Computer methods and programs in biomedicine update最新文献

Public health surveillance systems play a crucial role in detecting and responding to disease outbreaks. Visualizations of surveillance data are important for decision-making, but little attention has been paid to the usability and interaction of such systems. In this paper, we developed a set of 10 heuristics to assess the visualization and usability of public health surveillance systems. The heuristics cover aspects of perception, cognition, and interaction. The perception deals with how the system looks in the first glance and whether it has pleasant effect on the user or otherwise. Cognition deals with the question of whether enough information is provided to use the system, while usability and interaction deal with whether the system is user-friendly in terms of the tools provided for interaction and use. We recruited a panel of experts to evaluate a set of systems using our heuristics. Results showed that there was variation in the scores of the experts' assessments, indicating the importance of multiple expert evaluations. Our heuristics provide a practical and comprehensive tool for assessing the visualization and usability of public health surveillance systems, which can lead to improved decision-making and ultimately better public health outcomes. The results suggest that the heuristic based evaluation through a panel of experts can provide meaningful results and insights into the usability aspects of public health systems. The results suggest that for some systems there can be agreement in terms of evaluation while for some other systems the experts’ opinions can vary based on the weightage and importance each expert gives to a particular aspect.

Pre-training and fine-tuning have emerged as a promising paradigm across various natural language processing (NLP) tasks. The effectiveness of pretrained large language models (LLM) has witnessed further enhancement, holding potential for applications in the field of medicine, particularly in the context of Traditional Chinese Medicine (TCM). However, the application of these general models to specific domains often yields suboptimal results, primarily due to challenges like lack of domain knowledge, unique objectives, and computational efficiency. Furthermore, their effectiveness in specialized domains, such as Traditional Chinese Medicine, requires comprehensive evaluation.

To address the above issues, we propose a novel domain specific TCMDA (TCM Domain Adaptation) approach, efficient pre-training with domain-specific corpus. Specifically, we first construct a large TCM-specific corpus, TCM-Corpus-1B, by identifying domain keywords and retrieving from general corpus. Then, our TCMDA leverages the LoRA which freezes the pretrained model’s weights and uses rank decomposition matrices to efficiently train specific dense layers for pre-training and fine-tuning, efficiently aligning the model with TCM-related tasks, namely TCM-GPT-7B. We further conducted extensive experiments on two TCM tasks, including TCM examination and TCM diagnosis. TCM-GPT-7B archived the best performance across both datasets, outperforming other models by relative increments of 17% and 12% in accuracy, respectively. To the best of our knowledge, our study represents the pioneering validation of domain adaptation of a large language model with 7 billion parameters in TCM domain. We will release both TCM-Corpus-1B and TCM-GPT-7B model once accepted to facilitate interdisciplinary development in TCM and NLP, serving as the foundation for further study.

Background

There is a need to explore models of eHealth literacy that serve as mediators in the relationship between health status and well-being from multidimensional perspectives among the elderly population.

Aims

To examine series models in which eHealth literacy dimensions, including awareness of sources, recognizing quality and meaning, understanding information, perceived efficiency, and validating information, serve as mediators between health status and factors related to well-being, such as financial, physical, eudaimonic, and hedonic well-being.

Methods

This cross-sectional study included 437 Israeli seniors aged 65 or above and employed the eHEALS-E scale with six dimensions to assess eHealth literacy in the first section of the questionnaire. The second section utilized a well-being scale with five categories to measure financial, physical, social, eudaimonic, and hedonic well-being. Ethical approval was obtained from the Institutional Review Board (IRB).

Results

eHealth literacy dimensions such as understanding information, awareness of sources, validating information, and recognizing quality play a crucial role in mediating the relationship between health status and different aspects of financial, social, eudaimonic and hedonic well-being.

Conclusions

Interventions and educational programs are needed to focus on enhancing eHealth literacy, specifically targeting the dimensions of understanding information, awareness of sources, validating information, and recognizing quality. By improving these eHealth literacy dimensions, individuals' financial well-being, social well-being, and overall eudaimonic and hedonic well-being can be positively influenced.

Autologous ear reconstruction is a surgical procedure performed in the case of defects of the outer ear in which the malformed anatomy is reconstructed with autologous cartilage tissue and often involves the use of surgical guides modelled from a digital reconstruction of the ear anatomy. To obtain such three-dimensional anatomy, traditional imaging methods, which are expensive and invasive, can be replaced by professional 3D scanners or low-cost commercial devices. In this context, this paper focuses on the evaluation of two devices for the acquisition of the outer ear, the Intel® RealSense D405™ (stereo camera) and the TrueDepth camera of the iPhone® 13 (structured light camera), proposing a comparison based on four parameters: accuracy, precision, deviation range and point-to-point distance, in order to assess their usability in the medical field, and in particular in the context of autologous ear reconstruction. The results show that, despite significantly different handling of the raw data, the performance of the two devices is comparable: average accuracy is 0.76 mm for the D405 and 0.95 mm for the iPhone 13, average precision is 0.071 mm for the D405 and 0.065 mm for the iPhone 13, average range of deviation is 3.12 mm for the D405 and 3.64 mm for the iPhone 13.

Background

Clinical prediction is integral to modern healthcare, leveraging current and historical medical data to forecast health outcomes. The integration of Artificial Intelligence (AI) in this field significantly enhances diagnostic accuracy, treatment planning, disease prevention, and personalised care leading to better patient outcomes and healthcare efficiency.

Methods

This systematic review implemented a structured four-step methodology, including an extensive literature search in academic databases (PubMed, Embase, Google Scholar), applying specific inclusion and exclusion criteria, data extraction focusing on AI techniques and their applications in clinical prediction, and a thorough analysis of the collected information to understand AI's roles in enhancing clinical prediction.

Results

Through the analysis of 74 experimental studies, eight key domains, where AI significantly enhances clinical prediction, were identified: (1) Diagnosis and early detection of disease; (2) Prognosis of disease course and outcomes; (3) Risk assessment of future disease; (4) Treatment response for personalised medicine; (5) Disease progression; (6) Readmission risks; (7) Complication risks; and (8) Mortality prediction. Oncology and radiology come on top of the specialties benefiting from AI in clinical prediction.

Discussion

The review highlights AI's transformative impact across various clinical prediction domains, including its role in revolutionising diagnostics, improving prognosis accuracy, aiding in personalised medicine, and enhancing patient safety. AI-driven tools contribute significantly to the efficiency and effectiveness of healthcare delivery.

Conclusion and recommendations

AI's integration in clinical prediction marks a substantial advancement in healthcare. Recommendations include enhancing data quality and accessibility, promoting interdisciplinary collaboration, focusing on ethical AI practices, investing in AI education, expanding clinical trials, developing regulatory oversight, involving patients in the AI integration process, and continuous monitoring and improvement of AI systems.

Background

Chronic kidney disease (CKD) poses a major global public health burden, with over 700 million affected. Early identification of those in whom the disease is likely to progress enables timely therapeutic interventions to delay advancement to kidney failure.

Methods

This study developed explainable machine learning models leveraging pathology data to accurately predict CKD trajectory, targeting improved prognostic capability even in early stages using limited datasets. Key variables used in this study include age, gender, most recent estimated glomerular filtration rate (eGFR), mean eGFR, and eGFR slope over time prior to the incidence of kidney failure. Supervised classification modelling techniques included decision tree and random forest algorithms selected for interpretability. Internal validation on an Australian tertiary centre cohort (n = 706; 353 with kidney failure and 353 without) achieved exceptional predictive accuracy. To address the inherent class imbalance, centroid-cluster-based under-sampling was applied to the Australian dataset. For external validation, the model was applied to a dataset (n = 597 adults) sourced from a Japanese CKD registry. Transfer learning was subsequently employed by fine-tuning machine learning models on 15 % of the external dataset (n = 89) before evaluating the remaining 508 patients.

Results

Internal validation achieved exceptional predictive accuracy, with the area under the receiver operating characteristic curve (ROC-AUC) reaching 0.94 and 0.98 on the binary task of predicting kidney failure for decision tree and random forest, respectively. External validation demonstrated performant results with an ROC-AUC of 0.88 for the decision tree and 0.93 for the random forest model. Decision tree model analysis revealed the most recent eGFR and eGFR slope as the most informative variables for prediction in the Japanese cohort.

Conclusion

The research highlights the utility of deploying explainable machine learning techniques to forecast CKD trajectory even in the early stages utilising limited real-world datasets.

Pneumonia remains a significant global health concern, claiming millions of lives annually. This study introduces a novel approach by developing and analyzing a Caputo fractional order pneumonia infection model that incorporates pneumonia asymptomatic carriers. Through a qualitative lens, we establish the existence and uniqueness of model solutions by applying the well-known Picard–Lindelöf criteria. Employing a next-generation approach, we compute the model's basic reproduction number, determine equilibrium points, and probe their stabilities. The main objective of this study is to investigate the transmission dynamics of pneumonia infection with a focus on asymptomatic carriers using fractional order modeling. Our findings reveal innovative outcomes as we showcase numerical simulations, providing a practical verification of the qualitative results. Notably, we explore the fractional order model solutions in-depth, examining the influence of specific model parameters and fractional orders on the dynamics of pneumonia disease transmission. The significant contributions of this study lie in advancing the theoretical foundation of infectious disease modeling, particularly in the context of pneumonia. Through rigorous analysis and numerical simulations, we provide valuable insights into the behavior of the proposed fractional order model. These findings hold practical implications for understanding and managing pneumonia transmission in real-world scenarios. Our study serves as a vital resource for researchers, policymakers, and healthcare practitioners involved in combating and preventing the spread of pneumonia, ultimately contributing to global efforts in public health.

A study of a community of people with disabilities in a virtual world sheds new light on an important issue of health literacy that has to date remained underreported in the current body of research. Participants revealed a community of individuals who are adults role-playing via child avatars as a coping and recovery mechanism for childhood trauma. One case follows the experience of a woman who role plays an adopted child of a caring adult while another attempts to recreate different ages of herself to unpack past trauma and find therapeutic healing. This phenomenon, as well as both its risks and opportunities, are examined with important considerations for the future of digital mental health support for people who have experienced abuse as children. Researchers, policy makers, and mental health professionals are encouraged to consider the role of social virtual worlds in the future of telemedicine for PTSD therapy.