This study investigates the feature representations produced by publicly available open source medical vision-language models (VLMs). While medical VLMs are expected to capture diagnostically relevant features, their learned representations remain underexplored, and standard evaluations like classification accuracy do not fully reveal if they acquire truly discriminative, lesion-specific features. Understanding these representations is crucial for revealing medical image structures and improving downstream tasks in medical image analysis. This study aims to investigate the feature distributions learned by medical VLMs and evaluate the impact of medical specialization. We analyze the feature distribution of multiple image modalities extracted by some representative medical VLMs across lesion classification datasets on multiple modalities. These distributions were compared them with non-medical VLMs to assess the domain-specific medical training. Our experiments showed that medical VLMs can extract discriminative features that are effective for medical classification tasks. Moreover, it was found that non-medical VLMs with recent improvement with contextual enrichment such as LLM2CLIP produce more refined feature representations. Our results imply that enhancing text encoder is more crucial than training intensively on medical images when developing medical VLMs. Notably, non-medical models are particularly vulnerable to biases introduced by overlaied text strings on images. These findings underscore the need for careful consideration on model selection according to downstream tasks besides potential risks in inference due to background biases such as textual information in images.
Medical Multimodal Large Language Models (Medical MLLMs) have achieved remarkable progress in specialized medical tasks; however, research into their safety has lagged, posing potential risks for real-world deployment. In this paper, we first establish a multidimensional evaluation framework to systematically benchmark the safety of current SOTA Medical MLLMs. Our empirical analysis reveals pervasive vulnerabilities across both general and medical-specific safety dimensions in existing models, particularly highlighting their fragility against cross-modality jailbreak attacks. Furthermore, we find that the medical fine-tuning process frequently induces catastrophic forgetting of the model's original safety alignment. To address this challenge, we propose a novel "Parameter-Space Intervention" approach for efficient safety re-alignment. This method extracts intrinsic safety knowledge representations from original base models and concurrently injects them into the target model during the construction of medical capabilities. Additionally, we design a fine-grained parameter search algorithm to achieve an optimal trade-off between safety and medical performance. Experimental results demonstrate that our approach significantly bolsters the safety guardrails of Medical MLLMs without relying on additional domain-specific safety data, while minimizing degradation to core medical performance.
Hallucinations in foundation models arise from autoregressive training objectives that prioritize token-likelihood optimization over epistemic accuracy, fostering overconfidence and poorly calibrated uncertainty. We define medical hallucination as any model-generated output that is factually incorrect, logically inconsistent, or unsupported by authoritative clinical evidence in ways that could alter clinical decisions. We evaluated 11 foundation models (7 general-purpose, 4 medical-specialized) across seven medical hallucination tasks spanning medical reasoning and biomedical information retrieval. General-purpose models achieved significantly higher proportions of hallucination-free responses than medical-specialized models (median: 76.6% vs 51.3%, difference = 25.2%, 95% CI: 18.7-31.3%, Mann-Whitney U = 27.0, p = 0.012, rank-biserial r = -0.64). Top-performing models such as Gemini-2.5 Pro exceeded 97% accuracy when augmented with chain-of-thought prompting (base: 87.6%), while medical-specialized models like MedGemma ranged from 28.6-61.9% despite explicit training on medical corpora. Chain-of-thought reasoning significantly reduced hallucinations in 86.4% of tested comparisons after FDR correction (q < 0.05), demonstrating that explicit reasoning traces enable self-verification and error detection. Physician audits confirmed that 64-72% of residual hallucinations stemmed from causal or temporal reasoning failures rather than knowledge gaps. A global survey of clinicians (n = 70) validated real-world impact: 91.8% had encountered medical hallucinations, and 84.7% considered them capable of causing patient harm. The underperformance of medical-specialized models despite domain training indicates that safety emerges from sophisticated reasoning capabilities and broad knowledge integration developed during large-scale pre-training, not from narrow optimization.
Medical Phrase Grounding (MPG) maps radiological findings described in medical reports to specific regions in medical images. The primary obstacle hindering progress in MPG is the scarcity of annotated data available for training and validation. We propose anatomical grounding as an in-domain pre-training task that aligns anatomical terms with corresponding regions in medical images, leveraging large-scale datasets such as Chest ImaGenome. Our empirical evaluation on MS-CXR demonstrates that anatomical grounding pre-training significantly improves performance in both a zero-shot learning and fine-tuning setting, outperforming state-of-the-art MPG models. Our fine-tuned model achieved state-of-the-art performance on MS-CXR with an mIoU of 61.2, demonstrating the effectiveness of anatomical grounding pre-training for MPG.
<p>The current medical education curricula are designed to address the competencies that result in the graduation of a holistic healthcare professionals. The global competency frameworks have highlighted the outcomes of future doctors that not only incorporate necessary medical knowledge and procedural skills but also equip the graduates with human skills of professionalism, ethics, leadership, lifelong learners, critical thinkers and problem solvers. This has led to the incorporation of these domains in the training curricula of undergraduate medical education in Pakistani universities as well. The focus of this paper is to highlight the structure of a Professionalism, Ethics, Research and Leadership skills (PERLs) module that can be implemented in health professions education to be able to graduate a humanistic doctor who can provide holistic care to the community.</p>
Daniel P. Jeong, Saurabh Garg, Zachary C. Lipton
et al.
Several recent works seek to develop foundation models specifically for medical applications, adapting general-purpose large language models (LLMs) and vision-language models (VLMs) via continued pretraining on publicly available biomedical corpora. These works typically claim that such domain-adaptive pretraining (DAPT) improves performance on downstream medical tasks, such as answering medical licensing exam questions. In this paper, we compare seven public "medical" LLMs and two VLMs against their corresponding base models, arriving at a different conclusion: all medical VLMs and nearly all medical LLMs fail to consistently improve over their base models in the zero-/few-shot prompting regime for medical question-answering (QA) tasks. For instance, across the tasks and model pairs we consider in the 3-shot setting, medical LLMs only outperform their base models in 12.1% of cases, reach a (statistical) tie in 49.8% of cases, and are significantly worse than their base models in the remaining 38.2% of cases. Our conclusions are based on (i) comparing each medical model head-to-head, directly against the corresponding base model; (ii) optimizing the prompts for each model separately; and (iii) accounting for statistical uncertainty in comparisons. While these basic practices are not consistently adopted in the literature, our ablations show that they substantially impact conclusions. Our findings suggest that state-of-the-art general-domain models may already exhibit strong medical knowledge and reasoning capabilities, and offer recommendations to strengthen the conclusions of future studies.
Changsun Lee, Sangjoon Park, Cheong-Il Shin
et al.
Recent medical vision-language models (VLMs) have shown promise in 2D medical image interpretation. However extending them to 3D medical imaging has been challenging due to computational complexities and data scarcity. Although a few recent VLMs specified for 3D medical imaging have emerged, all are limited to learning volumetric representation of a 3D medical image as a set of sub-volumetric features. Such process introduces overly correlated representations along the z-axis that neglect slice-specific clinical details, particularly for 3D medical images where adjacent slices have low redundancy. To address this limitation, we introduce MS-VLM that mimic radiologists' workflow in 3D medical image interpretation. Specifically, radiologists analyze 3D medical images by examining individual slices sequentially and synthesizing information across slices and views. Likewise, MS-VLM leverages self-supervised 2D transformer encoders to learn a volumetric representation that capture inter-slice dependencies from a sequence of slice-specific features. Unbound by sub-volumetric patchification, MS-VLM is capable of obtaining useful volumetric representations from 3D medical images with any slice length and from multiple images acquired from different planes and phases. We evaluate MS-VLM on publicly available chest CT dataset CT-RATE and in-house rectal MRI dataset. In both scenarios, MS-VLM surpasses existing methods in radiology report generation, producing more coherent and clinically relevant reports. These findings highlight the potential of MS-VLM to advance 3D medical image interpretation and improve the robustness of medical VLMs.
In medical information extraction, medical Named Entity Recognition (NER) is indispensable, playing a crucial role in developing medical knowledge graphs, enhancing medical question-answering systems, and analyzing electronic medical records. The challenge in medical NER arises from the complex nested structures and sophisticated medical terminologies, distinguishing it from its counterparts in traditional domains. In response to these complexities, we propose a medical NER model based on Machine Reading Comprehension (MRC), which uses a task-adaptive pre-training strategy to improve the model's capability in the medical field. Meanwhile, our model introduces multiple word-pair embeddings and multi-granularity dilated convolution to enhance the model's representation ability and uses a combined predictor of Biaffine and MLP to improve the model's recognition performance. Experimental evaluations conducted on the CMeEE, a benchmark for Chinese nested medical NER, demonstrate that our proposed model outperforms the compared state-of-the-art (SOTA) models.
The availability of medical devices such as glucose levels and blood pressure measuring devices is continuously increasing. These devices have gained user interest as they are easy to use. However, medical devices introduce extra complexity to the network by being numerous, heterogeneous, and more vulnerable to cyber-attacks. For better network management and overall network security, it is important to understand the network traffic characteristics of the devices. Thus, in this paper, we analyze in detail the traffic characteristics of 8 medical devices both at the device level and at the level of individual functionality of each device. We collect and share both network and Bluetooth traffic from a total of 51 functionalities of the devices. Our analysis includes different metrics such as protocols, amount of incoming/outgoing traffic, DNS queries, and analysis of traffic destinations. We observed that devices have unique network and Bluetooth traffic characteristics, that might be useful in developing networking tools for medical devices.
One of the key goals of artificial intelligence (AI) is the development of a multimodal system that facilitates communication with the visual world (image and video) using a natural language query. Earlier works on medical question answering primarily focused on textual and visual (image) modalities, which may be inefficient in answering questions requiring demonstration. In recent years, significant progress has been achieved due to the introduction of large-scale language-vision datasets and the development of efficient deep neural techniques that bridge the gap between language and visual understanding. Improvements have been made in numerous vision-and-language tasks, such as visual captioning visual question answering, and natural language video localization. Most of the existing work on language vision focused on creating datasets and developing solutions for open-domain applications. We believe medical videos may provide the best possible answers to many first aid, medical emergency, and medical education questions. With increasing interest in AI to support clinical decision-making and improve patient engagement, there is a need to explore such challenges and develop efficient algorithms for medical language-video understanding and generation. Toward this, we introduced new tasks to foster research toward designing systems that can understand medical videos to provide visual answers to natural language questions, and are equipped with multimodal capability to generate instruction steps from the medical video. These tasks have the potential to support the development of sophisticated downstream applications that can benefit the public and medical professionals.
In the field of biomedical natural language processing, medical concept normalization is a crucial task for accurately mapping mentions of concepts to a large knowledge base. However, this task becomes even more challenging in low-resource settings, where limited data and resources are available. In this thesis, I explore the challenges of medical concept normalization in a low-resource setting. Specifically, I investigate the shortcomings of current medical concept normalization methods applied to German lay texts. Since there is no suitable dataset available, a dataset consisting of posts from a German medical online forum is annotated with concepts from the Unified Medical Language System. The experiments demonstrate that multilingual Transformer-based models are able to outperform string similarity methods. The use of contextual information to improve the normalization of lay mentions is also examined, but led to inferior results. Based on the results of the best performing model, I present a systematic error analysis and lay out potential improvements to mitigate frequent errors.
Accurate segmentation of lesion regions is crucial for clinical diagnosis and treatment across various diseases. While deep convolutional networks have achieved satisfactory results in medical image segmentation, they face challenges such as loss of lesion shape information due to continuous convolution and downsampling, as well as the high cost of manually labeling lesions with varying shapes and sizes. To address these issues, we propose a novel medical visual prompting (MVP) framework that leverages pre-training and prompting concepts from natural language processing (NLP). The framework utilizes three key components: Super-Pixel Guided Prompting (SPGP) for superpixelating the input image, Image Embedding Guided Prompting (IEGP) for freezing patch embedding and merging with superpixels to provide visual prompts, and Adaptive Attention Mechanism Guided Prompting (AAGP) for pinpointing prompt content and efficiently adapting all layers. By integrating SPGP, IEGP, and AAGP, the MVP enables the segmentation network to better learn shape prompting information and facilitates mutual learning across different tasks. Extensive experiments conducted on five datasets demonstrate superior performance of this method in various challenging medical image tasks, while simplifying single-task medical segmentation models. This novel framework offers improved performance with fewer parameters and holds significant potential for accurate segmentation of lesion regions in various medical tasks, making it clinically valuable.
Medical operations (MOs) are essential in healthcare,and they are also a big concept that includes various operations during the perioperative period.Traditional operation exposes its limitations during the perioperative period,reflected in medical training,surgical preparation,and postoperative rehabilitation.Serious Games for Medical Operation (SGMO) offer new ways and complementary solutions to support MOs.As a review,this paper analyzes the development of SGMO and considers various aspects of the SGMO,such as interface,functions,and technologies.By combining MO and serious games characteristics,the paper classifies SGMO and analyzes their features and functions for different groups of users and at various stages of the perioperative period (before,during,and after an MO).Interactive technologies used in SGMO are presented from a visual,haptic,and auditory perspective.This paper reviews the development of SGMO,summarizes its functions and technologies.Besides,it presents representative products and suggests future research directions.
Conversational AI systems can engage in unsafe behaviour when handling users' medical queries that can have severe consequences and could even lead to deaths. Systems therefore need to be capable of both recognising the seriousness of medical inputs and producing responses with appropriate levels of risk. We create a corpus of human written English language medical queries and the responses of different types of systems. We label these with both crowdsourced and expert annotations. While individual crowdworkers may be unreliable at grading the seriousness of the prompts, their aggregated labels tend to agree with professional opinion to a greater extent on identifying the medical queries and recognising the risk types posed by the responses. Results of classification experiments suggest that, while these tasks can be automated, caution should be exercised, as errors can potentially be very serious.
The role of software in society has changed drastically since the start of the 21st century. Software can now partially or fully facilitate anything from diagnosis to treatment of a disease, regardless of whether it is psychological or pathological, with the consequence of software being comparable to any other type of medical equipment, and this makes discovering when software must comply with such rules vital to both manufacturers and regulators. In lieu of the Medical Device Regulation we expand on the idea of intention, and identify the criteria software must fulfil to be considered medical devices within EU-law.
Transformers have dominated the field of natural language processing, and recently impacted the computer vision area. In the field of medical image analysis, Transformers have also been successfully applied to full-stack clinical applications, including image synthesis/reconstruction, registration, segmentation, detection, and diagnosis. Our paper aims to promote awareness and application of Transformers in the field of medical image analysis. Specifically, we first overview the core concepts of the attention mechanism built into Transformers and other basic components. Second, we review various Transformer architectures tailored for medical image applications and discuss their limitations. Within this review, we investigate key challenges revolving around the use of Transformers in different learning paradigms, improving the model efficiency, and their coupling with other techniques. We hope this review can give a comprehensive picture of Transformers to the readers in the field of medical image analysis.
Daniel Barrejón, Pablo M. Olmos, Antonio Artés-Rodríguez
Medical data sets are usually corrupted by noise and missing data. These missing patterns are commonly assumed to be completely random, but in medical scenarios, the reality is that these patterns occur in bursts due to sensors that are off for some time or data collected in a misaligned uneven fashion, among other causes. This paper proposes to model medical data records with heterogeneous data types and bursty missing data using sequential variational autoencoders (VAEs). In particular, we propose a new methodology, the Shi-VAE, which extends the capabilities of VAEs to sequential streams of data with missing observations. We compare our model against state-of-the-art solutions in an intensive care unit database (ICU) and a dataset of passive human monitoring. Furthermore, we find that standard error metrics such as RMSE are not conclusive enough to assess temporal models and include in our analysis the cross-correlation between the ground truth and the imputed signal. We show that Shi-VAE achieves the best performance in terms of using both metrics, with lower computational complexity than the GP-VAE model, which is the state-of-the-art method for medical records.