Journal of Medical Systems最新文献

Inventory management for emergency carts is one of the routine tasks in hospitals. It is highly desirable to simplify the workflow of the inventory task since healthcare staffs always work under high pressure and heavy workload. In this study, we exploit computer vision technology to develop an automated inventory management system for emergency carts. We have conducted a user study evaluate our proposed system. The subjects are 37 nurses from the internal medicine department and the pediatrics department of Chia-Yi Christian Hospital. The time spending on the inventory task before and after using our proposed system are compared. We also evaluate several state-of-the-art object detection algorithms (YOLOv4, YOLOv5, YOLOv7, and YOLOv5-OBB) for the task of automatic drug counting. We collect 500 images to train these object detection models and evaluate their accuracy. After using the proposed system, the time spent on the inventory task is reduced from 9.59 min to 5.32 min in average. The user study also indicates that the average workload score of the nurses is reduced from 3.81 to 2.70. For the evaluation of automatic drug counting, our proposed hybrid approach, which combines YOLOv5-OBB and YOLOv7, yields the highest accuracy. The resulting over detection rate and missed detection rate are 1.29% and 3.27%, respectively. In this study, we demonstrate that smartphone-based inventory management system using computer vision technology, specifically YOLO series object detectors, can effectively streamline inventory workflow and reduce the workload for healthcare personnel.

The rapid evolution of large language models (LLMs) in the medical field, particularly in automating medical tasks and supporting diagnosis and treatment, has shown promising potential. However, their accuracy, comprehensiveness, and safety in managing complex cardiovascular diseases have not been systematically assessed. This study aims to evaluate and compare the diagnostic, therapeutic, and safety performance of two large language models-ChatGPT-4o and Kimi-in managing complex cardiovascular diseases, and to explore their potential for future clinical application. A total of 200 complex cardiovascular cases published in JACC: Case Reports between January 2020 and August 2024 were included. All cases were standardized and de-identified before being input into ChatGPT-4o and Kimi using identical prompts. Each model independently generated diagnostic, treatment, and long-term management plans. Three cardiovascular specialists independently evaluated the outputs in a blinded manner, scoring accuracy and comprehensiveness using Likert scales. Safety was assessed using a risk matrix analysis. Additionally, 50 cases were randomly selected for triangulation to compare model-generated recommendations with clinical guidelines. Statistical analysis was performed using the Wilcoxon signed-rank test with Benjamini-Hochberg correction for multiple comparisons. In preliminary diagnostic accuracy, the two models performed similarly (P = 0.663, r = 0.044), but ChatGPT-4o showed superior comprehensiveness (P < 0.001, r = 0.484). For treatment recommendations, ChatGPT-4o outperformed Kimi in both accuracy (P = 0.004, r = 0.321) and comprehensiveness (P < 0.001, r = 0.644). In long-term management, ChatGPT-4o demonstrated significant advantages in accuracy (P < 0.001, r = 0.717) and comprehensiveness (P < 0.001, r = 0.690). Safety assessment showed a lower proportion of high-risk outputs with ChatGPT-4o (1.5%) compared to Kimi (4.5%). LLMs, particularly ChatGPT-4o, exhibit significant promise in the diagnosis and treatment of complex cardiovascular diseases, showing superior accuracy, comprehensiveness, and safety compared to Kimi. Despite their high accuracy and safety, LLMs still require clinician oversight, especially in the formulation of personalized treatment plans and complex decision-making scenarios, to ensure their reliable integration into clinical practice.

Background: Nuclear imaging is the cornerstone of clinical practice across many disciplines. Few innovations in imaging have addressed occupational health of radiographers exposed to radiation in their daily work. In this proof-of-concept study, we hypothesized that the use of autonomous robots in a high-volume Myocardial Perfusion Imaging (MPI) clinic to help reduce radiation risk for radiographers.

Methods: After initial assessment of the radiographer's workflow, an autonomous robot was set up to navigate and deliver radiotracer doses between the radio pharmacy and the injection room. 8 radiographers were briefed on robot usage and allowed to use the robot for their daily operations. Radiation exposure was measured as part of the regular bi-monthly Thermoluminescent Dosimeter (TLD) dose tracking. Radiation exposure before and after robot implementation were compared to assess whether the implementation of the autonomous robot significantly reduced radiation exposure for the radiographers.

Results: We observed a significant reduction in mean bi-monthly radiation exposure following the implementation of the autonomous robot. The mean radiation dose decreased from 0.67 mSv (95% CI: 0.60-0.74) pre-implementation to 0.49 mSv 95% (CI: 0.44-0.54) post-implementation, corresponding to a relative reduction of approximately 27%. A Bayesian independent t-test revealed strong evidence for this reduction, with Bayes Factors (BF10) of 17.04 for skin dose and 17.76 for whole-body dose, supporting the hypothesis that the autonomous robot effectively reduced radiation exposure among radiographers.

Conclusion: In this study, we provided a proof-of-concept on the use of autonomous robots in a MPI clinic as an additional tool to help radiographers manage radiation risk in their work. The innovations in technologies could expand the strategies available for managing occupational radiation risks in alignment with as low as reasonably achievable (ALARA) principles. Future work on scalability across diverse clinical and operational contexts would be next steps.

Background: DeepSeek-R1, an open-source reasoning large language model (LLM) clinically deployed in Chinese hospitals, still lacks validation in ophthalmology.

Aims: To compare DeepSeek-R1 against OpenAI's o1 and upgraded o3 models in diagnostic accuracy and reasoning capability across diverse ophthalmic conditions.

Methods: We evaluated 98 standardized case vignettes covering13 ophthalmic sub-specialties, each supplied with an expert-validated diagnostic hierarchy, differential list, and reasoning chain. Model performance was assessed with a diagnosis matrix focused on final-diagnosis (FDx) accuracy; incorrect outputs were resubmitted with key diagnostic clues (reasoning-augmented, RA prompt) to test self-correction. Reasoning capacity was quantified by the number/score of diagnostic clues retrieved per case across 13 predefined domains.

Results: DeepSeek-R1 achieved an 87.8% FDx accuracy, comparable to o3 (91.8%, P = .34) and higher than o1 (58.2%, P < .001). Similar trends were observed for others accuracy (global P < .001). Agreement was moderate-high between R1 and o3 (κ = 0.42-1.00), but slight with o1 (κ = 0.12-0.32). R1 and o3 identified more diagnostic clues than o1 (median count = 4 vs. 3, median score = 100 vs. 80; P < .001). RA prompts corrected 50.0%, 62.5% and 41.5% of FDx errors for R1, o3, and o1, raising FDx accuracy to 93.9%, 96.9%, and 80.6% respectively.

Conclusions: DeepSeek-R1 matched o3 and outperformed o1 in diagnostic accuracy and reasoning, retrieving nearly all expert-defined clues. Its open-source nature, low cost and strong performance support its use as a practical aid for ophthalmic decision-making.

Accurate polyp detection is essential for the early diagnosis and effective treatment of colorectal cancer (CRC). However, colonoscopy videos in real-world clinical settings present significant challenges, often causing existing algorithms to fail. Compared to single images, videos contain richer temporal and contextual information, making them valuable for developing deep-learning-based detection systems. To address these challenges, we propose an end-to-end Two-Stream Polyp Detection Transformer (TS-PDTR) network. First, our framework uses a two-stream feature extraction network to capture both spatial and temporal features from the RGB frames and optical flow. Then, the proposed Detail-Aware Convolution (DAConv) module enhances fine-grained contextual information in low-level features. Following this, the Detail-Guided Attention (DGA) module generates channel-specific Spatial Attention Maps (SAMs) to refine deep feature maps, improving the model's sensitivity to small and camouflaged polyps. Finally, a Flow Fusion Encoder (FFE) module combines temporal cues from optical flow to increase robustness against poor single-frame image quality. Experiments on three benchmark video colonoscopy datasets show that TS-PDTR consistently outperforms previous state-of-the-art image- and video-based polyp detection methods. Notably, our model achieves a mean Average Precision (mAP) of 33.2 on the most challenging LDPolypVideo dataset. It also improves the mAP to 64.0 and 55.6 on the SUN Colonoscopy Video Database and CVC-VideoClinicDB, respectively. In summary, TS-PDTR is a promising video-based polyp detection method with strong potential for further development and real-world clinical application.

Artificial intelligence (AI), specifically large language models (LLM), have gained significant popularity over the last decade with increased performance and expanding applications. AI could improve the quality of patient care in medicine but hidden biases introduced during training could be harmful. This work utilizes GPT-4o-mini to generate patient communications based on systematically generated, synthetic patient data that would be commonly available in a patient's medical record. To evaluate the AI generated communications for disparities, GPT-4o-mini was used to score the generated communications on empathy, encouragement, accuracy, clarity, professionalism, and respect. Disparities in scores associated with specific components of a patient's history were used to detect potential biases. A patient's sex and religious preference were found to have a statistically significant impact on scores. However, further work is needed to evaluate a wider collection of LLMs utilizing more specific and human validated scoring criteria. Overall, this work proposes a novel method of evaluating bias in LLMs by creating synthetic patient histories to formulate AI generated communications and score them with opportunities for further investigation.

Healthcare delivery systems face mounting administrative complexity that contributes to clinician burnout, medical errors, and reduced access to care for patients. This editorial explores how automation and artificial intelligence (AI) can address key operational inefficiencies-specifically in prior authorization, quality metric reporting, and clinical documentation-by leveraging informatics-driven solutions. We examine the current landscape, quantify the impact of administrative burden, and propose informatics strategies to realign healthcare delivery around patient-centered, efficient care.

Normative ranges for vital signs under general anesthesia are well established for healthy pediatric patients, but the influence of The American Society of Anesthesiologists Physical Status (ASA-PS) classification on these normative ranges remains unexplored. The purpose of this study is to develop age-based normative ranges for heart rate (HR) and blood pressure (BP) in patients undergoing general anesthesia for noncardiac surgery in our institution and assess differences by ASA-PS classification. This is a retrospective observational single-center study. We reviewed all anesthetic records from the Hospital for Sick Children, Canada between March 1st and December 31st, 2023. We extracted physiological data from our in-house high-resolution physiological data repository (AtriumDB) to develop normative ranges for physiological parameters and compared them according to ASA-PS classification. We developed age-based normative ranges for BP and HR. We found significant differences between ASA-PS groups, most notably between ASA-PS 1 and 5. We found a statistically significant difference between ASA-PS 1-2 and 3-5 across all physiological parameters. This study validates existing pediatric anesthesia reference ranges while demonstrating the feasibility of incorporating patients across the spectrum of ASA-PS. Further multicenter studies are needed to generalize these findings.

This paper presents a trust-aware architecture for personalized digital health that combines user modeling, symbolic reasoning, and adaptive trust mechanisms. The proposed system uses Blueprint Personas to capture detailed patient profiles, including clinical, behavioral, and emotional traits. These profiles guide an intelligent agent that interacts with patients and healthcare professionals to provide context-sensitive support. Personalization is achieved through an ontology-based reasoning layer that interprets user needs and integrates real-time data from electronic health records, wearable devices, and environmental sources. To promote transparency and foster long-term user engagement, the system includes a formal trust modeling component based on a Reference Ontology of Trust (ROT), allowing the system to flexibly tailor communication strategies in response to user feedback and evolving trust levels. A simulated scenario involving a patient with chronic obstructive pulmonary disease demonstrates how the system delivers proactive and personalized healthcare interventions, such as medication reminders and air quality alerts. While the architecture is modular and designed for scalability, it has not yet been deployed in real-world clinical settings. Empirical validation and integration with clinical platforms remain part of future work. Nevertheless, this ongoing work contributes to the development of explainable and ethically aligned AI systems that enhance autonomy, accessibility, and trust in digital health environments through explainable reasoning.