Hasil untuk "Medical legislation"

Ziad Elshaer, Essam A. Rashed

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.

Zhenyu Yang, Haiming Zhu, Rihui Zhang et al.

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.

Brian W. Pogue, Alexander P. Niver

Medical Physics education is delivered through accredited programs with admissions and funding for students determined by individual institutions providing the educational experiences. Public data from accredited graduate programs, along with funding data, were used to analyze institutional trends in this educational market. Temporal trends from 2017 to 2023 show robust growth in MS graduates, increasing at an average of 17.7 per year, as compared to steady but modest growth in PhDs, increasing by 3.6 per year. The current status is there are nearly two MS graduates for every PhD graduate. Trends in funding show self-funding of students is a dominant pathway in domestic programs. Those programs dominated by accredited MS education have their largest fraction of faculty in radiation oncology departments, whereas those dominated by PhD education have their largest fraction of faculty in radiology departments. Overall NIH funding in the space of radiation diagnostics and therapeutics has been largely static over this timeframe, but with a notable 5 year rise in NCI funding. This can be contrasted to a substantial 5X-6X rise in NIH funding for engineering research in this same period, with significant increases in trainee funding there. Taken as a whole, this survey shows that growth in the field of medical physics education is dominated by MS graduates, presumably servicing the expanded growth needs for well-trained clinical physicists. However, the research infrastructure that supports PhD training in medical physics seems likely to be growing modestly and missing the growth trend of NIH funding that appears to show substantially more growth in non-accredited programs such as biomedical engineering. This data is useful to informing accreditation guidance on numbers of graduates to match the workforce needs or for inter-institutional planning around education goals.

Lee J. Evitts, Philip W. Miller, Chiara Da Pieve et al.

Background: The emergence and growth of fusion technology enables investigative studies into its applications beyond typical power production facilities. This study seeks to determine the viability of medical isotope production with the neutrons produced in an example large fusion device. Using FISPACT-II (a nuclear inventory code) and a simulated fusion spectrum, the production yields of a significant number of potentially clinically relevant (both in use and novel) medical isotopes were calculated. Comparative calculations were also conducted against existing production routes. Results: Depending on the neutron flux of the fusion device, it could be an ideal technology to produce alpha-emitters such as 212Bi/212Pb, it may be able to contribute to the production of 99mTc/99Mo, and could offer an alternative route in the production a few Auger-emitting candidates. There is also a long list of beta-emitting nuclides where fusion technology may be best placed to produce over existing technologies including 67Cu, 90Y and 47Sc. Conclusions: It is theoretically viable to produce existing and novel medical isotopes with fusion technology. However, a significant number of assumptions form the basis of this study which would need to be studied further for any particular nuclide of interest.

Heather M. Whitney, Hui Li, Karen Drukker et al.

Health disparities (differences in non-genetic conditions that influence health) can be associated with differences in burden of disease by groups within a population. Social determinants of health (SDOH) are domains such as health care access, dietary access, and economics frequently studied for potential association with health disparities. Evaluating SDOH-related phenotypes using routine medical images as data sources may enhance health disparities research. We developed a pipeline for using quantitative measures automatically extracted from medical images as inputs into health disparities index calculations. Our study focused on the use case of two SDOH demographic correlates (sex and race) and data extracted from chest radiographs of 1,571 unique patients. The likelihood of severe disease within the lung parenchyma from each image type, measured using an established deep learning model, was merged into a single numerical image-based phenotype for each patient. Patients were then separated into phenogroups by unsupervised clustering of the image-based phenotypes. The health rate for each phenogroup was defined as the median image-based phenotype for each SDOH used as inputs to four imaging-derived health disparities indices (iHDIs): one absolute measure (between-group variance) and three relative measures (index of disparity, Theil index, and mean log deviation). The iHDI measures demonstrated feasible values for each SDOH demographic correlate, showing potential for medical images to serve as a novel probe for health disparities. Large-scale AI analysis of medical images can serve as a probe for a novel data source for health disparities research.

Harry G. Saavedra, Ramiro Moro

More than a century ago, Karl Bernhard Zoeppritz derived the equations that determine the reflected and transmitted coefficients at a planar interface for an incident seismic wave. The coefficients so obtained are a function of the elastic parameters of the media on each side of the interface and the angle of incidence. Approximations of the equations have been proposed and used in geophysical exploration, however, full use of the equations and their generalization to multiple layers could offer richer information about the properties of the media and be helpful in medical diagnosis via ultrasound. In this work, we investigate how to extract information from the angle-dependent reflection coefficients, including critical angles and the wave distortion at the interface between two and three media. It is shown that it is possible to separate the effect of density from speed of sound mismatch by measuring amplitudes as a function of angle of incidence (AVA). And examining the critical angle and waveform distortion of the reflected waves can reveal the thickness of an intermediate layer, even with subwavelength resolution. These studies could be integrated into medical imaging and also into the training of artificial intelligence systems that assist in diagnosis. In particular, they could help prevent cerebrovascular accidents by early detection of the formation and hardening of plaque in the arteries that irrigate the brain.

Telma Rodrigues Caldeira, Ana Valéria Machado Mendonça

As consultas públicas são um mecanismo de participação social no processo de regulamentação, caracterizadas pelo discurso da garantia da eficácia dos resultados e para promover a democratização do processo e atender a interesses capitalistas. À Agência Nacional de Vigilância Sanitária cabe controlar o risco sanitário de uma série de produtos e serviços, constituindo uma ação de saúde e um instrumento de organização econômica da sociedade em um espaço de disputa de interesses. Este artigo é uma análise documental, que tem como objetivo investigar os atributos e os processos de desenvolvimento das consultas públicas da Agência Nacional de Vigilância Sanitária, de janeiro/1999 a abril/2023. Os documentos encontrados foram categorizados em quatro grupos: legislação específica, relatórios e publicações da agência, relatórios e publicações do Palácio do Planalto e evidências científicas. Foram coletados 654 documentos e evidências, selecionados 350 para leitura de texto completo e extraídos dados de 50 referências. A institucionalização dos procedimentos de melhoria da qualidade regulatória em normas internas, impulsionadas ou não por força de lei e decretos, foi fundamental para estabelecer padrões para publicidade, transparência, prazos e formas de contribuições às consultas públicas, corroborando com evidências de avanços normativos e de desempenho institucional em favor da participação social em espaços decisórios técnicos. Observou-se ainda que um contexto favorável no governo federal decorreu de publicações de normas relacionadas às boas práticas regulatórias na agência em 2008, 2018 e 2021, consolidando institucionalmente as consultas públicas como prática de participação social na normatização sanitária. Entretanto, a participação social nessas consultas é desigual e a diversidade de mecanismos de engajamento na regulamentação pode favorecer grupos com maior poder de organização. Em conclusão, entende-se que são necessários investimentos em iniciativas de educação, enquanto mediadora da participação qualificada, propulsora de consciência política cidadã, poderosa habilidade para democratizar a regulação do risco sanitário.

S. E. Erdni-Garyaev, D. D. Mamedov, D. S. Yurochkin et al.

INTRODUCTION. The Russian Federation has decided to restore the system of compounding pharmacies as an element of the critical national healthcare infrastructure and pharmaceutical supply chain. To improve Russian regulatory practices, develop novel advanced approaches to pharmacy compounding and quality control, and implement these approaches, it is necessary to study relevant good compounding practices applied in healthcare systems of other countries.AIM. This study aimed to analyse the German experience in the organisation and regulation of pharmaceutical compounding to suggest recommendations for the development and implementation of Russian guidelines on good compounding and dispensing practices.DISCUSSION. This work continues a comprehensive study that delves into the current provisions of German legislation governing the system of good compounding practices. This article examines the differences between approaches to compounding by medical and pharmaceutical specialists. According to the findings, compounding by medical specialists is guided by the “free practice of medicine” principle and is subject to minimal regulatory oversight. All pharmacy organisations operate as compounding pharmacies, thereby enhancing the physical accessibility of compounded medicinal products to the population. The authors highlight the features of a process-based quality assurance system encompassing production process controls and quality control methods for compounded medicinal products. Additionally, the authors discuss the applicability of rapid test methods to compounding. In accordance with the German concept of good pharmacy practices, a compounding pharmacy may use the available validation, qualification, and verification tools in the pharmaceutical development of compounded medicinal products and may outsource its internal quality control function. Using the concept of risk ranking, compounding pharmacies may design their own sampling programmes. The article describes approaches to and requirements for organising the evaluation of compounding prescriptions. In Germany, compounding pharmacies may independently assign shelf lives to the compounded medicinal products they produce.CONCLUSIONS. Some German solutions are of considerable practical importance and are applicable to the development and implementation of Russian guidelines on good compounding and dispensing practices. In particular, pharmacy organisations may outsource the quality control of their compounded medicinal products. Additionally, pharmacies can conduct the full-scale pharmaceutical development of compounding technologies and quality control methods (including rapid test methods) using the validation, qualification, and verification tools available to drug manufacturers. Moreover, pharmacy organisations may independently assign shelf lives to the compounded medicinal products they produce. Finally, pharmacy orga­nisations may design their individual sampling programmes.

Yushen Xu, Xiaosong Li, Yuchan Jie et al.

In clinical practice, tri-modal medical image fusion, compared to the existing dual-modal technique, can provide a more comprehensive view of the lesions, aiding physicians in evaluating the disease's shape, location, and biological activity. However, due to the limitations of imaging equipment and considerations for patient safety, the quality of medical images is usually limited, leading to sub-optimal fusion performance, and affecting the depth of image analysis by the physician. Thus, there is an urgent need for a technology that can both enhance image resolution and integrate multi-modal information. Although current image processing methods can effectively address image fusion and super-resolution individually, solving both problems synchronously remains extremely challenging. In this paper, we propose TFS-Diff, a simultaneously realize tri-modal medical image fusion and super-resolution model. Specially, TFS-Diff is based on the diffusion model generation of a random iterative denoising process. We also develop a simple objective function and the proposed fusion super-resolution loss, effectively evaluates the uncertainty in the fusion and ensures the stability of the optimization process. And the channel attention module is proposed to effectively integrate key information from different modalities for clinical diagnosis, avoiding information loss caused by multiple image processing. Extensive experiments on public Harvard datasets show that TFS-Diff significantly surpass the existing state-of-the-art methods in both quantitative and visual evaluations. Code is available at https://github.com/XylonXu01/TFS-Diff.

Amirreza Hashemi, Sayantan Dutta, Bertrand Georgeot et al.

Background noise in many fields such as medical imaging poses significant challenges for accurate diagnosis, prompting the development of denoising algorithms. Traditional methodologies, however, often struggle to address the complexities of noisy environments in high dimensional imaging systems. This paper introduces a novel quantum-inspired approach for image denoising, drawing upon principles of quantum and condensed matter physics. Our approach views medical images as amorphous structures akin to those found in condensed matter physics and we propose an algorithm that incorporates the concept of mode resolved localization directly into the denoising process. Notably, unlike previous studies that considered localization as a hindrance, our approach considers quantum localization as a fundamental component of image reconstruction which is used to differentiate between noisy and non-noisy modes based on diffusivity and localization measurements. This perspective eliminates the need for hyperparameter tuning, making the proposed method a standalone algorithm which can be implemented with minimal manual intervention and can perform automatic filtering of noise regardless of noise level. Through numerical validation, we showcase the effectiveness of our approach in addressing noise-related challenges in imaging and especially medical imaging, underscoring its relevance for possible quantum computing applications.

Karamat Ali, Abdul Hamid

In recent years, the sensitive issue of child labour has received world-wide attention and has become the focus of serious discussion in developing as well as developed countries. Any exact information on child labour is usually hard to come by as most of the children work in the unorganised informal sector, which is neither regulated by labour laws nor is monitored by any organisation. These working children are usually illiterate and start working at a very early age, are inexperienced and vulnerable, they usually work long hours in deplorable conditions, have no medical cover, go without sufficient and proper food and clothing, and get little rest and recreation. In this paper, an attempt has been made to analyse the major causes of female child labour in the city of Multan and certain measures and policies have been suggested which could help in bringing an end to this inhumane practice. Legislation against child labour is not an ideal solution in a country such as Pakistan. The child labour phenomena is not as simple as it appears and needs consideration in the context of the microeconomics of the family and population growth and macroeconomics of the social security structure of a country, unemployment, underemployment, opportunity cost and productivity of formal education. There are very few studies on child labour in Pakistan and on female child labour, hardly any study can be found. Data has been collected for 60 female child labourers, employed as maidservants, baby sitters and other household activities etc. Most of these female children work in the houses of educated and well off people who are usually against child labour. This exploitation of child labour cannot be stopped by child labour laws only. In this regard, other measures such as more facilities for education and vocational training are indispensable. A group of social volunteers comprising workers, employers, government officers, media experts, members of non-government organisations and educationists should make earnest and sincere efforts to achieve the objective of minimising child labour and improve their living conditions as much as possible.

Isabela Tavares Amaral, Felipe Dutra Asensi, Laisa Naiara Euzébio Sá

A judicialização da saúde é um fenômeno que vem ganhando destaque no âmbito nacional e trazendo impactos das mais diversas ordens para o Poder Executivo. Tendo em vista a importância do diálogo institucional para a mitigação dos efeitos desse fenômeno, o objetivo principal do presente artigo foi descrever a operacionalização do arranjo institucional proposto pelo projeto “Construindo o SUS com a Defensoria Pública do Estado do Rio de Janeiro”, discutindo-o em municípios metropolitanos do estado. Trata-se de parte de um estudo de caso único integrado do tipo explicativo. Os dados apresentados neste artigo foram coletados por meio de entrevistas e pesquisa documental. A técnica de tratamento dos dados selecionada foi a construção da explicação. A percepção gerada por meio da experiência apresentada foi a de que a união da instituição jurídica com o poder público municipal, em um mesmo ambiente de trabalho, e a (co)responsabilização pelas demandas em saúde da população criaram em alguns casos e fortaleceram em outros uma atmosfera muito mais propícia ao diálogo institucional, na busca por caminhos de resolutividade das demanda sem saúde. São caminhos em que a via judicial não é uma possibilidade negada e a via administrativa não é uma opção negligenciada, sendo ambas consideradas no atendimento às demandas dos usuários.

Mu Tian, Qinzhu Yang, Yi Gao

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.

Siyi Du, Nourhan Bayasi, Ghassan Hamarneh et al.

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.

Nicholas Konz, Haoyu Dong, Maciej A. Mazurowski

Automated tumor detection in Digital Breast Tomosynthesis (DBT) is a difficult task due to natural tumor rarity, breast tissue variability, and high resolution. Given the scarcity of abnormal images and the abundance of normal images for this problem, an anomaly detection/localization approach could be well-suited. However, most anomaly localization research in machine learning focuses on non-medical datasets, and we find that these methods fall short when adapted to medical imaging datasets. The problem is alleviated when we solve the task from the image completion perspective, in which the presence of anomalies can be indicated by a discrepancy between the original appearance and its auto-completion conditioned on the surroundings. However, there are often many valid normal completions given the same surroundings, especially in the DBT dataset, making this evaluation criterion less precise. To address such an issue, we consider pluralistic image completion by exploring the distribution of possible completions instead of generating fixed predictions. This is achieved through our novel application of spatial dropout on the completion network during inference time only, which requires no additional training cost and is effective at generating diverse completions. We further propose minimum completion distance (MCD), a new metric for detecting anomalies, thanks to these stochastic completions. We provide theoretical as well as empirical support for the superiority over existing methods of using the proposed method for anomaly localization. On the DBT dataset, our model outperforms other state-of-the-art methods by at least 10\% AUROC for pixel-level detection.

Dmitry V. Mikhel

The increased attention of the Orthodox Church to issues of medical education in our country was the result of the fact that in the 1990s it once again became one of the most active forces in our society. The connection between the church and the medical community, which goes back to a time when the doctoring of the mind and bodily health was in fact the work of the same people, cannot leave the church indifferent to the professional formation of healthcare workers. The Soviet era saw the forced de-Christianization of the medical profession and measures taken to abolish medical ethics rooted in the Hippocratic Oath and the Gospel commandments. The restoration of dialogue between church and medicine began after the collapse of the Soviet state, but it is still insufficiently regular. Currently, factors complicating this dialogue are liberal medical legislation and capitalist economies in health care. The former, by legalizing abortion, artificial insemination, and sterilization, absolves the doctor of moral responsibility in matters concerning the management of human life; the latter encourages him to view his profession not as a service, but as a means of making money from other people’s suffering. If the dialogue between church and medicine were to be carried on permanently within the walls of medical schools, it would strengthen the spiritual and moral foundations of the medical profession, upon which it has always existed. One of the most significant forms of this dialogue should undoubtedly be the teaching of biomedical ethics, which should be grounded in modern theology and the values of traditional spiritual cultures of Russia.

Petra Hoogendoorn, Anke Versluis, Sanne van Kampen et al.

BackgroundThe lack of an international standard for assessing and communicating health app quality and the lack of consensus about what makes a high-quality health app negatively affect the uptake of such apps. At the request of the European Commission, the international Standard Development Organizations (SDOs), European Committee for Standardization, International Organization for Standardization, and International Electrotechnical Commission have joined forces to develop a technical specification (TS) for assessing the quality and reliability of health and wellness apps. ObjectiveThis study aimed to create a useful, globally applicable, trustworthy, and usable framework to assess health app quality. MethodsA 2-round Delphi technique with 83 experts from 6 continents (predominantly Europe) participating in one (n=42, 51%) or both (n=41, 49%) rounds was used to achieve consensus on a framework for assessing health app quality. Aims included identifying the maximum 100 requirement questions for the uptake of apps that do or do not qualify as medical devices. The draft assessment framework was built on 26 existing frameworks, the principles of stringent legislation, and input from 20 core experts. A follow-up survey with 28 respondents informed a scoring mechanism for the questions. After subsequent alignment with related standards, the quality assessment framework was tested and fine-tuned with manufacturers of 11 COVID-19 symptom apps. National mirror committees from the 52 countries that participated in the SDO technical committees were invited to comment on 4 working drafts and subsequently vote on the TS. ResultsThe final quality assessment framework includes 81 questions, 67 (83%) of which impact the scores of 4 overarching quality aspects. After testing with people with low health literacy, these aspects were phrased as “Healthy and safe,” “Easy to use,” “Secure data,” and “Robust build.” The scoring mechanism enables communication of the quality assessment results in a health app quality score and label, alongside a detailed report. Unstructured interviews with stakeholders revealed that evidence and third-party assessment are needed for health app uptake. The manufacturers considered the time needed to complete the assessment and gather evidence (2-4 days) acceptable. Publication of CEN-ISO/TS 82304-2:2021 Health software – Part 2: Health and wellness apps – Quality and reliability was approved in May 2021 in a nearly unanimous vote by 34 national SDOs, including 6 of the 10 most populous countries worldwide. ConclusionsA useful and usable international standard for health app quality assessment was developed. Its quality, approval rate, and early use provide proof of its potential to become the trusted, commonly used global framework. The framework will help manufacturers enhance and efficiently demonstrate the quality of health apps, consumers, and health care professionals to make informed decisions on health apps. It will also help insurers to make reimbursement decisions on health apps.

Ontran Sumantri Riyanto, Hetty W.A. Panggabean, Erik Adik Putra Bambang Kurniawan et al.

Surgery is an effort to treat or help to treat a patient's disease. The increasing demand for health care is to increase, whether it is expected that a nurse improves her knowledge and skills, such as a surgical nurse specialist. This research uses normative juridical research methods, using secondary data sources including legislation, books, journals, court decisions, and other literature. The result of the study is that the surgical nurse while performing his duties both in the surgical room, before surgery, during surgery and after surgery must be based on SOPs and mandates from doctors made in writing. That way a surgical nurse if working is in accordance with the procedures and directions of the doctor, the surgical nurse cannot be sanctioned in the event of a medical dispute because his responsibility rests with the surgeon.

Cheng Ouyang, Chen Chen, Surui Li et al.

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.

Mengting Xu, Tao Zhang, Zhongnian Li et al.

Deep learning models (with neural networks) have been widely used in challenging tasks such as computer-aided disease diagnosis based on medical images. Recent studies have shown deep diagnostic models may not be robust in the inference process and may pose severe security concerns in clinical practice. Among all the factors that make the model not robust, the most serious one is adversarial examples. The so-called "adversarial example" is a well-designed perturbation that is not easily perceived by humans but results in a false output of deep diagnostic models with high confidence. In this paper, we evaluate the robustness of deep diagnostic models by adversarial attack. Specifically, we have performed two types of adversarial attacks to three deep diagnostic models in both single-label and multi-label classification tasks, and found that these models are not reliable when attacked by adversarial example. We have further explored how adversarial examples attack the models, by analyzing their quantitative classification results, intermediate features, discriminability of features and correlation of estimated labels for both original/clean images and those adversarial ones. We have also designed two new defense methods to handle adversarial examples in deep diagnostic models, i.e., Multi-Perturbations Adversarial Training (MPAdvT) and Misclassification-Aware Adversarial Training (MAAdvT). The experimental results have shown that the use of defense methods can significantly improve the robustness of deep diagnostic models against adversarial attacks.