Lung cancer is a leading cause of cancer-related morbidity and mortality worldwide. Coupled with the substantial workload, the clinical management of lung cancer is challenged by the critical need to efficiently and accurately process increasingly complex medical information. In recent years, large language models (LLMs) technology has undergone explosive development, demonstrating unique advantages in handling complex medical data by leveraging its powerful natural language processing capabilities, and its application value in the field of lung cancer diagnosis and treatment is continuously increasing. The paper systematically analyzes that the exceptional potential of LLMs in lung cancer auxiliary diagnosis, tumor feature extraction, automatic staging, progression/outcome analysis, treatment recommendations, medical documentation generation, and patient education. However, they face critical technical and ethical challenges including inconsistent performance in complex integrated decision-making (e.g., TNM staging, personalized treatment suggestions) and "black box" opacity issues, along with dilemmas such as training data biases, model hallucinations, data privacy concerns, and cross-lingual adaptation challenges ("data colonization"). Future directions should prioritize constructing high-quality multimodal corpora specific to lung cancer, developing interpretable and compliant specialized models, and achieving seamless integration with existing clinical workflows. Through dual drivers of technological innovation and ethical standardization, LLMs should be prudently advanced for holistic lung cancer management processes, ultimately promoting efficient, standardized, and personalized diagnosis and treatment practices.

With the continuous deepening of the concept of precision diagnosis and treatment for lung cancer, how to achieve higher efficiency and accuracy in the screening, diagnosis, and treatment pathways in clinical practice has become an important issue that urgently needs to be overcome. The current clinical difficulty lies in the fact that despite continuous advancements in imaging and molecular diagnostic technologies, there are still limitations in manual efficiency and subjective experience when it comes to massive data analysis and multi-scale feature extraction. Artificial intelligence (AI), especially algorithm systems based on deep learning, is an innovative technology capable of deeply empowering medical big data. This method utilizes algorithms such as convolutional neural networks, combined with radiomics, pathomics, and multi-modal data fusion analysis, demonstrating immense potential in early precise detection and benign-malignant differentiation of pulmonary nodules, digital pathological subtype recognition and non-invasive prediction of driver genes, precise 3D surgical planning and automatic delineation of radiotherapy target volumes, as well as dynamic risk warning during follow-up. This innovative technology provides a brand-new solution for realizing intelligent and individualized lung cancer diagnosis and treatment models. This consensus, based on the latest evidence from evidence-based medicine and combined with the development trends in the AI field and real-world clinical needs, was ultimately formed by gathering the consensus opinions of multidisciplinary experts in radiology, pathology, thoracic surgery, and other fields. The main content covers the application specifications of AI in the three core scenarios of lung cancer screening, diagnosis, and treatment, the technical standards for data collection and algorithm validation, as well as the ethical and regulatory challenges faced at the current stage. It aims to clarify the applicable boundaries of AI as a clinical auxiliary decision support tool, providing scientific guidance and standardized exploration directions for peers currently engaged in or planning to carry out AI-assisted clinical diagnosis, treatment, and translation of lung cancer.

To address the growing challenge of an increasing number of elderly lung cancer patients amidst China's aging population and to fill the gap in quality control standards for surgical treatment in this special population, this study aimed to develop a standardized surgical evaluation system for elderly lung cancer patients tailored to China's national conditions. The system was established through a literature review, integrated the pathophysiological characteristics of elderly patients, and was constructed following review, feedback, and revision by experts from multiple thoracic surgery centers. Employing a 100-point scoring system, it comprises three primary domains: physical infrastructure and geriatric adaptability foundational conditions (10 points); management level and perioperative care models (20 points); and technical proficiency and clinical outcomes (70 points). The system places a strong emphasis on geriatric adaptability, proposing specific, quantifiable indicators for age-friendly facility modifications, control of elderly-specific complications, multidisciplinary collaboration, and standardized perioperative management. It provides a convenient and measurable assessment tool for quality control in the surgical treatment of elderly lung cancer in China, which is expected to promote the standardization and homogenization of diagnosis and treatment.

ObjectiveTo systematically review the effect of dual-task training (DTT) on mobility and balance in stroke patients. MethodsCNKI, Wanfang Data, VIP, PubMed, Web of Science, Embase and Cochrane Library were searched for randomized controlled trials (RCT) on DTT in stroke rehabilitation from inception to June, 2025. The quality of the included studies was assessed using the Cochrane Risk of Bias Tool and the PEDro scale. Meta-analysis was performed using RevMan 5.4. ResultsA total of eleven RCT involving 517 stroke patients were included, and the scores of PEDro scale were four to eight. DTT was more effective on the score of Fugl-Meyer Assessment-Lower Extremity (MD = 3.70, 95%CI 1.88 to 5.53, P < 0.001) and Berg Balance Scale (BBS) (MD = 3.04, 95%CI 1.58 to 4.49, P < 0.001), fall incidence (OR = 0.32, 95%CI 0.11 to 0.87, P = 0.025), gait speed under a single-task condition (MD = 0.08, 95%CI 0.03 to 0.12, P = 0.002), gait speed under a dual-task condition (MD = 0.13, 95%CI 0.10 to 0.16, P < 0.001), Timed Up and Go Test time (MD = -2.78, 95%CI -3.85 to -1.72, P < 0.001), and score of modified Barthel Index (MD = 4.77, 95%CI 1.88 to 7.66, P = 0.001). Subgroup analysis indicated that motor-motor DTT was more effective than cognitive-motor DTT on BBS scores and single-task gait speed. ConclusionsDTT can improve mobility and balance in stroke patients, especially with motor-motor tasks.

Artificial intelligence (AI) has made preliminary advances in the management of postoperative follow up for lung cancer; however, a systematic review of its application value across the entire follow-up continuum remains lacking. Taking the four core dimensions of postoperative follow-up management as its framework, including surveillance candidate selection, follow-up interval optimization, follow-up protocol design, and follow-up modality selection, this review examines the progress of AI in patient risk stratification, surveillance frequency optimization, content design, and supportive platform development. The review further delineates the current challenges and future directions for AI in this domain, and ultimately seeks to promote the standardized and scaled application of AI in postoperative follow-up management for lung cancer, with the goal of establishing a "care beyond hospitalization" life-cycle management framework that improves the long-term quality of survival for patients undergoing lung cancer surgery.

Objective To evaluate the performance of lightweight Chinese large language models (LLMs) in answering specialized lung cancer questions, and to explore the impact of retrieval-augmented generation (RAG) on model performance. Methods Eleven lightweight Chinese LLMs with parameter sizes ranging from 7B to 32B were included. A lung cancer-specific evaluation dataset consisting of 200 questions [100 A1-type (basic knowledge) and 100 A2-type (clinical case) questions], constructed based on clinical guidelines and thoracic surgery textbooks, was used for assessment. Model performance was evaluated under two conditions (with and without RAG). Accuracy was used to assess model performance, and response latency was recorded to reflect inference efficiency. An accuracy–latency scatter plot was constructed for descriptive analysis of overall model performance. Results All models successfully completed the evaluation. With the introduction of RAG, the overall average accuracy improved from 61.68% to 76.36%. Smaller models demonstrated the most significant improvement (e.g., the accuracy of DeepSeek-7B increased from 32.50% to 60.00%, P<0.001). The average response latency increased from 12.58 s to 13.80 s. The Qwen3 series showed the best overall performance, and Qwen3-32B achieved the highest accuracy under both conditions (76.50% and 84.00%, respectively). After RAG integration, performance differences among model families were markedly reduced. Based on the accuracy-latency trade-off, Qwen3-32B achieved the best balance between accuracy and response latency under the baseline condition, whereas Qwen3-14B demonstrated superior overall performance in terms of accuracy, latency, and computational cost after RAG integration. Conclusion The integration of RAG technology improves the ability of lightweight Chinese LLMs to answer specialized lung cancer questions. Under the dual practical constraints of limited computational resources and medical data security requirements, the "lightweight model+RAG" technical framework may represent a promising deployment solution.

With the widespread adoption of low-dose CT screening and the extensive application of high-resolution CT, the detection rate of sub-centimeter lung nodules has significantly increased. How to scientifically manage these nodules while avoiding overtreatment and diagnostic delays has become an important clinical issue. Among them, lung nodules with a consolidation tumor ratio less than 0.25, dominated by ground-glass shadows, are particularly worthy of attention. The therapeutic challenge for this group is how to achieve precise and complete resection of nodules during surgery while maximizing the preservation of the patient's lung function. The "watershed topography map" is a new technology based on big data and artificial intelligence algorithms. This method uses Dicom data from conventional dose CT scans, combined with microscopic (22-24 levels) capillary network anatomical watershed features, to generate high-precision simulated natural segmentation planes of lung sub-segments through specific textures and forms. This technology forms fluorescent watershed boundaries on the lung surface, which highly fit the actual lung anatomical structure. By analyzing the adjacent relationship between the nodule and the watershed boundary, real-time, visually accurate positioning of the nodule can be achieved. This innovative technology provides a new solution for the intraoperative positioning and resection of lung nodules. This consensus was led by four major domestic societies, jointly with expert teams in related fields, oriented to clinical practical needs, referring to domestic and foreign guidelines and consensus, and finally formed after multiple rounds of consultation, discussion, and voting. The main content covers the theoretical basis of the "watershed topography map" technology, indications, operation procedures, surgical planning details, and postoperative evaluation standards, aiming to provide scientific guidance and exploration directions for clinical peers who are currently or plan to carry out lung nodule resection using the fluorescent microscope watershed analysis method.

Surgical treatment is one of the key approaches for non-small cell lung cancer (NSCLC). Regular postoperative follow-up is crucial for early detection and timely management of tumor recurrence, metastasis, or second primary tumors. A scientifically sound and reasonable follow-up strategy not only extends patient survival but also significantly improves quality of life, thereby enhancing overall prognosis. This consensus aims to build upon the previous version by incorporating the latest clinical research advancements and refining postoperative follow-up protocols for early-stage NSCLC patients based on different treatment modalities. It provides a scientific and practical reference for clinicians involved in the postoperative follow-up management of NSCLC. By optimizing follow-up strategies, this consensus seeks to promote the standardization and normalization of lung cancer diagnosis and treatment in China, helping more patients receive high-quality care and long-term management. Additionally, the release of this consensus is expected to provide insights for related research and clinical practice both domestically and internationally, driving continuous development and innovation in the field of postoperative management for NSCLC.

Objective To evaluate the diagnostic performance of the minimum-cost-path-based CT angiography-derived fractional flow reserve(MCP-FFR)and the deep learning-driven CT angiography-derived fractional flow reserve(DeepCL-FFR),and to particularly explore the potential value of the DeepCL algorithm in improving diagnostic accuracy within the"gray zone."Methods A retrospective analysis was conducted on 151 coronary vessels from 109 patients with coronary artery disease,who were hospitalized at the General Hospital of the People's Liberation Army between January 2020 and June 2021.Pearson correlation and Bland-Altman plots were employed to assess the correlation and agreement of the two CT-FFR methods with invasive FFR.A CT-FFR range of 0.70-0.80 was defined as the diagnostic"gray zone."The accuracy,sensitivity,specificity,positive predictive value,and negative predictive value for detecting hemodynamic abnormalities were calculated and analyzed.The DeLong test was used to compare the areas under the receiver operating characteristic curves(AUC)between the two CT-FFR calculation methods.Results Both CT-FFR methods exhibited a positive correlation with invasive FFR(MCP-FFR:r=0.75,P＜0.001;DeepCL-FFR:r=0.86,P＜0.001)and showed good agreement(MCP-FFR:mean difference=0.010,P=0.351;DeepCL-FFR:mean difference=-0.003,P=0.772).Both DeepCL-FFR(AUC 0.97,95％CI 0.94-0.99)and MCP-FFR(AUC 0.92,95％CI 0.88-0.97)demonstrated favorable diagnostic performance for detecting hemodynamic abnormalities(P=0.122).In the"gray zone"for hemodynamic abnormality,the diagnostic accuracy of MCP-FFR was 68.8％,whereas DeepCL-FFR increased it to 89.7％.DeepCL-FFR also exhibited superior diagnostic performance(AUC 0.89,95％CI 0.73-0.99)within the"gray zone,"which was significantly higher than that of MCP-FFR(AUC 0.71,95％CI 0.54-0.87)(P＜0.001).Conclusions The deep learning-driven coronary centerline extraction algorithm,DeepCL,demonstrates superior diagnostic performance in CT-FFR for detecting hemodynamic abnormalities,particularly by significantly improving diagnostic accuracy in the"gray zone."

Objective:To analyze the differences in determination of fluoride level in drinking water by ion-selective electrode method and high-throughput rapid determination method.Methods:The precision test was carried out by using the two methods to measure two kinds of fluoride standard substances, water samples of external quality control assessment from 2021 to 2023 (two kinds each year) and the fluoride level in three drinking water samples (for 5 times/each sample). Accuracy testing was conducted by measuring the external quality control assessment water samples and the spiked recovery rates drinking water, and water samples were grouped (water fluoride ≤1.00, > 1.00 mg/L) and analyzed according to the "Hygienic Standards for Drinking Water" (GB 5749-85). SPSS 23.0 software was used for statistical analysis of the measurement results.Results:(1) The correlation coefficients ( r) of the working curves of the two methods were both > 0.990, meeting the quality control requirements. (2) In the precision test, when comparing the results of the two methods for detecting two kinds of fluoride standard substances, there was no statistically significant difference ( F = 0.36, 0.15, P = 0.564, 0.707), and the coefficients of variation ( CV) were all < 5%. The CV of the detection results of the external quality control assessment water samples and drinking water samples were < 5%. (3) In the accuracy test, when the fluoride concentration in water was ≤1.00 mg/L, there was no statistically significant difference in the spiked recovery rates between the two methods ( F = 0.49, P = 0.504). When the fluoride concentration in water was > 1.00 mg/L, there was a statistically significant difference in the spiked recovery rates between the two methods ( F = 24.75, P = 0.003). Conclusions:The ion-selective electrode method has the advantages of wide detection range and wide adaptability, while the high-throughput rapid determination method has high accuracy. Testing personnel can weigh and choose the appropriate determination method based on the actual laboratory conditions and sample concentration range.