An audio classifier detects home-care–relevant events (cough, cane taps, thuds, knocks, and speech). A large language model integrates transcripts, acoustic features, and a consented household knowledge base to produce a daily caregiver report covering orientation/disorientation (person, place, and time), delusion themes, agitation events, health proxies, and safety flags (e.g., exit seeking and falling). The pipeline targets real-time monitoring in homes and facilities, and it is an adjunct to caregiving, not a diagnostic device. Evaluation focuses on human-in-the-loop review, various audio/speech modalities, and the ability of AI to integrate information and reason. Intended users are low-income households in remote settings where in-person caregiving cannot be secured, enabling remote monitoring support for older adults with dementia. Full article