2025年 · 第05卷第03期

Coronary computed tomography angiography (CCTA), which enables noninvasive assessment of luminal stenosis and atherosclerotic plaque components, has become the first-line technique for evaluating coronary artery disease. Artificial intelligence (AI) has the potential to revolutionize the CCTA workflow. However, it is crucial to evaluate the effectiveness and feasibility of AI algorithms before their clinical deployment. This expert consensus proposes three fundamental elements of research designs of AI in CCTA and offers corresponding recommendations. The consensus also reviews the existing evidence on AI applications in CCTA and provides recommendations on the current clinical applications of AI, including image acquisition and reconstruction, postprocessing, diagnosis, prognostic prediction, guiding prevention and treatment, and cardiovascular disease prevention.

Artificial intelligence (AI) and digital twin technologies exhibit significant potential in analyzing and integrating multidimensional datasets and offer novel perspectives for the management of chronic diseases including diabetes. These technologies offer opportunities for personalizing treatment and potentially reversing the conditions. This review systematically evaluated the advantages and limitations of AI applications, potential for predictive analytics in formulating personalized management strategies, and practical roles of AI and digital twin technologies in diabetes diagnosis and treatment. Special attention was given to their strengths and weaknesses in disease prediction, early detection, and development of individualized management strategies.

AI algorithms have demonstrated great efficiency in analyzing large datasets, aiding in the early identification and intervention of prediabetes. Machine learning algorithms, including deep learning neural networks, integrate lifestyle, genetic, and other influencing factors to accurately predict the progression of prediabetes to diabetes. Moreover, AI-driven wearable devices and mobile applications provide real-time monitoring and personalized guidance, thereby effectively mitigating diabetes. This study also explored the challenges of integrating AI and digital twin technologies into clinical practice for diabetes management and broader healthcare domains, focusing on data privacy, need for diverse and comprehensive datasets, and the importance of integrating AI tools into clinical workflows.

Rectal cancer remains a major global health challenge, prompting ongoing efforts to optimize treatment strategies. In recent years, organ-preserving approaches—particularly the "watch-and-wait" strategy—have gained growing interest. Concurrently, the advent of artificial intelligence (AI) has opened new avenues in personalized oncology. This review explored the emerging role of AI in the individualized management of rectal cancer, with a focus on its potential to improve treatment outcomes and patient prognosis. Herein, we provide a comprehensive synthesis of recent studies investigating AI applications in predicting pathological complete response, metastasis, and disease-free survival following neoadjuvant therapy. These studies employ diverse data modalities, including radiomics (magnetic resonance imaging (MRI), computerized tomography (CT), and endoscopy), clinical parameters, and other omics-based features. The study evaluated the predictive models developed using machine learning and deep learning algorithms, discussing their performance metrics, strengths, and limitations. Despite the ongoing challenges—such as limited data availability, lack of model interpretability, and suboptimal predictive accuracy—AI has demonstrated potential to outperform conventional assessment methods in select areas. These findings may highlight the growing significance of AI in supporting personalized, evidence-based decision-making in rectal cancer care.

Narcolepsy is a chronic neurological disorder that disrupts the sleep-wake cycle and manifests in symptoms like excessive daytime sleepiness (EDS), cataplexy, and rapid transitions into rapid eye movement (REM) sleep. Its variable prevalence, genetics, and clinical presentations pose considerable challenges in diagnosis and management. Here, we synthesized the advances in neuroimaging techniques and their substantial contributions to the narcolepsy complex pathology. We analyzed the structural magnetic resonance imaging (MRI) scan findings that highlight gray matter reductions and cortical thinning in patients with narcolepsy. Additionally, we explored findings from diffusion tensor imaging (DTI) scans that shed light on compromises in white matter integrity. Functional MRI and positron emission tomography (PET) scan studies further illuminated neurochemical deficits and altered brain connectivity. The implications of these findings extend beyond diagnosis, suggesting potential targets for neuromodulation therapies and calling for larger, more standardized studies to enhance both our understanding and treatment approaches for narcolepsy. Despite such advances, this field continues to meet challenges, including limitations in sample size and the need for comprehensive longitudinal and multimodal studies. This review highlighted the potential of neuroimaging combined with machine learning and advanced analytics, which help to discover novel biomarkers, refine the comprehension of narcolepsy and its neurochemical intricacies, and improve the therapeutic strategies.

Accurate early warning of sepsis onset is crucial for reducing mortality. However, the inter-individual heterogeneity in clinical manifestations of sepsis leads to significant sparsity of data. The current time series analysis methods attempt to interpolate highly sparse sepsis data, yielding unsatisfactory results. In this study, we aimed to develop an efficient artificial intelligence approach for early warning of sepsis onset.

The I2former model, an incident-induced attention-based architecture, was proposed to address the challenges posed by sparse medical data. This model employs a novel increment entropy encoding strategy to extract clinically significant features from sparse data, effectively transforming the unavailable data into valuable insights. The training data were sourced from MIMIC-IV v2.2 and eICU v2.0, with external validation from Beijing Tsinghua Changgung Hospital. Five advanced models, including the Autoformer, Timesnet, Informer, Reformer, and DLinear, currently in use were used for comparison.

Five metrics used for classification indicated that the I2former significantly outperformed the 5 advanced time series analysis methods, achieving area under the receiver operating characteristic (AUROC), area under the precision-recall curve (AUPRC), Matthews correlation coefficient (MCC), F1-score, and accuracy of 0.886, 0.529, 0.449, and 0.917, respectively. Furthermore, external validation using the data from Beijing Tsinghua Changgung Hospital demonstrated that the model provides accurate early warnings, on average of 15.5 h prior to sepsis onset.

Therefore, I2former is proposed for accurate early warning of sepsis onset. Five crucial metrics for classification underscored the substantial advantages of I2former in managing sparse data, while highlighting its potential application and value in the field of medical data analysis.

Generalized knee tissue segmentation, such as cartilage and meniscus in magnetic resonance imaging (MRI), plays a vital role in the clinical assessment of knee osteoarthritis (OA). However, domain variability between MRI datasets poses a significant challenge for the application of robust segmentation methods in real-world clinical settings. Existing unsupervised domain adaptation (UDA) approaches, which rely on one-to-one assumptions between the source and target domains, often fail to preserve knee tissues such as cartilage and meniscus, which are critical for OA diagnosis in diverse clinical settings.

We propose a source-independent segmentation approach tailored for multi-domain knee MRI datasets. Our method emphasizes knee tissue regions to reduce domain gaps and label inconsistencies. By introducing a stepwise adaptation strategy, segmentation performance was refined progressively from intermediate domains to the final target domain. Pseudo-label attention mechanisms were integrated into the adaptation pipeline, enabling iterative fine-tuning of domain-specific segmentations while leveraging unidirectional generative adversarial networks to enhance tissue-specific adaptation. This iterative training process ensures the generation of reliable pseudo-labels, thereby improving segmentation accuracy in diverse clinical MRI datasets.

We demonstrated the effectiveness of our approach on the OA initiative dataset as the source domain and self-collected, T1-weighted fast field echo (T1FFE) as the intermediate domain and three-dimensional fast spin echo (3D FSE) as the final target domain. Our method achieved an average dice scores of 0.8701 and 0.7990 for source and target domains, respectively, surpassing the typical UDA methods explored in our experiments.

The experiments conducted on clinical MRI data, spanning OA severity from healthy knees to KL Grades 1-4, validated the effectiveness of the proposed domain adaptation method in precise segmentation of the cartilage and meniscus.

Based on the development background of digital medical technology, this study aimed to establish design guidelines and references in relevant fields to better serve clinical medical treatment using intelligent technology to enhance the usability of the interaction interface of robotic surgical systems and reduce potential human-factor risks during digital surgery.

Considering the robotic liver cancer ablation surgery system as the research object, subjective and objective evaluation indicators were established from 3 dimensions of effectiveness, efficiency, and satisfaction based on the usability theory. Using the hierarchical task analysis method, usability experiments were conducted to collect relevant data. Feedback on issues during the experimental process was obtained through observation and interviews. Failure mode and effect analysis and fault tree analysis were used to assess risk levels and formulate design strategies.

The interface design of the liver cancer ablation surgery robot was iteratively optimized. The results showed that the interface after iteration improved in skilled operation time, subjective evaluation scores, risk priority number value, and risk level. The rationality of the scheme was verified, and interface design paradigm was constructed based on intelligent technology.

After improving the design, the interface effectively reduced the frequency of problems and average skilled operation time, thereby, improving the subjective satisfaction score of users.

对小编来说，

对小编来说，

对小编来说，

对小编来说，

对小编来说，