OBJECTIVE To preliminarily assess the horizontal sound localization and its influencing factors in patients with unilateral sudden sensorineural hearing loss during the acute phase.METHODS The azimuth discrimination test and azimuth identification test were completed,with the speech sound(65 dB SPL)as the stimulus.The minimum audible angle(MAA)and root-mean-square error(RMSE)were obtained,and the RMSE of the affected side and the healthy side were calculated respectively.According to the WHO(2021)hearing loss classification criteria,the data were analyzed based on the pure-tone average(PTA)of the affected ear.And the best resident hearing at each frequency of the affected ear was recorded.RESULTS The performance of the unilateral sudden sensorineural hearing loss patients in the sound localization varied greatly.Some performed close to the normal level,while others completely lost the ability to localize sound.The RMSE of the moderate hearing loss group(≥35 dB HL)was significantly higher than that of the normal hearing group(P<0.01),the MAA of the moderate to severe hearing loss group(≥50 dB HL)showed statistically significant differencescompared with normal hearing group(P<0.001).The RMSE of the affected side of patients in the severe and above hearing loss group was significantly larger than that of the healthy side.Regression analysis showed that the best resident hearing at each frequency of the affected ear was the most significant factor affecting MAA(R2=0.572,P<0.001)and RMSE(R2=0.768,P<0.001).CONCLUSION The horizontal sound localization of unilateral sudden sensorineural hearing loss patients in the acute phase varies greatly.When the PTA of the affected side reaches moderate hearing loss,the localization ability is significantly lower than that of normal-hearing individuals.The best resident hearing at each frequency of the affected ear is the key factor affecting the localization ability.

To investigate the effectiveness of a self-developed intelligent medical record writing assistant in enhancing the efficiency of discharge record writing and improving the quality of discharge records, and to assess physicians' satisfaction with the assistant.

With the rapid advancement of artificial intelligence technology, chatbots have shown great potential in the healthcare sector. From personalized health advice to chronic disease management and psychological support, chatbots have demonstrated significant advantages in improving the efficiency and quality of healthcare services. As the scope of their applications expands, the relationship between technological complexity and practical application scenarios has become increasingly intertwined, necessitating a more comprehensive evaluation of both aspects. This paper, from the perspective of he althcare applications, systematically reviews the technological pathways and development of chatbots in the medical field, providing an in-depth analysis of their performance across various medical scenarios. It thoroughly examines the advantages and limitations of chatbots, aiming to offer theoretical support for future research and propose feasible recommendations for the broader adoption of chatbot technologies in healthcare.

Generative Artificial Intelligence ishows a broad application prospect in the field of healthcare and has become an important technical means to promote the development of medical informatization.It addresses the multi-faceted challenges of medical record documentation,including efficiency,quality,and doctor-patient communica-tion.It analyzes the adaptability and feasibility of Generative Artificial Intelligence in different clinical scenarios of intelli-gent medical record generation.Additionally,it explores the issues present in current applications and proposes corre-sponding solutions,providing references for the effective application and continuous optimization of Generative Artifi-cial Intelligence in medical record documentation.This provides a theoretical foundation for further expanding the appli-cation scenarios of automatic medical record documentation in China's healthcare industry.

Generative Artificial Intelligence ishows a broad application prospect in the field of healthcare and has become an important technical means to promote the development of medical informatization.It addresses the multi-faceted challenges of medical record documentation,including efficiency,quality,and doctor-patient communica-tion.It analyzes the adaptability and feasibility of Generative Artificial Intelligence in different clinical scenarios of intelli-gent medical record generation.Additionally,it explores the issues present in current applications and proposes corre-sponding solutions,providing references for the effective application and continuous optimization of Generative Artifi-cial Intelligence in medical record documentation.This provides a theoretical foundation for further expanding the appli-cation scenarios of automatic medical record documentation in China's healthcare industry.

Objectives:To examine the feasibility of using foundation model in computer vision for echocardiographic left ventricular ejection fraction measurement. Methods:Based on the most extensive publicly accessible repository of echocardiographic loops,EchoNet-Dynamic,featuring 10024 recordings from individual patients,a foundation model in computer vision,VideoMAE V2,was fine-tuned,validated,tested using 7460,1288,and 1276 echocardiographic loops,respectively. Results:The mean absolute error between left ventricular ejection fraction measurements of VideoMAE V2 and expert's measurements was 3.94％ (95％CI:3.79％-4.11％).The Pearson's correlation coefficient was 0.91 (95％CI:0.89-0.92).Additionally,VideoMAE V2 demonstrated exceptional accuracy in identifying patients with a left ventricular ejection fraction below 50％,achieving an AUC of 0.96 (95％CI:0.95-0.97). Conclusions:This study validates the feasibility of using foundation model in computer vision for measuring left ventricular ejection fraction in echocardiographic loops and lays the foundation for the development of a generalized multimodal automated interpretation system for echocardiography.

Objectives:To develop a novel prediction model for mitral valve repair failure based on machine learning algorithms.Methods:Clinical and echocardiographic data were analyzed on patients,who underwent mitral valve repair in Fuwai Hospital from 2009 January 1st to 2022 December 31st.End points included immediate mitral valve repair failure (mitral replacement secondary to mitral repair failure) and recurrence regurgitation (moderate or severe mitral regurgitation before discharge).Risk factors of mitral valve repair failure were analyzed by XGBoost and shapley additive explanation (SHAP),and a machine learning model was established based on mixture of experts (MoE) as a risk prediction model and compared with conventional mitral valve repair complexity scores.Results:A total of 2314 patients were included in this study.Mitral repair was unsuccessful in 4.2％ (98 of 2314) of patients.Patient factors such as tricuspid regurgitation pressure gradient,A3 and A3P3 lesions,left ventricular end-systolic volume,and left atrium anterior and posterior diameter are associated with mitral valve repair failure;in addition,surgeon factors,such as cumulative repair failure rate,cumulative repair volume,and surgeon seniority,are also risk factors for mitral valve repair failure.The MoE model has an AUC value of 0.79,and the prediction performance is significantly better than traditional complexity scores.Conclusions:The MoE based machine learning model can predict the risk of mitral valve repair failure well.This evaluation system can effectively assist surgeons in assessing the risk of mitral valve repair failure and in selecting suitable treatment options for patients.

Objectives:To develop a novel prediction model for mitral valve repair failure based on machine learning algorithms.Methods:Clinical and echocardiographic data were analyzed on patients,who underwent mitral valve repair in Fuwai Hospital from 2009 January 1st to 2022 December 31st.End points included immediate mitral valve repair failure (mitral replacement secondary to mitral repair failure) and recurrence regurgitation (moderate or severe mitral regurgitation before discharge).Risk factors of mitral valve repair failure were analyzed by XGBoost and shapley additive explanation (SHAP),and a machine learning model was established based on mixture of experts (MoE) as a risk prediction model and compared with conventional mitral valve repair complexity scores.Results:A total of 2314 patients were included in this study.Mitral repair was unsuccessful in 4.2％ (98 of 2314) of patients.Patient factors such as tricuspid regurgitation pressure gradient,A3 and A3P3 lesions,left ventricular end-systolic volume,and left atrium anterior and posterior diameter are associated with mitral valve repair failure;in addition,surgeon factors,such as cumulative repair failure rate,cumulative repair volume,and surgeon seniority,are also risk factors for mitral valve repair failure.The MoE model has an AUC value of 0.79,and the prediction performance is significantly better than traditional complexity scores.Conclusions:The MoE based machine learning model can predict the risk of mitral valve repair failure well.This evaluation system can effectively assist surgeons in assessing the risk of mitral valve repair failure and in selecting suitable treatment options for patients.

Objective: To investigate the correlations between single nucleotide polymorphisms (SNPs) in the D-loop of mitochondrial DNA (mtDNA) and the disease risk as well as the prognosis of diffuse large B cell lymphoma (DLBCL). Methods: Blood samples from 108 DLBCL patients treated at the Department of Hematology of the Fourth Hospital of Heibei Medical University during July, 1991 and July 2012 were collected for this study; in addition, blood samples from 159 healthy controls during the same period were also collected. DNA was extracted according to the standard protocols for PCR amplification and SNP locus genotype analyses. The risk of D-loop SNPs was investigated by case-control study. Results: The minor alleles of nucleotides 73A/G, 263A/G, 315C/C insert were associated with a decreased risk for DLBCL. The minor allele of the nucleotides 200G/Awas associated with an increased risk for DLBCL. To further evaluate the predictive function of D-loop SNPs in DLBCL patients, five SNP sites were identified by Log-Rank test that with statistically significant prediction value of DLBCL survival in a univariate analysis. In a multivariate analysis, allele 16304 was identified as an independent predictor of DLBCL prognosis. The survival time of DLBCL patients with 16304C was significantly shorter than that of patients with 16304T (RR=0.513, 95% CI=0.266-0.989, P<0.05). Conclusion: The analysis of D-loop SNPs in mtDNA can help identifying the occurrence risks and poor prognosis subtypes of DLBCL.

Objective To evaluate the effect of single-nucleotide polymorphisms at the miRNA binding site rs3660 in the 3'-untranslated region of the KRT81 gene (miR-SNPs) on the cancer risk and clinical prognosis of non-Hodgkin's lymphomas (NHL).Methods The single-nucleo-tide polymorphisms of rs3660 was genotyped with ligation detection reaction method.The association of rs3660 with NHL survival was calculated with log-rank test using Kaplan-Meier method.Multivariate survival analysis was performed using a Cox proportional hazards model.Results The rs3660 genotype distribution difference was not statistically significant between the case and control group (P =0.50).Patients carrying the rs3660 CG/CC genotype exhibited a significantly longer survival time than patients carrying the GG genotype (P =0.012).In addition,rs3660 was associated independently with the survival of NHL patients in multivariate analysis (RR=0.589,95％ CI:0.415-0.832,P =0.004).The association of this miR-SNP with NHL survival was further confirmed in the peripheral T cell lymphoma (PTCL) subtype.Conclusion Our results indicate that KRT81 rs3660 GG type is an independent prognostic marker in NHL.