Hasil untuk "Internal medicine"

Jihoon Jeong

Small language models (SLMs) in the 100M-10B parameter range increasingly power production systems, yet whether they possess the internal emotion representations recently discovered in frontier models remains unknown. We present the first comparative analysis of emotion vector extraction methods for SLMs, evaluating 9 models across 5 architectural families (GPT-2, Gemma, Qwen, Llama, Mistral) using 20 emotions and two extraction methods (generation-based and comprehension-based). Generation-based extraction produces statistically superior emotion separation (Mann-Whitney p = 0.007; Cohen's d = -107.5), with the advantage modulated by instruction tuning and architecture. Emotion representations localize at middle transformer layers (~50% depth), following a U-shaped curve that is architecture-invariant from 124M to 3B parameters. We validate these findings against representational anisotropy baselines across 4 models and confirm causal behavioral effects through steering experiments, independently verified by an external emotion classifier (92% success rate, 37/40 scenarios). Steering reveals three regimes -- surgical (coherent text transformation), repetitive collapse, and explosive (text degradation) -- quantified by perplexity ratios and separated by model architecture rather than scale. We document cross-lingual emotion entanglement in Qwen, where steering activates semantically aligned Chinese tokens that RLHF does not suppress, raising safety concerns for multilingual deployment. This work provides methodological guidelines for emotion research on open-weight models and contributes to the Model Medicine series by bridging external behavioral profiling with internal representational analysis.

Xinyue Wang, Liming Li, Michael Roman et al.

With its extreme axial tilt, radiant energy budget and internal heat of Uranus remain among the most intriguing mysteries of our Solar System. Here, we present the global radiant energy budget spanning a complete orbital period, revealing significant seasonal variations driven primarily by the highly variable solar flux. Despite these fluctuations, emitted thermal power consistently exceeds absorbed solar power, indicating a net energy loss and ongoing global cooling. Based on the seasonal variations of radiant energy budget, we determine a statistically significant internal heat flux. This finding resolves a long-standing debate over whether Uranus possesses internal heat. We also examine the energy budget of the weather layer by combining the internal heat with the radiant energies, revealing significant energy imbalances at both global and hemispheric scales. These global and hemispheric imbalances should be considered in theoretical and numerical models. The Uranus flagship mission, as recommended by the recent survey, will provide crucial observations to address more unresolved questions and advance our understanding of this enigmatic ice giant.

Qingjuan Chen, Zhongqiang Yao, Jianfeng Duan et al.

Abstract This study aimed to investigate the potential value of circulating miRNAs in the diagnosis of peritoneal carcinomatosis (PC) in patients with gastric cancer (GC). A quantitative reverse-transcription polymerase chain reaction (qRT-PCR) method was optimized for the measurement of plasma miRNAs. The concentrations of 11 plasma miRNA transcripts were analyzed in 13 pairs of GC patients with PC (GC/PC) and without PC (GC/NPC) using qRT-PCR. The plasma levels of miR-9 and miR-106a were further validated in 30 pairs of GC/PC and GC/NPC patients, as well as in 35 healthy controls (HC), followed by receiver operating characteristic (ROC) curve analysis. Serum levels of carbohydrate antigen 125 (CA125) and carcinoembryonic antigen (CEA) were also measured. Primary screening and further validation revealed significantly lower plasma miR-9 levels in the GC/PC group than in the GC/NPC and HC groups (p < 0.001). Conversely, plasma miR-106a and serum CA125 levels, but not CEA levels, were significantly higher in the GC/PC group than in the GC/NPC and HC groups (p < 0.001). No significant differences were observed in plasma miR-9 and miR-106a levels between the GC/NPC and HC groups. ROC analyses indicated that plasma miR-9 yielded an area under the curve (AUC) of 0.776 (95% confidence interval [CI] 0.673–0.859, p < 0.001) with 67.4% sensitivity and 93% specificity, and miR-106a had an AUC of 0.830 (95% CI 0.743–0.916, p < 0.001) with 72.1% sensitivity and 83.7% specificity in distinguishing the GC/PC group from the GC/NPC group. Moreover, the diagnostic performances of plasma miR-9 and miR-106a were comparable with that of CA125 (p > 0.05). Kaplan–Meier survival analysis demonstrated that high plasma levels of miR-106a (hazard ratio (HR) = 0.44, 95% CI 0.19–1.00, p = 0.040) and low levels of miR-9 (HR = 0.43, 95% CI 0.18–1.02, p = 0.042) were significantly associated with reduced overall survival in GC/PC patients. Taken together, these findings suggest that plasma miR-9 and miR-106a may serve as promising non-invasive biomarkers for both the diagnosis and prognosis of PC in patients with GC.

Selene S Mak, Laura M Nally, Juanita Montoya et al.

BackgroundNumerous barriers to moderate to vigorous physical activity exist for youths with type 1 diabetes (T1D). The virtual exercise games for youth with T1D (ExerT1D) intervention implement synchronous support of moderate to vigorous physical activity including T1D peers and role models. ObjectiveThis study aims to understand the acceptability of this intervention to participants. MethodsWe conducted postprogram, semistructured, televideo interviews with participating youths to elicit perspectives on the acceptability of the intervention and experience with the program. Two coders independently reviewed and analyzed each transcript using a coding scheme developed inductively by senior researchers. Discrepancies were resolved by team discussion, and multiple codes were grouped together to produce 4 main thematic areas. ResultsAll 15 participants provided interviews (aged 14-19 years; 2 nonbinary, 6 females; median hemoglobin A1c level of 7.8% (IQR 7.4%-11.2%), 5 with a hemoglobin A1c level of ≥10%). Qualitative data revealed four themes: (1) motivation to engage in physical activity (PA)—improving their physical capabilities and stabilizing glucose levels were cited as motivation for PA and challenges of living with T1D were cited as PA barriers; (2) experience with and motivation to manage diabetes while engaging in PA—participants provided details of accommodating the inherent uncertainty or limitations of PA with diabetes and sometimes preparing for PA involved psychological and motivational adjustments while some relayed feelings of avoidance; (3) peer support encouraged engagement with the intervention—participants appreciated the peer aspects of components of ExerT1D and participants’ reflections of the facilitated group experience highlight many benefits of a small-group virtual program; and (4) improvements in PA and diabetes self-management efficacy—all participants credited the program with improving or at least raising awareness of T1D management skills. ConclusionsOur virtual PA intervention using an active video game and discussion component provided adolescents with T1D the confidence and peer support to engage in PA, improved awareness of diabetes-specific tasks to prepare for exercise, and improved understanding of the effect of PA on glucose levels. Engaging youths with a virtual video game intervention is a viable approach to overcome barriers to PA for adolescents with T1D. Trial RegistrationClinicalTrials.gov NCT05163912; https://clinicaltrials.gov/ct2/show/NCT05163912

Jinesh Mukesh Shah, Nijanth Manohararaj, Koh Yeow Hoay

We aim to report an extremely rare case of a primary thoracic intramedullary cavernoma with Brown-Sequard syndrome (BSS), its transcranial magnetic stimulation (TMS)/somatosensory evoked potential (SSEP) neurophysiology tests, and their localizing value. A 53-year-old Chinese male with a history of multiple arteriovenous malformations (AVMs) presented with an intermittent 3-year history of the left lower limb weakness with recent worsening and findings of dissociated sensory loss. Neurophysiological testing showed prolonged central motor conduction time to his left lower limb on TMS while tibial SSEP showed prolonged P37 latencies. Magnetic resonance imaging spine showed a T4-5 intramedullary expansile enhancing cord lesion, suggestive of a thoracic cavernoma, with surrounding acute hematomyelia and cord edema from C7 to T6. A spinal angiogram did not reveal any vascular malformation. He was conservatively treated for possible T4-5 cavernoma with hematomyelia. Repeat imaging showed complete resolution of edema with a T3-5 internal T2-weighted hyperintensity and residual susceptibility focus likely representing a cavernoma that had bled with no evidence of AVM. A repeat tibial SSEP still showed prolonged tibial SSEPs, but TMS was now normal. Primary thoracic intramedullary cavernomas may be a rare cause of BSS. TMS and SSEP may have a role in the diagnostic evaluation of BSS.

Qingfeng Sun, Kai Zhang, Yuanlong Xu et al.

Abstract Background HIV/TB co-infection presents substantial public-health challenges, showing greater treatment-failure and mortality rates than tuberculosis alone. Recent advances in machine learning (ML) provide a robust means of identifying high-risk patients early in the disease course. Methods This retrospective study enrolled 359 patients co-infected with HIV and TB at a single tertiary-care hospital. We extracted clinical and immunological data. The cohort was subsequently divided into training (0%) and test (0%) subsets, and class imbalance was addressed with the Synthetic Minority Over-sampling Technique (SMOTE). Six ML classifiers—Random Forest, XGBoost, LightGBM, Support Vector Machine, Extra Trees and CatBoost—were trained after grid-search hyper-parameter tuning. Model performance was assessed with the area under the receiver-operating-characteristic curve (AUC), accuracy, recall, precision, specificity and F1-score. Multi-criteria ranking was then conducted with the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS). The leading model was interpreted using SHapley Additive exPlanations (SHAP). Results Overall, 304 of 359 patients (84.7%) had favourable outcomes, whereas 55 (15.3%) had unfavourable outcomes. LightGBM achieved the best overall performance (AUC = 0.771; accuracy = 84.72%; F1 = 0.522) and was ranked first by TOPSIS. SHAP analysis highlighted age, CD4 and CD8 counts, body-mass index and occupation as key predictors. Lower BMI, pronounced immunosuppression and older age were strongly associated with unfavourable outcomes, findings that align with established clinical evidence. Conclusion A gradient-boosted model (LightGBM) combined with SHAP interpretation demonstrated reliable predictive performance in HIV/TB co-infection and highlighted clinically actionable risk factors. Incorporating this tool into routine workflows could enable healthcare providers to identify high-risk individuals earlier, allocate resources more efficiently and, ultimately, improve TB-treatment success. Clinical trial registration Not applicable.

Minda Hu, Qiyuan Zhang, Yufei Wang et al.

As a crucial step to enhance LLMs alignment with human intentions, Instruction Fine-Tuning (IFT) has a high demand on dataset quality. However, existing IFT datasets often contain knowledge that is inconsistent with LLMs' internal knowledge learned from the pre-training phase, which can greatly affect the efficacy of IFT. To address this issue, we introduce NILE (iNternal consIstency aLignmEnt) framework, aimed at optimizing IFT datasets to unlock LLMs' capability further. NILE operates by eliciting target pre-trained LLM's internal knowledge corresponding to instruction data. The internal knowledge is leveraged to revise the answer in IFT datasets. Additionally, we propose a novel Internal Consistency Filtering (ICF) method to filter training samples, ensuring its high consistency with LLM's internal knowledge. Our experiments demonstrate that NILE-aligned IFT datasets sharply boost LLM performance across multiple LLM ability evaluation datasets, achieving up to 66.6% gain on Arena-Hard and 68.5% on Alpaca-Eval V2. Further analysis confirms that each component of the NILE}framework contributes to these substantial performance improvements, and provides compelling evidence that dataset consistency with pre-trained internal knowledge is pivotal for maximizing LLM potential.

Chaoyi Wu, Pengcheng Qiu, Jinxin Liu et al.

In this study, we present MedS-Bench, a comprehensive benchmark designed to evaluate the performance of large language models (LLMs) in clinical contexts. Unlike existing benchmarks that focus on multiple-choice question answering, MedS-Bench spans 11 high-level clinical tasks, including clinical report summarization, treatment recommendations, diagnosis, named entity recognition, and medical concept explanation, among others. We evaluated six leading LLMs, e.g., MEDITRON, Mistral, InternLM 2, Llama 3, GPT-4, and Claude-3.5 using few-shot prompting, and found that even the most sophisticated models struggle with these complex tasks. To address these limitations, we developed MedS-Ins, a large-scale instruction tuning dataset for medicine. MedS-Ins comprises 58 medically oriented language corpora, totaling 13.5 million samples across 122 tasks. To demonstrate the dataset's utility, we conducted a proof-of-concept experiment by performing instruction tuning on a lightweight, open-source medical language model. The resulting model, MMedIns-Llama 3, significantly outperformed existing models across nearly all clinical tasks. To promote further advancements in the application of LLMs to clinical challenges, we have made the MedS-Ins dataset fully accessible and invite the research community to contribute to its expansion.Additionally, we have launched a dynamic leaderboard for MedS-Bench, which we plan to regularly update the test set to track progress and enhance the adaptation of general LLMs to the medical domain. Leaderboard: https://henrychur.github.io/MedS-Bench/. Github: https://github.com/MAGIC-AI4Med/MedS-Ins.

Paris Kantaras, Niki Mourouti, Theodora Mouratidou et al.

In total, 3274 adults (65.2% females) from six European countries were included in this cross-sectional analysis using data from the baseline assessment of the Feel4Diabetes study. Anthropometric, sociodemographic, dietary and behavioral data were assessed, and the existence of metabolic syndrome (MetS) was recorded. Beverage consumption patterns (BCPs) were derived via principal component analysis. Three BCPs were derived explaining 39.5% of the total variation. BCP1 was labeled as “Alcoholic beverage pattern”, which loaded heavily on high consumption of beer/cider, wine and other spirits; BCP2 was labeled as “High in sugars beverage pattern” that was mainly characterized by high consumption of soft drinks with sugar, juice containing sugar and low consumption of water; and BCP3 was labeled as “Healthy beverage pattern” that was mainly characterized by high consumption of water, tea, fruit juice freshly squeezed or prepacked without sugar and low consumption of soft drinks without sugar. After adjusting for various confounders, BCP2 was positively associated with elevated triglycerides (<i>p</i> = 0.001), elevated blood pressure (<i>p</i> = 0.001) elevated fasting glucose (<i>p</i> = 0.008) and the existence of MetS (<i>p</i> = 0.006), while BCP1 was inversely associated with reduced HDL-C (<i>p</i> = 0.005) and BCP3 was inversely associated with elevated blood pressure (<i>p</i> = 0.047). The establishment of policy actions as well as public health nutritional education can contribute to the promotion of a healthy beverage consumption.

Ethan R. Roy, Sanming Li, Sepideh Saroukhani et al.

Abstract Background Aging significantly elevates the risk of developing neurodegenerative diseases. Neuroinflammation is a universal hallmark of neurodegeneration as well as normal brain aging. Which branches of age-related neuroinflammation, and how they precondition the brain toward pathological progression, remain ill-understood. The presence of elevated type I interferon (IFN-I) has been documented in the aged brain, but its role in promoting degenerative processes, such as the loss of neurons in vulnerable regions, has not been studied in depth. Methods To comprehend the scope of IFN-I activity in the aging brain, we surveyed IFN-I-responsive reporter mice at multiple ages. We also examined 5- and 24-month-old mice harboring selective ablation of Ifnar1 in microglia to observe the effects of manipulating this pathway during the aging process using bulk RNA sequencing and histological parameters. Results We detected age-dependent IFN-I signal escalation in multiple brain cell types from various regions, especially in microglia. Selective ablation of Ifnar1 from microglia in aged mice significantly reduced overall brain IFN-I signature, dampened microglial reactivity, lessened neuronal loss, restored expression of key neuronal genes and pathways, and diminished the accumulation of lipofuscin, a core hallmark of cellular aging in the brain. Conclusions Overall, our study demonstrates pervasive IFN-I activity during normal mouse brain aging and reveals a pathogenic, pro-degenerative role played by microglial IFN-I signaling in perpetuating neuroinflammation, neuronal dysfunction, and molecular aggregation. These findings extend the understanding of a principal axis of age-related inflammation in the brain, one likely shared with multiple neurological disorders, and provide a rationale to modulate aberrant immune activation to mitigate neurodegenerative process at all stages.

José Nunes de Alencar

Chaoyi Wu, Weixiong Lin, Xiaoman Zhang et al.

Recently, Large Language Models (LLMs) have showcased remarkable capabilities in natural language understanding. While demonstrating proficiency in everyday conversations and question-answering situations, these models frequently struggle in domains that require precision, such as medical applications, due to their lack of domain-specific knowledge. In this paper, we describe the procedure for building a powerful, open-source language model specifically designed for medicine applications, termed as PMC-LLaMA. Our contributions are threefold: (i) we systematically investigate the process of adapting a general-purpose foundation language model towards medical domain, this involves data-centric knowledge injection through the integration of 4.8M biomedical academic papers and 30K medical textbooks, as well as comprehensive fine-tuning for alignment with domain-specific instructions; (ii) we contribute a large-scale, comprehensive dataset for instruction tuning. This dataset encompasses medical question-answering (QA), rationale for reasoning, and conversational dialogues, comprising a total of 202M tokens; (iii) we conduct thorough ablation studies to demonstrate the effectiveness of each proposed component. While evaluating on various public medical question-answering benchmarks, our lightweight PMCLLaMA, which consists of only 13 billion parameters, exhibits superior performance, even surpassing ChatGPT. All models, codes, datasets can be found in https://github.com/chaoyi-wu/PMC-LLaMA.

Leo Law, BingKan Xue

In varying environments it is beneficial for organisms to utilize available cues to infer the conditions they may encounter and express potentially favorable traits. However, external cues can be unreliable or too costly to use. We consider an alternative strategy where organisms exploit internal sources of information. Even without sensing environmental cues, their internal states may become correlated with the environment as a result of selection, which then form a memory that helps predict future conditions. To demonstrate the adaptive value of such internal memory in varying environments, we revisit the classic example of seed dormancy in annual plants. Previous studies have considered the germination fraction of seeds and its dependence on environmental cues. In contrast, we consider a model of germination fraction that depends on the seed age, which is an internal state that can serve as a memory. We show that, if the environmental variation has temporal structure, then age-dependent germination fractions will allow the population to have an increased long-term growth rate. The more organisms can remember through their internal states, the higher growth rate a population can potentially achieve. Our results suggest experimental ways to infer internal memory and its benefit for adaptation in varying environments.

Nelson Martins-Ferreira

Regardless of its environment, the category of internal groupoids is shown to be equivalent to the full subcategory of involutive-2-links that are unital and associative. The new notion of involutive-2-link originates from the study of triangulated surfaces and their application in additive manufacturing and 3d-printing. Thus, this result establishes a bridge between the structure of an internal groupoid and an abstract triangulated surface. An example is provided which can be thought of as a crossed-module of magmas rather than groups.

Kavita Dharamraj

Aim: To determine the association between the frequency of physical activity and blood pressure (BP) levels among persons with type 2 diabetes at a public health center in Southwest Trinidad. Settings and Design: In 2011, the Penal Health Center, Diabetes Patient Self-Care Study enrolled 523 persons with type 2 diabetes in routine care in Southwest Trinidad aiming to obtain information on health status including diabetes and cardiovascular disease. The study was cross-sectional and included both males and females aged 25–87 years, having the exposure – physical activity and the outcome – BP levels. Subjects and Methods: Adults with type 2 diabetes aged 25–87 years with available information on physical activity and BP (n = 469). The main outcomes measures were systolic and diastolic BP (DBP) levels. Linear regression models examined the association between the frequency of physical activity (infrequent: <3x/week or frequent: ≥3x/week) and systolic BP (SBP)/DBP adjusting for potential confounders. Episodes of physical activity were defined as continuous physical activity, averaging ≥ 20 min/episode/week. Results: BP among hypertensive participants who exercise ≥ 3x/week was 5.3 mmHg lower than those who exercise <3x/week (Unadjusted β = −5.3, [95% confidence interval (CI) −10.0, −0.6], P = 0.026). DBP among hypertensive participants who exercise ≥3x/week was 0.4 mmHg lower than those who exercise <3x/week (Model 2: Adjusted β = −0.4, [95% CI – −3.5, 2.8], P = 0.818). Conclusion: Our findings may suggest an association between the frequency of physical inactivity and SBP levels in persons with type 2 diabetes.

Peizhi Zhang, Jiayi Li, Zicheng Wang et al.

PurposeThe mitogen-activated protein kinase (MAPK) signaling pathway is often studied in oncology as the most easily mentioned signaling pathway. This study aims to establish a new prognostic risk model of MAPK pathway related molecules in kidney renal clear cell carcinoma (KIRC) based on genome and transcriptome analysis.MethodsIn our study, RNA-seq data were acquired from the KIRC dataset of The Cancer Genome Atlas (TCGA) database. MAPK signaling pathway-related genes were obtained from the gene enrichment analysis (GSEA) database. We used “glmnet” and the “survival” extension package for LASSO (Least absolute shrinkage and selection operator) regression curve analysis and constructed a prognosis-related risk model. The survival curve and the COX regression analysis were used the “survival” expansion packages. The ROC curve was plotted using the “survival ROC” extension package. We then used the “rms” expansion package to construct a nomogram plot. We performed a pan-cancer analysis of CNV (copy number variation), SNV (single nucleotide variant), drug sensitivity, immune infiltration, and overall survival (OS) of 14 MAPK signaling pathway-related genes using several analysis websites, such as GEPIA website and TIMER database. Besides, the immunohistochemistry and pathway enrichment analysis used The Human Protein Atlas (THPA) database and the GSEA method. Finally, the mRNA expression of risk model genes in clinical renal cancer tissues versus adjacent normal tissues was further verified by real-time quantitative reverse transcription (qRT-PCR).ResultsWe performed Lasso regression analysis using 14 genes and created a new KIRC prognosis-related risk model. High-risk scores suggested that KIRC patients with lower-risk scores had a significantly worse prognosis. Based on the multivariate Cox analysis, we found that the risk score of this model could serve as an independent risk factor for KIRC patients. In addition, we used the THPA database to verify the differential expression of proteins between normal kidney tissues and KIRC tumor tissues. Finally, the results of qRT-PCR experiments suggested large differences in the mRNA expression of risk model genes.ConclusionsThis study constructs a KIRC prognosis prediction model involving 14 MAPK signaling pathway-related genes, which is essential for exploring potential biomarkers for KIRC diagnosis.

Takehiro Yamashita, MD, PhD, Naoki Iwakiri, MD, Eigo Kurebayashi, MD et al.

Antoine Allioux, Eric Finster, Matthieu Sozeau

By extending type theory with a universe of definitionally associative and unital polynomial monads, we show how to arrive at a definition of opetopic type which is able to encode a number of fully coherent algebraic structures. In particular, our approach leads to a definition of $\infty$-groupoid internal to type theory and we prove that the type of such $\infty$-groupoids is equivalent to the universe of types. That is, every type admits the structure of an $\infty$-groupoid internally, and this structure is unique.

Seyed Vahid Moravvej, Abdolreza Mirzaei, Mehran Safayani

Text summarization in medicine can help doctors for reducing the time to access important information from countless documents. The paper offers a supervised extractive summarization method based on conditional generative adversarial networks using convolutional neural networks. Unlike previous models, which often use greedy methods to select sentences, we use a new approach for selecting sentences. Moreover, we provide a network for biomedical word embedding, which improves summarization. An essential contribution of the paper is introducing a new loss function for the discriminator, making the discriminator perform better. The proposed model achieves results comparable to the state-of-the-art approaches, as determined by the ROUGE metric. Experiments on the medical dataset show that the proposed method works on average 5% better than the competing models and is more similar to the reference summaries.

Gurpreet K. Aulakh, Sushmita Maltare, Baljit Singh

Abstract Background CD34, a pan-selectin binding protein when glycosylated, has been shown to be involved in leukocyte migration to the site of inflammation. However, only one report is available on the expression and role of CD34 in neutrophil recruitment during acute lung inflammation. Methods We proceeded to study the role of CD34 in lung neutrophil migration using mouse model of endotoxin induced acute lung inflammation and studied over multiple time points, in generic CD34 knock-out (KO) strain. Results While there was no difference in BAL total or differential leukocyte counts, lung MPO content was lower in LPS exposed KO compared to WT group at 3 h time-point (p = 0.0308). The MPO levels in CD34 KO mice begin to rise at 9 h (p = 0.0021), as opposed to an early 3 h rise in WT mice (p = 0.0001), indicating that KO mice display delays in lung neutrophil recruitment kinetics. KO mice do not loose endotoxin induced lung vascular barrier properties as suggested by lower BAL total protein at 3 h (p = 0.0452) and 24 h (p = 0.0113) time-points. Several pro-inflammatory cytokines and chemokines (TNF-α, IL-1β, KC, MIP-1α, IL-6, IL-10 and IL-12 p70 sub-unit; p < 0.05) had higher levels in WT compared to KO group, at 3 h. Lung immunofluorescence in healthy WT mice reveals CD34 expression in the bronchiolar epithelium, in addition to alveolar septa. Conclusion Thus, given CD34′s pan-selectin affinity, and expression in the bronchiolar epithelium as well as alveolar septa, our study points towards a role of CD34 in lung neutrophil recruitment but not alveolar migration, cytokine expression and lung inflammation.