Exploring the Use of AI-Assisted Video Monitoring to Predict Accidental Events in ICU Patients
试验速览
- 阶段
- 不适用
- 状态
- 尚未招募
- 发起方
- 入组人数
- 300
- 主要终点
- Accuracy of AI-Assisted Video Monitoring in Predicting Accidental Events
研究概览
简要总结
This study aims to improve the safety and care of patients in the Intensive Care Unit (ICU) by using artificial intelligence (AI) to analyze video monitoring. ICU patients often face serious risks such as delirium, accidental removal of breathing tubes or lines, and sleep problems. These events can lead to medical emergencies, longer ICU stays, higher costs, and worse outcomes.
To address these challenges, we will place a small video camera above each ICU bed. The camera will record patient movements, body activity, and sleep patterns. At the same time, routine medical monitors will record heart rate, blood oxygen levels, and other vital signs. Noise levels in the room will also be measured. All these data help us understand the patient's behavior and condition more accurately.
The video recording does not involve extra treatment or additional procedures. All data are collected passively and safely. Patient privacy is strictly protected: the system will blur faces or replace them with digital avatars, and any information that could identify the patient or the environment will be masked. All videos are stored securely inside the hospital and are processed only after privacy protection.
Using these recordings, an AI model will be trained to recognize early warning signs of dangerous situations. For example, the system may detect early movements that suggest the patient is becoming agitated, confused, or trying to remove medical tubes. It may also identify severe sleep disturbance that may lead to delirium. If the AI can recognize these early changes, medical staff can intervene sooner and prevent harm.
About 300 patients from Fudan University Zhongshan Hospital will participate. Participation is voluntary. Patients or families will sign an informed consent form before being enrolled. The study has three stages:
Screening - understanding the study and signing consent. Data collection - video and medical monitor data are collected during the ICU stay.
Follow-up - telephone or in-person follow-up at 1 month and 6 months after discharge to evaluate recovery, sleep, mental status, and overall safety.
There are no direct medical risks from participating in this study because it only collects behavioral and monitoring data. The cameras do not interfere with treatment. Privacy and data security are the main considerations, and all measures strictly follow national laws and hospital regulations.
Participants may benefit from earlier identification of dangerous situations, which may help prevent accidental tube removal, severe agitation, or other emergencies. Even if no direct benefit occurs, the information collected may help improve future ICU care by enabling safer and more accurate monitoring systems.
Taking part in the study will not affect the patient's medical care. Patients may withdraw at any time without any consequences or loss of benefits.
This study hopes to build a reliable AI tool that can assist nurses and doctors in recognizing early signs of trouble, improving safety, and enhancing the quality of care for ICU patients.
详细描述
This study investigates whether artificial intelligence (AI)-assisted video monitoring can identify early behavioral changes that precede accidental or harmful events in Intensive Care Unit (ICU) patients. ICU patients are vulnerable to a series of sudden and potentially dangerous events-such as agitation, delirium, accidental device removal, and significant sleep disruption-many of which develop gradually and are difficult to detect solely from routine physiological monitoring. This project aims to determine whether AI analysis of continuous bedside video recordings, combined with noise-level information and vital-sign data already collected during standard ICU care, can provide clinicians with timely warnings before these events occur.
Rationale Traditional ICU monitoring systems focus on physiological parameters such as heart rate, blood pressure, and oxygen saturation. While essential, these measurements do not fully represent patient behavior. Many high-risk events are preceded by subtle motor patterns or behavioral cues-for example, repeated reaching toward tubes, rising restlessness, or disturbed sleep cycles. Such cues are often intermittent, brief, or masked by sedation or other treatments, making them difficult for staff to detect in busy clinical environments.
Computer vision and AI technologies offer an opportunity to objectively observe and interpret patient movements and behavioral trends continuously, without adding clinical workload. By integrating video information with physiologic data and environmental noise levels, the AI system may identify patterns that indicate emerging delirium, increased agitation, or imminent attempts to remove medical devices. Early identification may support timely preventive interventions and reduce the rates of adverse events.
Study Overview The study will prospectively enroll ICU patients who consent to video monitoring and data use. A small camera will be installed above each bed to continuously capture patient movement and posture. The camera view is restricted to the patient zone, excluding unnecessary areas such as the nursing station. All recordings follow strict privacy-protection procedures, including automated face masking, background blurring, and removal of identifying information from objects in the frame.
Environmental noise is recorded through a decibel meter, and routine vital-sign data are synchronized with the video timeline. These combined multimodal data will serve as input for AI model development.
研究设计
- 研究类型
- Observational
- 观察模型
- Other
- 时间视角
- Prospective
入排标准
- 性别
- All
- 接受健康志愿者
- 否
入选标准
- •Adult or pediatric patients admitted to the Intensive Care Unit (ICU).
- •Patient or legally authorized representative is capable of understanding the study information and providing informed consent.
- •Patient is expected to remain in the ICU long enough to allow video and physiologic data collection.
- •Agreement to participate and allow video monitoring during the ICU stay.
排除标准
- •Refusal to participate from the patient or legally authorized representative.
- •Patients for whom continuous video monitoring is medically inappropriate or not feasible (e.g., isolation conditions preventing camera installation).
- •Patients whose condition or legal status requires special restrictions on video recording (e.g., certain forensic or custodial cases).
- •Any situation judged by the clinical team to place the patient at increased privacy or safety risk by participation.
- •Withdrawal of consent at any point during the study.
结局指标
主要结局
Accuracy of AI-Assisted Video Monitoring in Predicting Accidental Events
时间窗: From ICU admission until ICU discharge (up to 30 days)
Accuracy will be measured as the proportion of correct AI predictions compared with clinically confirmed accidental events, including delirium-related agitation, unplanned device removal, and significant sleep disruption. Accuracy includes both true positive and true negative predictions. The AI model's output will be compared with event labels validated through clinical documentation and human annotation of video data. This outcome assesses whether the AI system can reliably identify high-risk behavioral patterns before the occurrence of actual adverse events.
次要结局
- Sensitivity of AI Predictions for Accidental Events(From ICU admission to ICU discharge (up to 30 days))
- Specificity of AI Predictions(From ICU admission to ICU discharge (up to 30 days))
- Lead Time Between AI Alert and Actual Event(From ICU admission to ICU discharge (up to 30 days))