Volume 45, Issue 6

Vision-language models (VLMs) have achieved impressive progress in natural image reasoning, yet their potential in medical imaging remains underexplored. Medical vision-language tasks demand precise understanding and clinically coherent answers, which are difficult to achieve due to complexity of medical data and the scarcity of high-quality expert annotations. These challenges limit the effectiveness of conventional supervised fine-tuning (SFT) and Chain-of-Thought (CoT) strategies that work well in general domains. To address these challenges, we propose Med-R1, a reinforcement learning (RL)-enhanced VLM designed to improve generalization and reliability in medical reasoning. Med-R1 adopts Group Relative Policy Optimization (GRPO) to encourage reward-guided learning beyond static annotations. We comprehensively evaluate Med-R1 across eight distinct medical imaging modalities. Med-R1 achieves a 29.94% improvement in average accuracy over its base model Qwen2-VL-2B, and even outperforms Qwen2-VL-72B—a model with $36\times $ more parameters . To assess cross-task generalization, we further evaluate Med-R1 on five question types. Med-R1 outperforms Qwen2-VL-2B by 32.06% in question-type generalization, also surpassing Qwen2-VL-72B. We further explore the thinking process in Med-R1, a crucial component of Deepseek-R1. Our results show that omitting intermediate rationales ( No-Thinking Med-R1 ) not only improves cross-domain generalization with less training, but also challenges the common assumption that more reasoning always helps. Nevertheless, we also find that the Think-After Med-R1 variant further improves performance while maintaining interpretability. These findings suggest that, in medical VQA, the mere presence of explicit reasoning does not guarantee better performance. Instead, performance depends on the quality of the reasoning and the position where the reasoning is generated.

视觉语言模型（VLM）在自然图像推理方面取得了显著进展，但其在医学影像中的潜力尚未得到充分探索。医学视觉语言任务要求精确的理解和临床一致的答案，但由于医学数据的复杂性和高质量专家标注的稀缺性，这些难以实现。这些挑战限制了在通用领域表现良好的传统监督微调（SFT）和思维链（CoT）策略的有效性。为了解决这些挑战，我们提出了Med-R1，一种强化学习（RL）增强的VLM，旨在提高医学推理的泛化性和可靠性。Med-R1采用组相对策略优化（GRPO）来鼓励超越静态标注的奖励引导学习。我们在八种不同的医学成像模态上全面评估了Med-R1。Med-R1在其基础模型Qwen2-VL-2B上实现了平均准确率29.94%的提升，甚至超越了参数多36倍的Qwen2-VL-72B模型。为了评估跨任务泛化能力，我们进一步在五种问题类型上评估了Med-R1。Med-R1在问题类型泛化上比Qwen2-VL-2B提高了32.06%，也超越了Qwen2-VL-72B。我们进一步探索了Med-R1中的思考过程，这是Deepseek-R1的关键组成部分。我们的结果表明，省略中间推理（无思考Med-R1）不仅以更少的训练提高了跨领域泛化能力，而且挑战了“更多推理总是更好”的常见假设。尽管如此，我们也发现“后思考”Med-R1变体在保持可解释性的同时进一步提高了性能。这些发现表明，在医学VQA中，仅仅存在显式推理并不能保证更好的性能。相反，性能取决于推理的质量以及推理生成的位置。

Tumor synthesis can generate challenging cases that AI often misses or over-detects. Training on these cases improves AI performance. However, most existing synthesis methods are either unconditional— generating images from random variables—or conditioned only on tumor shape. As a result, they lack control over clinically important tumor characteristics, such as texture, heterogeneity, boundary, a...

肿瘤合成可以生成AI经常漏检或过度检测的具有挑战性的病例。在这些病例上训练可以提高AI的性能。然而，大多数现有的合成方法要么是无条件的——从随机变量生成图像，要么仅以肿瘤形状为条件。因此，它们缺乏对临床上重要的肿瘤特征的控制，如纹理、异质性、边界等。

Accurate segmentation of cervical structures in transvaginal ultrasound (TVS) is critical for assessing the risk of spontaneous preterm birth (PTB), yet the scarcity of labeled data limits the performance of supervised learning approaches. This paper introduces the Fetal Ultrasound Grand Challenge (FUGC), the first benchmark for semi-supervised learning in cervical segmentation, hosted at ISBI 202...

在经阴道超声（TVS）中准确分割宫颈结构对于评估自发性早产（PTB）风险至关重要，然而标记数据的稀缺限制了监督学习方法的性能。本文介绍了胎儿超声大挑战（FUGC），这是首个用于宫颈分割的半监督学习基准，在ISBI 202...上举办。

Diffusion models have shown great promise in medical image denoising and reconstruction, but their application to Positron Emission Tomography (PET) imaging remains limited by tracer-specific contrast variability and high computational demands. In this work, we proposed an anatomical prior–guided PET image reconstruction method based on diffusion models, inspired by the deep diffusion image prior ...

扩散模型在医学图像去噪和重建中显示出巨大潜力，但它们应用于正电子发射断层扫描（PET）成像仍然受到示踪剂特异性对比度变化和高计算需求的限制。在本工作中，我们提出了一种基于扩散模型的解剖先验引导的PET图像重建方法，受深度扩散图像先验的启发。

One of the main challenges in medical microwave imaging is to provide meaningful images when only minimal a priori information is available. Concerning the case of brain stroke imaging, this paper proposes a novel method for generating a low-resolution patient-specific approximation of the head by processing the same data used for the diagnosis. The method assumes knowledge of the head shape and a...

医学微波成像的主要挑战之一是在仅有最小先验信息时提供有意义的图像。针对脑卒中成像的情况，本文提出了一种新方法，通过处理用于诊断的相同数据来生成低分辨率的患者特异性头部近似。该方法假设已知头部形状和...

Functional ultrasound localization micro- scopy (fULM) enables brain-wide mapping of neural activity at micron-scale resolution but suffers from limited sensitivity due to sparse and noisy microbubble (MB) detections. Extending fULM into three dimensions (3D) further exacerbates these challenges because of low-frequency matrix arrays, reduced localization efficiency, and severe data sparsity. To a...

功能超声定位显微镜（fULM）能够以微米级分辨率实现全脑神经活动成像，但由于稀疏且嘈杂的微泡检测，其灵敏度有限。将fULM扩展到三维进一步加剧了这些挑战，因为低频矩阵阵列、降低的定位效率和严重的数据稀疏性。为了...

Scribble supervision reduces annotation costs in multi-organ segmentation. However, its sparsity results in insufficient supervision for most regions and inadequate feature learning in hard areas (e.g., organ boundaries). These hard areas cause model confirmation bias and high epistemic uncertainty, which existing methods fail to address. To overcome these core challenges, we propose an epistemic-...

涂鸦监督减少了多器官分割中的标注成本。然而，其稀疏性导致大多数区域监督不足，且在困难区域（例如器官边界）特征学习不充分。这些困难区域导致模型确认偏差和高认知不确定性，现有方法未能解决。为了克服这些核心挑战，我们提出了一种认知驱动的...

Generative diffusion models have received increasing attention in medical imaging, particularly in limited-angle computed tomography (LACT). Standard diffusion models achieve high-quality image reconstruction but require a large number of sampling steps during inference, imposing a heavy computational burden. Although skip-sampling strategies have been proposed to improve efficiency, they often le...

中文摘要翻译待生成

Plane-wave (PW) imaging, which provides high temporal resolution, has gained a significant attention. However, it is constrained by inherent lack of focus, requiring support from beamforming methods, such as Coherent Plane-Wave Compounding (CPWC), which can diminish the temporal advantages. Therefore, it is crucial to develop an advanced approach capable of addressing the trade-off between quality...

中文摘要翻译待生成

Performing unsupervised anomaly detection in retinal optical coherence tomography (OCT) images involves training a model solely on anomaly-free samples and detecting anomalies during inference, which reduces the cost of collecting large-scale annotated anomalous data. However, retinal OCT images exhibit significant variations in shape, thickness, and orientation, and lesions often have similar ref...

中文摘要翻译待生成

Given the rich and complementary information contained in multi-phase CT images (CTs), they play an indispensable role in liver cancer diagnosis and prognosis, wherein an important prerequisite is liver tumor segmentation. However, spatial misalignments across phases and the limited availability of high-quality multi-phase CT datasets significantly hinder the performance of liver tumor segmentatio...

中文摘要翻译待生成

Central lumbar spinal stenosis, a prevalent degenerative spinal disorder, severely impacts the quality of life for those affected. Axial and sagittal MRI images offer diverse information on tissue structure and lesions, which is crucial for accurate diagnosis. However, MRI-based diagnostic approaches still have poor lesion localization, insufficient cross-view alignment, underutilization of multi-...

中文摘要翻译待生成

Integrating multimodal radiological images and clinical data is critical for survival prediction in rectal cancer. However, existing methods often lack sufficient consideration of 1) modality heterogeneity (caused by rectal peristalsis, noise artifacts, and missing modalities) and 2) site heterogeneity (caused by different imaging protocols and patient populations). These factors hinder the model ...

中文摘要翻译待生成

Direct parametric reconstruction algorithms have been developed to improve the statistical reliability of parametric images estimated from dynamic PET imaging data. However, these estimates are degraded by noise due to measurement error and noise propagation during reconstruction. In this study, we develop a deep image prior (DIP) regularized direct reconstruction method, where the DIP network is ...

中文摘要翻译待生成

Surgical action triplet recognition plays a vital role in computer-assisted and anatomical surgery. A fine-grained formulation of this task aims to simultaneously identify the surgical instrument, operative verb, and anatomical target, which are represented as triplets in the form of . Existing methods face two key challenges: 1) Relying solely on the visual features of...

中文摘要翻译待生成

Probing diffusion in myelin water using diffusion-weighted ${T} {_{{1}}}$ -/ ${T} {_{{2}}}$ -selective MRI acquisitions enables noninvasive measurement of myelinated axon diameter. Its application for in vivo measurements requires numerical verification through diffusion simulations. Here, we propose the theory of myelin water diffusion as measured with a diffusion MRI pulse sequence with wide gr...

中文摘要翻译待生成

Predicting stroke outcome remains challenging due to inherent heterogeneity, misalignment of multimodal clinical data, and the availability of well-annotated longitudinal datasets. Current methodologies often lack robustness and generalizability across these tasks. We propose a few-shot contrastive learning framework that integrates brain MRI images and structured clinical records for cross-task p...

中文摘要翻译待生成

The integration of laboratory tests and medical images is crucial in making accurate disease prediction. However, imaging data exhibits temporal sparsity, compared to frequently collected laboratory tests. This temporal sparsity limits effective multi-modal interaction, which in turn degrades the prediction accuracy. We address this issue by generating additional medical images at more time points...

中文摘要翻译待生成

Automated chest X-ray report generation requires not only clinical accuracy but also transparent and interpretable diagnostic reasoning. In this work, we propose FiR-Rad, a two-stage framework that combines explicit structured reasoning with targeted fine-grained optimization. In the first stage, a supervised chain-of-thought approach guides the model to sequentially analyze and describe a compreh...

中文摘要翻译待生成

Vision foundation models have demonstrated strong generalization in medical image segmentation by leveraging large-scale, heterogeneous pretraining. However, they often struggle to generalize to specialized clinical tasks under limited annotations or rare pathological variations, due to a mismatch between general priors and task-specific requirements. To address this, we propose Uncertainty-inform...

中文摘要翻译待生成

Pseudo-healthy image synthesis aims to generate subject-specific, pathology-free images from pathological scans. Such images can be helpful in certain tasks, such as anomaly detection and understanding changes induced by pathology and disease. A participant cannot be “healthy” and “unhealthy” at the same time, and thus, directly obtaining pathological and healthy paired images of the same individu...

中文摘要翻译待生成

Histology image segmentation is a critical prerequisite to pathological diagnosis. Accurate segmentation of these images can significantly aid physicians by facilitating quicker and more precise diagnostic decisions. A notable challenge in this area arises from the fact that different objects within histology images require segmentation at varying magnifications. However, most existing models are ...

中文摘要翻译待生成

Self-supervised endoscopic depth estimation seeks to reconstruct dense depth information from clinical endoscopic video sequences without the necessity of ground-truth depth annotations, thereby offering significant potential for widespread application in surgical environments. Nevertheless, the majority of current approaches treat each video frame as an independent static image, neglecting the in...

中文摘要翻译待生成

Segmentation of the pulmonary vessel from computed tomography (CT) images plays a crucial role in the diagnosis and treatment of various lung diseases. Although deep learning-based approaches have shown remarkable progress in recent years, their performance is often hindered by the lack of high-quality annotated datasets, in which the complex anatomy and morphology of pulmonary vessels make manual...

中文摘要翻译待生成

Biomechanical modelling of soft tissue provides a method for constraining medical image registration, such that the estimated spatial transformation is considered biophysically plausible. Existing methods either directly optimize the loss function containing the biomechanical-constrained regularization term over deformations, which takes much computational time, or are trained using biomechanicall...

中文摘要翻译待生成