Hasil untuk "Internal medicine"

Anran Li, Lingfei Qian, Mengmeng Du et al.

Large Language Models (LLMs) have demonstrated significant potential in medicine, with many studies adapting them through continued pre-training or fine-tuning on medical data to enhance domain-specific accuracy and safety. However, a key open question remains: to what extent do LLMs memorize medical training data. Memorization can be beneficial when it enables LLMs to retain valuable medical knowledge during domain adaptation. Yet, it also raises concerns. LLMs may inadvertently reproduce sensitive clinical content (e.g., patient-specific details), and excessive memorization may reduce model generalizability, increasing risks of misdiagnosis and making unwarranted recommendations. These risks are further amplified by the generative nature of LLMs, which can not only surface memorized content but also produce overconfident, misleading outputs that may hinder clinical adoption. In this work, we present a study on memorization of LLMs in medicine, assessing its prevalence (how frequently it occurs), characteristics (what is memorized), volume (how much content is memorized), and potential downstream impacts (how memorization may affect medical applications). We systematically analyze common adaptation scenarios: (1) continued pretraining on medical corpora, (2) fine-tuning on standard medical benchmarks, and (3) fine-tuning on real-world clinical data, including over 13,000 unique inpatient records from Yale New Haven Health System. The results demonstrate that memorization is prevalent across all adaptation scenarios and significantly higher than that reported in the general domain. Moreover, memorization has distinct characteristics during continued pre-training and fine-tuning, and it is persistent: up to 87% of content memorized during continued pre-training remains after fine-tuning on new medical tasks.

Zheng Hui, Yijiang River Dong, Ehsan Shareghi et al.

As large language models (LLMs) are increasingly deployed in high-risk domains such as law, finance, and medicine, systematically evaluating their domain-specific safety and compliance becomes critical. While prior work has largely focused on improving LLM performance in these domains, it has often neglected the evaluation of domain-specific safety risks. To bridge this gap, we first define domain-specific safety principles for LLMs based on the AMA Principles of Medical Ethics, the ABA Model Rules of Professional Conduct, and the CFA Institute Code of Ethics. Building on this foundation, we introduce Trident-Bench, a benchmark specifically targeting LLM safety in the legal, financial, and medical domains. We evaluated 19 general-purpose and domain-specialized models on Trident-Bench and show that it effectively reveals key safety gaps -- strong generalist models (e.g., GPT, Gemini) can meet basic expectations, whereas domain-specialized models often struggle with subtle ethical nuances. This highlights an urgent need for finer-grained domain-specific safety improvements. By introducing Trident-Bench, our work provides one of the first systematic resources for studying LLM safety in law and finance, and lays the groundwork for future research aimed at reducing the safety risks of deploying LLMs in professionally regulated fields. Code and benchmark will be released at: https://github.com/zackhuiiiii/TRIDENT

William Troiani

We describe how finite colimits can be described using the internal lanuage, also known as the Mitchell-Benabou language, of a topos, provided the topos admits countably infinite colimits. This description is based on the set theoretic definitions of colimits and coequalisers, however the translation is not direct due to the differences between set theory and the internal language, these differences are described as internal versus external. Solutions to the hurdles which thus arise are given.

Chengfeng Dou, Ying Zhang, Zhi Jin et al.

Evidence-based medicine (EBM) plays a crucial role in the application of large language models (LLMs) in healthcare, as it provides reliable support for medical decision-making processes. Although it benefits from current retrieval-augmented generation~(RAG) technologies, it still faces two significant challenges: the collection of dispersed evidence and the efficient organization of this evidence to support the complex queries necessary for EBM. To tackle these issues, we propose using LLMs to gather scattered evidence from multiple sources and present a knowledge hypergraph-based evidence management model to integrate these evidence while capturing intricate relationships. Furthermore, to better support complex queries, we have developed an Importance-Driven Evidence Prioritization (IDEP) algorithm that utilizes the LLM to generate multiple evidence features, each with an associated importance score, which are then used to rank the evidence and produce the final retrieval results. Experimental results from six datasets demonstrate that our approach outperforms existing RAG techniques in application domains of interest to EBM, such as medical quizzing, hallucination detection, and decision support. Testsets and the constructed knowledge graph can be accessed at \href{https://drive.google.com/file/d/1WJ9QTokK3MdkjEmwuFQxwH96j_Byawj_/view?usp=drive_link}{https://drive.google.com/rag4ebm}.

Svyatoslav M. Yurin, Dmitry A. Apalkov, Tatiana A. Minenkova et al.

Background. Congenital disorders of sex development (DSD) represent a heterogeneous group of dysontogenetic conditions characterized by a discordance between chromosomal, gonadal, and phenotypic sex. Among them, idiopathic forms of male pseudohermaphroditism with a 46,XY karyotype are among the most diagnostically challenging variants, as the presence of testicular tissue is accompanied by incomplete masculinization of the external genitalia in the absence of detectable mutations within the androgen-regulatory gene system. Given the clinico-endocrinological ambiguity of this pathology, an integrative diagnostic strategy combining hormonal, cytogenetic, and morphofunctional assessments, along with the timely determination of optimal timing for surgical and hormonal correction, acquires particular clinical importance. Objective. To perform a detailed clinico-endocrinological characterization of the idiopathic variant of DSD in a prepubertal child with a 46,XY karyotype and to determine the principles of rational diagnostic and therapeutic management. Materials and methods. The study is based on the clinical observation of a 9-year-old boy examined in the Endocrinology Department of the Kursk Regional Children’s Clinical Hospital. The analysis included medical history, physical status, serum levels of LH, FSH, testosterone, and anti-Müllerian hormone, cytogenetic data, and ultrasonographic features of the gonads and pelvic organs, correlated with up-to-date literature sources. Conclusions. Idiopathic forms of male pseudohermaphroditism with a normal male karyotype require a multidisciplinary approach and prolonged follow-up. Early surgical correction and subsequent hormonal monitoring contribute to the formation of an adequate phenotypic outcome, reduction of endocrine complications, and improvement of psychosocial adaptation during puberty.

Suvijak Untaaveesup, MD, Thipsukon Amnartpanich, MD, Noraworn Jirattikanwong, MD et al.

Background: Chronic systemic inflammation in individuals with moderate-to-severe atopic dermatitis (AD) potentially predisposes them to metabolic and cardiovascular diseases. Nevertheless, evidence with regard to such association is limited. Objective: To assess the association between metabolic and cardiovascular outcomes and moderate-to-severe AD. Methods: A systematic search was performed through PubMed, Scopus, EMBASE, and Cochrane for population-based studies that addressed the effects of moderate-to-severe AD on metabolic and cardiovascular outcomes compared with the general population from inception to August 31, 2023. Meta-analysis was performed using the random effects model. The pooled odds ratio (OR) and certainty of evidence for each outcome were reported. Results: We included 11 studies, 4 retrospective cohorts, 1 prospective cohort, 4 cross-sectional, and 2 case-control studies involving 405,170 moderate-to-severe AD patients compared to 4,591,478 unaffected controls. Moderate-to-severe AD was associated with a higher risk of myocardial infarction with an OR (95% CI) of 1.33 (1.07, 1.65), angina 1.33 (1.06, 1.66), heart failure 1.56 (1.28, 1.90), stroke 1.45 (1.21, 1.74), hypertension 1.38 (1.18, 1.63), dyslipidemia 1.27 (1.15, 1.41), and metabolic syndrome 1.24 (1.05, 1.42) with very low certainty of evidence. No significantly increased risk of cardiovascular death with an odds ratio (95% CI) of 1.81 (0.96, 3.44) and diabetes of 1.24 (0.91, 1.68) was observed. High heterogeneity was observed in most studies for all of the outcomes. Conclusion: Our meta-analysis demonstrated a modest but significant association between moderate-to-severe AD and increased susceptibility to metabolic and cardiovascular diseases. Initial assessment of cardiovascular and metabolic risk for patients with moderate-to-severe AD should be considered to enable early management strategies.

Bhargav Teja Nallapu, Ali Ezzati, Helena M. Blumen et al.

ABSTRACT INTRODUCTION Understanding the heterogeneity of brain structure in individuals with the Motoric Cognitive Risk Syndrome (MCR) may improve the current risk assessments of dementia. METHODS We used data from six cohorts from the MCR consortium (N = 1987). A weakly‐supervised clustering algorithm called HYDRA (Heterogeneity through Discriminative Analysis) was applied to volumetric magnetic resonance imaging (MRI) measures to identify distinct subgroups in the population with gait speeds lower than one standard deviation (1SD) above mean. RESULTS Three subgroups (Groups A, B, and C) were identified through MRI‐based clustering with significant differences in regional brain volumes, gait speeds, and performance on Trail Making (Part‐B) and Free and Cued Selective Reminding Tests. DISCUSSION Based on structural MRI, our results reflect heterogeneity in the population with moderate and slow gait, including those with MCR. Such a data‐driven approach could help pave new pathways toward dementia at‐risk stratification and have implications for precision health for patients. Highlights Different patterns of brain atrophy were observed among the people with moderate and slow gait speeds Slower gait speeds were associated with substantial cortical atrophy, higher rates of Motoric Cognitive Risk Syndrome (MCR), and worse cognitive performance This approach can aid patient stratification at early asymptomatic stages and have implications for precision health.

Tsion Michael, Solomon Moges Demeke

IntroductionCommon mental disorders (CMDs) and suicidality are two of the most common psychological and mental health issues associated with acute and chronic sexual and gender-based violence (SGBV). Thus, the purpose of this study was to determine the magnitude of symptoms of CMDs, and suicidality among females experienced SGBV in Ethiopia.MethodA cross-sectional study was conducted among 407 female survivors of SGBV in the One Stop Centers of the Amhara region. Data analysis was performed using SPSS version 25. The odds ratio at a p-value of 0.05 was used to determine the strength of the association of the independent variables with CMDs and suicidality.ResultsA total of 407 women participated in the study. Suicidality was reported by a quarter of the survivors (24.1%), while CMDs were reported by nearly two-thirds (61.7%). Being widowed (AOR = 3.0, 95% CI = 3.0 [1.22, 7.66]), having a family history of mental illnesses (AOR = 7.1, 95% CI = 7.1 [4.07, 12.39)], being low-income (AOR = 2.8, 95% CI = 2.8 [1.64, 5.06]), and current drug use (AOR = 2.9, 95% CI = 2.9 [1.63, 5.16]) were all linked with CMDs. Having a history of abortion (AOR = 4.1, 95% CI = 4.1 [1.9, 8.5]), CMDs (AOR = 4.6, 95% CI = 4.6 [2.0, 10.74]), and history of suicide (AOR = 3.41, 95% CI = 3.41 [1.22, 9.55]) were some of the characteristics that were substantially linked with suicidality.ConclusionFemales with SGBV had a high prevalence of CMDs and suicidality and calls for comprehensive remedies.

Sibo Wei, Xueping Peng, Yi-Fei Wang et al.

The surge of large language models (LLMs) has driven significant progress in medical applications, including traditional Chinese medicine (TCM). However, current medical LLMs struggle with TCM diagnosis and syndrome differentiation due to substantial differences between TCM and modern medical theory, and the scarcity of specialized, high-quality corpora. To this end, in this paper we propose BianCang, a TCM-specific LLM, using a two-stage training process that first injects domain-specific knowledge and then aligns it through targeted stimulation to enhance diagnostic and differentiation capabilities. Specifically, we constructed pre-training corpora, instruction-aligned datasets based on real hospital records, and the ChP-TCM dataset derived from the Pharmacopoeia of the People's Republic of China. We compiled extensive TCM and medical corpora for continual pre-training and supervised fine-tuning, building a comprehensive dataset to refine the model's understanding of TCM. Evaluations across 11 test sets involving 31 models and 4 tasks demonstrate the effectiveness of BianCang, offering valuable insights for future research. Code, datasets, and models are available on https://github.com/QLU-NLP/BianCang.

Ahmed Abdeen Hamed, Tamer E. Fandy

The objective of this research is to introduce a network specialized in predicting drugs that can be repurposed by investigating real-world evidence sources, such as clinical trials and biomedical literature. Specifically, it aims to generate drug combination therapies for complex diseases (e.g., cancer, Alzheimer's). We present a multilayered network medicine approach, empowered by a highly configured ChatGPT prompt engineering system, which is constructed on the fly to extract drug mentions in clinical trials. Additionally, we introduce a novel algorithm that connects real-world evidence with disease-specific signaling pathways (e.g., KEGG database). This sheds light on the repurposability of drugs if they are found to bind with one or more protein constituents of a signaling pathway. To demonstrate, we instantiated the framework for breast cancer and found that, out of 46 breast cancer signaling pathways, the framework identified 38 pathways that were covered by at least two drugs. This evidence signals the potential for combining those drugs. Specifically, the most covered signaling pathway, ID hsa:2064, was covered by 108 drugs, some of which can be combined. Conversely, the signaling pathway ID hsa:1499 was covered by only two drugs, indicating a significant gap for further research. Our network medicine framework, empowered by GenAI, shows promise in identifying drug combinations with a high degree of specificity, knowing the exact signaling pathways and proteins that serve as targets. It is noteworthy that ChatGPT successfully accelerated the process of identifying drug mentions in clinical trials, though further investigations are required to determine the relationships among the drug mentions.

Kishi Kobe Yee Francisco, Andrane Estelle Carnicer Apuhin, Myles Joshua Toledo Tan et al.

Personalized medicine (PM) promises to transform healthcare by providing treatments tailored to individual genetic, environmental, and lifestyle factors. However, its high costs and infrastructure demands raise concerns about exacerbating health disparities, especially between high-income countries (HICs) and low- and middle-income countries (LMICs). While HICs benefit from advanced PM applications through AI and genomics, LMICs often lack the resources necessary to adopt these innovations, leading to a widening healthcare divide. This paper explores the financial and ethical challenges of PM implementation, with a focus on ensuring equitable access. It proposes strategies for global collaboration, infrastructure development, and ethical frameworks to support LMICs in adopting PM, aiming to prevent further disparities in healthcare accessibility and outcomes.

Anukul Ghimire, Samveg Shah, Utkarsh Chauhan et al.

Abstract Background There is a lack of contemporary data describing global variations in vascular access for hemodialysis (HD). We used the third iteration of the International Society of Nephrology Global Kidney Health Atlas (ISN-GKHA) to highlight differences in funding and availability of hemodialysis accesses used for initiating HD across world regions. Methods Survey questions were directed at understanding the funding modules for obtaining vascular access and types of accesses used to initiate dialysis. An electronic survey was sent to national and regional key stakeholders affiliated with the ISN between June and September 2022. Countries that participated in the survey were categorized based on World Bank Income Classification (low-, lower-middle, upper-middle, and high-income) and by their regional affiliation with the ISN. Results Data on types of vascular access were available from 160 countries. Respondents from 35 countries (22% of surveyed countries) reported that > 50% of patients started HD with an arteriovenous fistula or graft (AVF or AVG). These rates were higher in Western Europe (n = 14; 64%), North & East Asia (n = 4; 67%), and among high-income countries (n = 24; 38%). The rates of > 50% of patients starting HD with a tunneled dialysis catheter were highest in North America & Caribbean region (n = 7; 58%) and lowest in South Asia and Newly Independent States and Russia (n = 0 in both regions). Respondents from 50% (n = 9) of low-income countries reported that > 75% of patients started HD using a temporary catheter, with the highest rates in Africa (n = 30; 75%) and Latin America (n = 14; 67%). Funding for the creation of vascular access was often through public funding and free at the point of delivery in high-income countries (n = 42; 67% for AVF/AVG, n = 44; 70% for central venous catheters). In low-income countries, private and out of pocket funding was reported as being more common (n = 8; 40% for AVF/AVG, n = 5; 25% for central venous catheters). Conclusions High income countries exhibit variation in the use of AVF/AVG and tunneled catheters. In low-income countries, there is a higher use of temporary dialysis catheters and private funding models for access creation.

Masanobu Kogure, Nobuhisa Kanahara, Atsuhiro Miyazawa et al.

BackgroundMost genetic analyses that have attempted to identify a locus or loci that can distinguish patients with treatment-resistant schizophrenia (TRS) from those who respond to treatment (non-TRS) have failed. However, evidence from multiple studies suggests that patients with schizophrenia who respond well to antipsychotic medication have a higher dopamine (DA) state in brain synaptic clefts whereas patients with TRS do not show enhanced DA synthesis/release pathways.Patients and methodsTo examine the contribution (if any) of genetics to TRS, we conducted a genetic association analysis of DA-related genes in schizophrenia patients (TRS, n = 435; non-TRS, n = 539) and healthy controls (HC: n = 489).ResultsThe distributions of the genotypes of rs3756450 and the 40-bp variable number tandem repeat on SLC6A3 differed between the TRS and non-TRS groups. Regarding rs3756450, the TRS group showed a significantly higher ratio of the A allele, whereas the non-TRS group predominantly had the G allele. The analysis of the combination of COMT and SLC6A3 yielded a significantly higher ratio of the putative low-DA type (i.e., high COMT activity + high SLC6A3 activity) in the TRS group compared to the two other groups. Patients with the low-DA type accounted for the minority of the non-TRS group and exhibited milder psychopathology.ConclusionThe overall results suggest that (i) SLC6A3 could be involved in responsiveness to antipsychotic medication and (ii) genetic variants modulating brain DA levels may be related to the classification of TRS and non-TRS.

Louis Martini, Sebastian Wolf

We develop the theory of topoi internal to an arbitrary $\infty$-topos $\mathcal B$. We provide several characterisations of these, including an internal analogue of Lurie's characterisation of $\infty$-topoi, but also a description in terms of the underlying sheaves of $\infty$-categories, and we prove a number of structural results about these objects. Furthermore, we show that the $\infty$-category of topoi internal to $\mathcal B$ is equivalent to the $\infty$-category of $\infty$-topoi over $\mathcal B$, and use this result to derive a formula for the pullback of $\infty$-topoi. Lastly, we use our theory to relate smooth geometric morphisms of $\infty$-topoi to internal locally contractible topoi.

Kostya Druzhkov

A description of how the principle of stationary action reproduces itself in terms of the intrinsic geometry of variational equations is proposed. A notion of stationary points of an internal Lagrangian is introduced. A connection between symmetries, conservation laws and internal Lagrangians is established. Noether's theorem is formulated in terms of internal Lagrangians. A relation between non-degenerate Lagrangians and the corresponding internal Lagrangians is investigated. Several examples are discussed.

Jie Xu, Lu Lu, Sen Yang et al.

METHODS: First, a set of evaluation criteria is designed based on a comprehensive literature review. Second, existing candidate criteria are optimized for using a Delphi method by five experts in medicine and engineering. Third, three clinical experts design a set of medical datasets to interact with LLMs. Finally, benchmarking experiments are conducted on the datasets. The responses generated by chatbots based on LLMs are recorded for blind evaluations by five licensed medical experts. RESULTS: The obtained evaluation criteria cover medical professional capabilities, social comprehensive capabilities, contextual capabilities, and computational robustness, with sixteen detailed indicators. The medical datasets include twenty-seven medical dialogues and seven case reports in Chinese. Three chatbots are evaluated, ChatGPT by OpenAI, ERNIE Bot by Baidu Inc., and Doctor PuJiang (Dr. PJ) by Shanghai Artificial Intelligence Laboratory. Experimental results show that Dr. PJ outperforms ChatGPT and ERNIE Bot in both multiple-turn medical dialogue and case report scenarios.

Xiang Tang, Guoqing Liu, Li Lin et al.

Abstract Background Neurodevelopmental disorder with absent language and variable seizures (NEDALVS, # 618707) are characterized by delayed speech and motor development, ocular abnormalities, and seizures. NEDAVLS is an autosomal dominant disorder caused by de novo mutations in the wasp protein family member 1 (WASF1) gene. Case presentation We identified a de novo nonsense variant c.1516 C > T (p.Arg506*) of WASF1 gene (NM_003931.3) in two pediatric female patients with delayed motor and language development. Conclusion This case demonstrates the effective role of WES in the diagnosis of NEDALVS. To the best of our knowledge, this variant has not been reported in the Chinese population. This contributes to our further understanding of the disease and to research related to the genetic and clinical heterogeneity, the treatment and prognosis of the disease.

Haim Shmuely, Yekaterina Shvartsman, Rita Berdinstein et al.

Abstract Purpose During the last few decades, the increased use of various types of antibiotics in the general population caused a significant change in regional Helicobacter pylori (H. pylori) antibiotic resistance. Our aim is to study the changes in H. pylori resistance in patients who had undergone an esophagogastroduodenoscopy (EGD) and susceptibility testing and found positive for H. pylori. The study was conducted in a university affiliated hospital between 2013–2020. Methods A cross-sectional study was performed on all consecutive patients who had undergone an EGD and tested positive for H. pylori at the Kaplan Medical Center, Israel. The study period was divided into two sub-periods: 2013–2016 and 2017–2020. Data on age, sex, comorbidities, previous treatments, and antimicrobial susceptibility testing for six antimicrobial agents were compared. Results The resistance rates of H. pylori to clarithromycin and dual resistance to clarithromycin and metronidazole were found significantly higher during the late period. Multivariable analysis showed that the later period, older age, and diabetes mellitus were independent predictors for antimicrobial resistance. Conclusions Our study has shown that there is an increasing resistance of H. pylori to clarithromycin and metronidazole while its susceptibility is unaffected with time to other antibiotics. More recent cross-sectional studies with larger samples are warranted in order to evaluate the changes in the resistance patterns of H. pylori to various antibiotics with time.

Rehanguli Maimaitituerxun, Wenhang Chen, Jingsha Xiang et al.

Abstract Background Depression and diabetes are major health challenges, with heavy economic social burden, and comorbid depression in diabetes could lead to a wide range of poor health outcomes. Although many descriptive studies have highlighted the prevalence of comorbid depression and its associated factors, the situation in Hunan, China, remains unclear. Therefore, this study aimed to identify the prevalence of comorbid depression and associated factors among hospitalized type 2 diabetes mellitus (T2DM) patients in Hunan, China. Methods This cross-sectional study involved 496 patients with T2DM who were referred to the endocrinology inpatient department of Xiangya Hospital affiliated to Central South University, Hunan. Participants’ data on socio-demographic status, lifestyle factors, T2DM-related characteristics, and social support were collected. Depression was evaluated using the Hospital Anxiety and Depression Scale-depression subscale. All statistical analyses were conducted using the R software version 4.2.1. Results The prevalence of comorbid depression among hospitalized T2DM patients in Hunan was 27.22% (95% Confidence Interval [CI]: 23.3–31.1%). Individuals with depression differed significantly from those without depression in age, educational level, per capita monthly household income, current work status, current smoking status, current drinking status, regular physical activity, duration of diabetes, hypertension, chronic kidney disease, stroke, fatty liver, diabetic nephropathy, diabetic retinopathy, insulin use, HbA1c, and social support. A multivariable logistic regression model showed that insulin users (adjusted OR = 1.86, 95% CI: 1.02–3.42) had a higher risk of depression, while those with regular physical activity (adjusted OR = 0.48, 95% CI: 0.30–0.77) or greater social support (adjusted OR = 0.20, 95% CI: 0.11–0.34) had a lower risk of depression. The area under the curve of the receiver operator characteristic based on this model was 0.741 with a sensitivity of 0.785 and specificity of 0.615. Conclusions Depression was moderately prevalent among hospitalized T2DM patients in Hunan, China. Insulin treatment strategies, regular physical activity, and social support were significantly independently associated with depression, and the multivariable model based on these three factors demonstrated good predictivity, which could be applied in clinical practice.

Nelson Martins-Ferreira

An answer to the question investigated in this paper brings a new characterization of internal groupoids such that: (a) it holds even when finite limits are not assumed to exist; (b) it is a full subcategory of the category of involutive-2-links, that is, a category whose objects are morphisms equipped with a pair of interlinked involutions. This result highlights the fact that even thought internal groupoids are internal categories equipped with an involution, they can equivalently be seen as tri-graphs with an involution. Moreover, the structure of a tri-graph with an involution can be further contracted into a simpler structure consisting of one morphism with two interlinked involutions. This approach highly contrasts with the one where groupoids are seen as reflexive graphs on which a multiplicative structure is defined with inverses.