High-performing medical Large Language Models (LLMs) typically require extensive fine-tuning with substantial computational resources, limiting accessibility for resource-constrained healthcare institutions. This study introduces a confidence-driven multi-model framework that leverages model diversity to enhance medical question answering without fine-tuning. Our framework employs a two-stage architecture: a confidence detection module assesses the primary model's certainty, and an adaptive routing mechanism directs low-confidence queries to Helper models with complementary knowledge for collaborative reasoning. We evaluate our approach using Qwen3-30B-A3B-Instruct, Phi-4 14B, and Gemma 2 12B across three medical benchmarks; MedQA, MedMCQA, and PubMedQA. Result demonstrate that our framework achieves competitive performance, with particularly strong results in PubMedQA (95.0\%) and MedMCQA (78.0\%). Ablation studies confirm that confidence-aware routing combined with multi-model collaboration substantially outperforms single-model approaches and uniform reasoning strategies. This work establishes that strategic model collaboration offers a practical, computationally efficient pathway to improve medical AI systems, with significant implications for democratizing access to advanced medical AI in resource-limited settings.
Background: Deep learning has significantly advanced medical image analysis, with Vision Transformers (ViTs) offering a powerful alternative to convolutional models by modeling long-range dependencies through self-attention. However, ViTs are inherently data-intensive and lack domain-specific inductive biases, limiting their applicability in medical imaging. In contrast, radiomics provides interpretable, handcrafted descriptors of tissue heterogeneity but suffers from limited scalability and integration into end-to-end learning frameworks. In this work, we propose the Radiomics-Embedded Vision Transformer (RE-ViT) that combines radiomic features with data-driven visual embeddings within a ViT backbone. Purpose: To develop a hybrid RE-ViT framework that integrates radiomics and patch-wise ViT embeddings through early fusion, enhancing robustness and performance in medical image classification. Methods: Following the standard ViT pipeline, images were divided into patches. For each patch, handcrafted radiomic features were extracted and fused with linearly projected pixel embeddings. The fused representations were normalized, positionally encoded, and passed to the ViT encoder. A learnable [CLS] token aggregated patch-level information for classification. We evaluated RE-ViT on three public datasets (including BUSI, ChestXray2017, and Retinal OCT) using accuracy, macro AUC, sensitivity, and specificity. RE-ViT was benchmarked against CNN-based (VGG-16, ResNet) and hybrid (TransMed) models. Results: RE-ViT achieved state-of-the-art results: on BUSI, AUC=0.950+/-0.011; on ChestXray2017, AUC=0.989+/-0.004; on Retinal OCT, AUC=0.986+/-0.001, which outperforms other comparison models. Conclusions: The RE-ViT framework effectively integrates radiomics with ViT architectures, demonstrating improved performance and generalizability across multimodal medical image classification tasks.
This paper provides an overview of recent historical research regarding scientifically-informed challenges to the idea that the stars are other suns orbited by other inhabited earths -- an idea that came to be known as "the Plurality of Worlds". Johannes Kepler in the seventeenth century, Jacques Cassini in the eighteenth, and William Whewell in the nineteenth each argued against "pluralism" based on what in their respective times was solid science. Nevertheless, pluralism remained popular despite these and other scientific challenges. This history will be of interest to the astronomical community so that it is better positioned to avoid difficulties should the historical trajectory of pluralism continue, especially as it persists in the popular imagination.
Large language models (LLMs) have exhibited remarkable capabilities across various domains and tasks, pushing the boundaries of our knowledge in learning and cognition. The latest model, OpenAI's o1, stands out as the first LLM with an internalized chain-of-thought technique using reinforcement learning strategies. While it has demonstrated surprisingly strong capabilities on various general language tasks, its performance in specialized fields such as medicine remains unknown. To this end, this report provides a comprehensive exploration of o1 on different medical scenarios, examining 3 key aspects: understanding, reasoning, and multilinguality. Specifically, our evaluation encompasses 6 tasks using data from 37 medical datasets, including two newly constructed and more challenging question-answering (QA) tasks based on professional medical quizzes from the New England Journal of Medicine (NEJM) and The Lancet. These datasets offer greater clinical relevance compared to standard medical QA benchmarks such as MedQA, translating more effectively into real-world clinical utility. Our analysis of o1 suggests that the enhanced reasoning ability of LLMs may (significantly) benefit their capability to understand various medical instructions and reason through complex clinical scenarios. Notably, o1 surpasses the previous GPT-4 in accuracy by an average of 6.2% and 6.6% across 19 datasets and two newly created complex QA scenarios. But meanwhile, we identify several weaknesses in both the model capability and the existing evaluation protocols, including hallucination, inconsistent multilingual ability, and discrepant metrics for evaluation. We release our raw data and model outputs at https://ucsc-vlaa.github.io/o1_medicine/ for future research.
З електрофізіологією тісно пов'язаний розвиток фізіології як самостійної науки. Закономірності, сформульовані в цьому параграфі, важливі для розуміння закономірностей функціонування внутрішніх органів. Розділ електрофізіології включає вивчення процесів збудливості, формування потенціалів спокою і дії, механізми м'язового скорочення, проведення збудження по нервових волокнах, механізми проведення збудження через синапси. Робота всіх внутрішніх органів заснована на електрофізіологічних принципах, оскільки збудливі тканини входять до складу функціонуючих внутрішніх органів. З цим матеріалом буде корисно ознайомитись як студентам медичних вузів, так і працюючим лікарям, що й визначає актуальність роботи. Метою статті ми ставимо аналіз історичних етапів становлення електрофізіології як самостійний розділ фізіології. Методологія дослідження реалізується переважно за допомогою описового та історичного методів. Новизна дослідження. У статті йдеться, що наявність «тваринної електрики» вперше помітив Луїджі Гальвані. Подальші дослідження Вольта, Маттеуччі, Дюбуа-Реймона розширили уявлення про механізми проведення збуджень на нервово-м'язових елементах організму. Акцентовано увагу на ролі українських фізіологів, які зробили вагомий внесок у наукову спадщину. Так, В. Чаговцем були відкриті фундаментальні закономірності формування потенціалу та основи збудливості, концепція критичного рівня деполяризації. Висновки. Отримані моделі є фундаментальними навіть сьогодні. Наведені в статті дані розширюють знання про фізіологію збудливих тканин і дають можливість скласти більш повну картину з цього розрізу.
History of medicine. Medical expeditions, Social Sciences
Amirreza Hashemi, Yuemeng Feng, Arman Rahmim
et al.
This work investigates use of equivariant neural networks as efficient and high-performance frameworks for image reconstruction and denoising in nuclear medicine. Our work aims to tackle limitations of conventional Convolutional Neural Networks (CNNs), which require significant training. We investigated equivariant networks, aiming to reduce CNN's dependency on specific training sets. Specifically, we implemented and evaluated equivariant spherical CNNs (SCNNs) for 2- and 3-dimensional medical imaging problems. Our results demonstrate superior quality and computational efficiency of SCNNs in both image reconstruction and denoising benchmark problems. Furthermore, we propose a novel approach to employ SCNNs as a complement to conventional image reconstruction tools, enhancing the outcomes while reducing reliance on the training set. Across all cases, we observed significant decrease in computational cost by leveraging the inherent inclusion of equivariant representatives while achieving the same or higher quality of image processing using SCNNs compared to CNNs. Additionally, we explore the potential of SCNNs for broader tomography applications, particularly those requiring rotationally variant representation.
The success of deep networks in medical image segmentation relies heavily on massive labeled training data. However, acquiring dense annotations is a time-consuming process. Weakly-supervised methods normally employ less expensive forms of supervision, among which scribbles started to gain popularity lately thanks to its flexibility. However, due to lack of shape and boundary information, it is extremely challenging to train a deep network on scribbles that generalizes on unlabeled pixels. In this paper, we present a straightforward yet effective scribble supervised learning framework. Inspired by recent advances of transformer based segmentation, we create a pluggable spatial self-attention module which could be attached on top of any internal feature layers of arbitrary fully convolutional network (FCN) backbone. The module infuses global interaction while keeping the efficiency of convolutions. Descended from this module, we construct a similarity metric based on normalized and symmetrized attention. This attentive similarity leads to a novel regularization loss that imposes consistency between segmentation prediction and visual affinity. This attentive similarity loss optimizes the alignment of FCN encoders, attention mapping and model prediction. Ultimately, the proposed FCN+Attention architecture can be trained end-to-end guided by a combination of three learning objectives: partial segmentation loss, a customized masked conditional random fields and the proposed attentive similarity loss. Extensive experiments on public datasets (ACDC and CHAOS) showed that our framework not just out-performs existing state-of-the-art, but also delivers close performance to fully-supervised benchmark. Code will be available upon publication.
Despite its clinical utility, medical image segmentation (MIS) remains a daunting task due to images' inherent complexity and variability. Vision transformers (ViTs) have recently emerged as a promising solution to improve MIS; however, they require larger training datasets than convolutional neural networks. To overcome this obstacle, data-efficient ViTs were proposed, but they are typically trained using a single source of data, which overlooks the valuable knowledge that could be leveraged from other available datasets. Naivly combining datasets from different domains can result in negative knowledge transfer (NKT), i.e., a decrease in model performance on some domains with non-negligible inter-domain heterogeneity. In this paper, we propose MDViT, the first multi-domain ViT that includes domain adapters to mitigate data-hunger and combat NKT by adaptively exploiting knowledge in multiple small data resources (domains). Further, to enhance representation learning across domains, we integrate a mutual knowledge distillation paradigm that transfers knowledge between a universal network (spanning all the domains) and auxiliary domain-specific branches. Experiments on 4 skin lesion segmentation datasets show that MDViT outperforms state-of-the-art algorithms, with superior segmentation performance and a fixed model size, at inference time, even as more domains are added. Our code is available at https://github.com/siyi-wind/MDViT.
Deep learning models usually suffer from domain shift issues, where models trained on one source domain do not generalize well to other unseen domains. In this work, we investigate the single-source domain generalization problem: training a deep network that is robust to unseen domains, under the condition that training data is only available from one source domain, which is common in medical imaging applications. We tackle this problem in the context of cross-domain medical image segmentation. Under this scenario, domain shifts are mainly caused by different acquisition processes. We propose a simple causality-inspired data augmentation approach to expose a segmentation model to synthesized domain-shifted training examples. Specifically, 1) to make the deep model robust to discrepancies in image intensities and textures, we employ a family of randomly-weighted shallow networks. They augment training images using diverse appearance transformations. 2) Further we show that spurious correlations among objects in an image are detrimental to domain robustness. These correlations might be taken by the network as domain-specific clues for making predictions, and they may break on unseen domains. We remove these spurious correlations via causal intervention. This is achieved by resampling the appearances of potentially correlated objects independently. The proposed approach is validated on three cross-domain segmentation tasks: cross-modality (CT-MRI) abdominal image segmentation, cross-sequence (bSSFP-LGE) cardiac MRI segmentation, and cross-center prostate MRI segmentation. The proposed approach yields consistent performance gains compared with competitive methods when tested on unseen domains.
This paper discusses the relation between the decoherent histories approach to quantum mechanics that is based on coarse-grained decoherent histories of a closed system, and the approximate quantum mechanics of measured subsystems, as in the Copenhagen interpretation. We show how the a classical world used in such formulations is not to something to be postulated but rather explained by suitable sets of alternative histories of quasiclassical variables. We discuss the general definition of measurement, the collapse of the wave function, and irreversibility from the perspective of decoherent histories quantum theory..
Deep learning models perform best when tested on target (test) data domains whose distribution is similar to the set of source (train) domains. However, model generalization can be hindered when there is significant difference in the underlying statistics between the target and source domains. In this work, we adapt a domain generalization method based on a model-agnostic meta-learning framework to biomedical imaging. The method learns a domain-agnostic feature representation to improve generalization of models to the unseen test distribution. The method can be used for any imaging task, as it does not depend on the underlying model architecture. We validate the approach through a computed tomography (CT) vertebrae segmentation task across healthy and pathological cases on three datasets. Next, we employ few-shot learning, i.e. training the generalized model using very few examples from the unseen domain, to quickly adapt the model to new unseen data distribution. Our results suggest that the method could help generalize models across different medical centers, image acquisition protocols, anatomies, different regions in a given scan, healthy and diseased populations across varied imaging modalities.
This article contributes to a global history of relativity, by exploring how Einstein's theory was appropriated in Belgium. This may sound as a contradiction in terms, yet the early-twentieth-century Belgian context, because of its cultural diversity and reflectiveness of global conditions (the principal example being the First World War), proves well-suited to expose transnational flows and patterns in the global history of relativity. The attempts of Belgian physicist Théophile de Donder to contribute to relativity physics during the 1910s and 1920s illustrate the role of the war in shaping the transnational networks through which relativity circulated. The local attitudes of conservative Belgian Catholic scientists and philosophers, who denied that relativity was philosophically significant, exemplify a global pattern: while critics of relativity feared to become marginalized by the scientific, political, and cultural revolutions that Einstein and his theory were taken to represent, supporters sympathized with these revolutions.
We explore the application of concepts developed in High Energy Physics (HEP) for advanced medical data analysis. Our study case is a problem with high social impact: clinically-feasible Magnetic Resonance Fingerprinting (MRF). MRF is a new, quantitative, imaging technique that replaces multiple qualitative Magnetic Resonance Imaging (MRI) exams with a single, reproducible measurement for increased sensitivity and efficiency. A fast acquisition is followed by a pattern matching (PM) task, where signal responses are matched to entries from a dictionary of simulated, physically-feasible responses, yielding multiple tissue parameters simultaneously. Each pixel signal response in the volume is compared through scalar products with all dictionary entries to choose the best measurement reproduction. MRF is limited by the PM processing time, which scales exponentially with the dictionary dimensionality, i.e. with the number of tissue parameters to be reconstructed. We developed for HEP a powerful, compact, embedded system, optimized for extremely fast PM. This system executes real-time tracking for online event selection in the HEP experiments, exploiting maximum parallelism and pipelining. Track reconstruction is executed in two steps. The Associative Memory (AM) ASIC first implements a PM algorithm by recognizing track candidates at low resolution. The second step, which is implemented into FPGAs (Field Programmable Gate Arrays), refines the AM output finding the track parameters at full resolution. We propose to use this system to perform MRF, to achieve clinically reasonable reconstruction time. This paper proposes an adaptation of the HEP system for medical imaging and shows some preliminary results.
This study focused on the socialist camp’s North Korean medical support and its effects on North Korean medical field from liberation to 1958. Except for the Soviet assistance from liberation to the Korean War, existing studies mainly have paid attention to the ‘autonomous’ growth of the North Korean medical field. The studies on the medical support of the Eastern European countries during the Korean War have only focused on one-sided support and neglected the interactions with the North Korean medical field. Failing in utilizing the materials produced in North Korea has led to the omission of detailed circumstances of providing support. Since the review of China’s support and the North Korea-China medical exchanges has been concentrated in the period after the mid-1950s, the impacts of China’s medical support on North Korea during the Korean War period and the post-war recovery period have not been taken into account.<br> In terms of these limitations, this study examined the medical activities by the Socialist camp of the Eastern European countries in North Korea after the Korean War. The medical aid teams from Hungary, Romania, Bulgaria, Czechoslovakia, Poland, and East Germany that came to North Korea in the wake of the Korean War continued to stay in North Korea after the war to build hospitals and train medical personnel. In the hospitals operated by these countries, cooperative medical care with North Korean medical personnel and medical technology education were conducted. Moreover, medical teams from each country in North Korea held seminars and conferences and exchanged knowledge with the North Korean medical field staffs. These activities by the Socialist countries in North Korea provided the North Korean medical personnel with the opportunity to directly experience the medical technology of each country.<br> China’s support was crucial to North Korea’s ‘rediscovery’ of Korean medicine in the mid-1950s. After the Korean War, North Korea began to apply the Chinese-Western medicine integration policy, which was performed in China at that time, to the North Korean health care field through China’s medical support and exchanges. In other words, China’s emphasis on Chinese medicine and the integration of the Chinese-Western medicine were presented as one of the directions for medical development of North Korea in the 1950s, and the experiences of China in this process convinced North Korea that Korean medicine policy was appropriate. The decision-makers of the North Korean medical policies, who returned to North Korea after studying abroad in China at that time, actively introduced the experiences from China and constantly sought to learn about them.<br> This study identified that a variety of external stimuli had complex impacts on the North Korean medical field in the gap between ‘Soviet learning’ in the late 1940s and the ‘autonomous’ medical development since the 1960s. The North Korean medical field was formed not by the unilateral or dominant influences of a single nation but by the stimulation from many nations and the various interactions in the process.
One usually refers the concept of Feynman path integral to the work of Norbert Wiener on Brownian motion in the early 1920s. This view is not false and we show in this article that Wiener used the first path integral of the history of physics to describe the Brownian motion. That said, Wiener, as he pointed out, was inspired by the work of some French mathematicians, particularly Gateaux and Levy. Moreover, although Richard Feynman has independently found this notion, we show that in the course of the 1930s, while searching a kind of geometrization of quantum mechanics, another French mathematician, Adolphe Buhl, noticed by the philosopher Gaston Bachelard, had himself been close to forge such a notion. This reminder does not detract from the remarkable discovery of Feynman, which must undeniably be attributed to him. We also show, however, that the difficulties of this notion had to wait many years before being resolved, and it was only recently that the path integral could be rigorously established from a mathematical point of view.
We explore the chemodynamical properties of a sample of barred galaxies in the Auriga magneto-hydrodynamical cosmological zoom-in simulations, which form boxy/peanut (b/p) bulges, and compare these to the Milky Way (MW). We show that the Auriga galaxies which best reproduce the chemodynamical properties of stellar populations in the MW bulge have quiescent merger histories since redshift $z\sim3.5$: their last major merger occurs at $t_{\rm lookback}>12\,\rm Gyrs$, while subsequent mergers have a stellar mass ratio of $\leq$1:20, suggesting an upper limit of a few percent for the mass ratio of the recently proposed Gaia Sausage/Enceladus merger. These Auriga MW-analogues have a negligible fraction of ex-situ stars in the b/p region ($<1\%$), with flattened, thick disc-like metal-poor stellar populations. The average fraction of ex-situ stars in the central regions of all Auriga galaxies with b/p's is 3% -- significantly lower than in those which do not host a b/p or a bar. While the central regions of these barred galaxies contain the oldest populations, they also have stars younger than 5Gyrs (>30%) and exhibit X-shaped age and abundance distributions. Examining the discs in our sample, we find that in some cases a star-forming ring forms around the bar, which alters the metallicity of the inner regions of the galaxy. Further out in the disc, bar-induced resonances lead to metal-rich ridges in the $V_φ-r$ plane -- the longest of which is due to the Outer Lindblad Resonance. Our results suggest the Milky Way has an uncommonly quiet merger history, which leads to an essentially in-situ bulge, and highlight the significant effects the bar can have on the surrounding disc.
Анистратенко Антонина. Функциональное значение хронотопа в уровневой структуре прозаического произведения (на материале современной романистики). Пятиуровневая структура художественного произведения объединяет референтный, образный, языковой, текстуальное и кодовый уровне. Референтный - воплощает общую картину
мира и концепцию человека, то есть опирается на природные знаковые системы. В теле произведения этот уровень
соответствует организации художественного пространства и времени и структуры повествования. Они возникают как
базовый, первичный уровень текстуальной реализации целостного произведения литературы. Так создается парадокс
литературного чтения: разграничение времени и пространства в литературном тексте создает новую синкретическую
сущность хронотопа в художественном произведении. Ведь литературный текст может быть охарактеризован как один
из уровней художественного произведения и обнаружен как концепт, хронотоп, архетип времени или как линейное
время. Следовательно, целью статьи является определение особенностей и своеобразия каждого из временных проявлений, как базовая характеристика литературного текста в теории структурализма.
Ключевые слова: литературное произведение, текст, уровни художественного произведения, концепт времени,
архетип времени, хронотоп.
Anistratenko Antonina. The functional significance of
chronotop in the leveled structure of prose (based on current
novels expirience). The five-level structure of a prose itemt combines reference, figurative, linguistic, textual and code levels. Reference one parcipate the general worldview and the concept of
man, that is based on the natural sign systems. In the body of the
text this level corresponds to the organization of the artistic space
and the time and structure of the narrator. It appears as a basic,
primary level of textual realization in the holistic work of literature.
In such way the paradox of literary reading uses to be: the delineation of time and space in the literary text creates a new syncretic essence of the chronotop in the artistic text. After all, literary text could be described as one of the levels of artistic text and is identified as a concept, chronotope, archetype of time or as linear time.
Therefore, the purpose of the article is to outline the peculiarities
and peculiarities of each of the temporal realisations as the basic
characteristics of the literary text in the theory of structuralism.
The concept of time correlates with the anthropological un-
derstanding of time. After all, an artwork serves always as a model
of a partial representation of the world, which runs to fullness forever. So the system is detected using the components of the artisticproduct, one of them is the concept of time. The chronotop of the
literary work is much more subjective to schematization. Unlike
the concept and archetype of time, it is part of the architectonics of
the text. In fixed plot schemes it identifites by the type of
chronotope and use to be determined by the following types of
novel: the historical novel (the linear chronotope, mostly unfolding
in the past), a fantastic novel (the cyclonic chronotope, mostly
unfolding in the present and future), a detective novel (the inverse
chronotope unfolding from the present to the past). The archetype
of time like a concept that works beyond the boundaries of the
national worldview and considered at the level of civilization that
is on the background of the most general images and realizes an
archetype of time. It could be compared with such archetypes as
Earth, Water, Fire, Air, Good, Evil, Mother, Father, Force, Enemy,
Home, archetype of Time is a little bit later. If to use it likely corre-
lated with a certain functional sense of time, then temporal arche-
type corresponds to the philosophical perception of time.
Key words: literary work, text, levels of artistic text, concept
of time, archetype of time, chronotope.
History of medicine. Medical expeditions, Social Sciences