Radiofrequency ablation (RFA) is a widely used minimally invasive treatment for non-surgical hepatocellular carcinoma. This review synthesizes the technical principles, core components, and key modeling aspects of RFA. RFA induces tumor necrosis via Joule heating from ionic vibration. Electrode needle design critically impacts ablation efficacy and safety, with multipolar needles offering larger zones yet posing power and tissue risks. Crucially, biological tissue parameters exhibit dynamic spatial and thermal variations, necessitating nonlinear modeling for accurate temperature prediction. While the Pennes bioheat model remains mainstream for its simplicity, more advanced models (e.g., porous medium) enhance physiological realism. Thermal damage assessment commonly employs the Arrhenius model and isothermal thresholds, aided by real-time monitoring for intraoperative precision. Future research should prioritize the development of smart electrodes, creation of personalized tissue parameter databases, and exploration of multi-energy techniques to shift RFA from an "empirically oriented" approach to an "accurate prediction" paradigm, ultimately improving hepatocellular carcinoma patient survival and quality of life.

Images taken under hazy weather conditions suffer from problems such as blurring, low contrast, and low saturation due to the scattering of atmospheric light by aerosol particles in the air, which affects the performance and judgment of image analysis equipment. With the rapid development of image processing technology and computer vision technology, researchers have proposed a large number of targeted haze removal algorithms to improve the quality of images taken under hazy weather conditions. According to the haze removal principle, mainstream haze removal algorithms can be classified into three categories: image enhancement-based, physics model-based, and neural network-based. This paper introduces and explores classic haze removal algorithms from the perspectives of principles, development, advantages, and disadvantages, and outlines the prospects for the future development and application direction of haze removal algorithms.

Objective: Diabetes clinical teaching has long relied on theoretical instruction and static case studies, which leaves residents ill-prepared to manage dynamic blood glucose fluctuations and complex clinical scenarios. This study aims to explore the effectiveness of integrating Ambulatory Glucose Profile (AGP) with scenario simulation in enhancing residents’ clinical reasoning for diabetes management. Methods: Sixty endocrinology residents were randomly assigned to a control group and an intervention group. Both groups underwent the same conventional clinical teaching curriculum. Additionally, the control group completed an 8-hour AGP training program consisting of 4 hours of didactic instruction and 4 hours of self-directed practice with static case analysis. The intervention group received an 8-hour integrated program combining AGP analysis with high-fidelity scenario simulation and structured debriefing according to Kolb’s experiential learning cycle. Teaching outcomes were evaluated through theoretical assessments (primary outcome), Objective Structured Clinical Examinations (OSCEs), Mini-Clinical Evaluation Exercises (Mini-CEX), and satisfaction surveys. Results: The intervention group demonstrated significantly higher scores than the control group in the primary outcome (theoretical knowledge), across all six OSCE stations (including history taking, physical examination, skill operation, medical communication, medical record writing, and case analysis), as well as in multiple Mini-CEX dimensions (history taking, physical examination, communication skills, clinical judgment, and organizational efficiency) (all P<0.05). No significant differences were observed between the two groups in humanistic care and teaching satisfaction (P>0.05). Conclusions: The integrated teaching approach combining AGP with scenario simulation demonstrated preliminary effectiveness in enhancing residents’ comprehensive clinical management skills and may serve as a valuable method for training clinical reasoning in diabetes care.

The management of pulmonary nodules is gradually shifting from traditional "single image judgment" to "continuous risk management" mode, especially in the fields of ground glass nodules and microinvasive lung adenocarcinoma. The importance of artificial intelligence (AI) and medical GPT assisted decision-making is becoming increasingly prominent. This case presents a typical case where postoperative pathology was successfully corrected through BAIGPT pulmonary nodule intelligent agent evaluation. The patient is a 67 year old female with no smoking history. She visited Professor Bai's AI clinic in July 2025 due to mixed nodules in the left lung. The AI system evaluation showed that the probability of malignancy of the nodule was as high as 88%. Although the first postoperative pathological report indicated "no evidence of malignant tumor", after AI assisted pathological review and molecular testing, it was ultimately confirmed to be microinvasive lung adenocarcinoma. This case suggests that pulmonary nodule intelligent agents not only have the ability to "detect lesions", but are also entering a new stage of "cognition reasoning error correction closed-loop follow-up", which has important clinical value in reducing missed diagnosis of early lung cancer.

Key Words: pulmonary nodule; pathology; agent

Minimally invasive surgery (MIS) has been developed based on the concept of minimum trauma and maximal function, epitomizing the advancement in modern surgery. Following the introduction of laparoscopy in the 1980s, MIS has continued to evolve into multiple modalities, including single-incision surgery, natural orifice transluminal endoscopic surgery, and robotic surgery. The advancements in imaging, artificial intelligence (AI) and robotization have come together to enhance not only the precision and safety of operations, but also the visualization and technical feasibility even in very complex cases. Numerous investigations have shown that MIS, compared with open surgery, decreases blood loss, reduces complication rates, shortens hospital day, and improves postoperative quality of life. The use of AI has brought surgery to the new age of machine intelligence and data-based decision making. Nevertheless, ethical and legal barriers, inadequate physician training programs, and uneven distribution of resources are obstacles to its wider implementation. The next wave of MIS, will be characterized by intelligence and personalization, incorporating AI navigation, augmented reality, and a multi-disciplinary approach.

Chronic respiratory diseases are characterized by prolonged courses, fluctuating symptoms, complex care pathways, and a substantial reliance on out-of-hospital management, placing high demands on patients’ disease understanding, operational skills, self-monitoring, and long-term adherence. Patient education and management are the core determinants of symptom control, exacerbation prevention, rehabilitation participation, and improvement of long-term outcomes. Recent advances in artificial intelligence, medical large language models, digital humans, virtual experts, the Internet of Things, and metaverse medicine, are reshaping respiratory health and management from one-time in-clinic instruction toward a model that is continuous, individualized, contextualized, and interactive. In chronic respiratory disease management, these technologies of digital humans and virtual experts have shown promise in disease cognition building, inhaler instruction, interpretation of written action plans, pulmonary rehabilitation training, remote follow-up, and long-term health support, creating new opportunities for integrated hospital-community-home care. Nonetheless, current evidence remains constrained by substantial heterogeneity, a lack of hard clinical endpoints, limited adaptation for older adults and people with low health literacy, insufficient algorithmic transparency, and unclear boundaries of responsibility. The concepts of metaverse medicine, medical GPT, and BAIMGPT proposed by Professor Chunxue Bai and colleagues provide important theoretical and technical support for localized practice in this field. Therefore, this study aims to integrate recent international reviews, landmark studies, consensus guidelines and research from Professor Bai’s group, and systematically review the theoretical foundation, core applications, technical system, challenges and future directions of digital humans, virtual experts and patient education in respiratory health communication, so as to inform innovative strategies for chronic respiratory disease management.

Key Words: digital human; virtual expert; BAIMGPT; internet of things; metaverse medicine; patient education and management; chronic respiratory disease