A Within-Subject Eye-Tracking Study Examining How the Timing of AI Decision Support Influences Visual Search Behaviour, Diagnostic Accuracy, and Trust During Chest X-ray Interpretation
试验速览
- 阶段
- 不适用
- 状态
- 招募中
- 发起方
- 入组人数
- 24
- 试验地点
- 1
- 主要终点
- Change in Chest X-ray Case Interpretation Time
研究概览
简要总结
Chest X-rays are commonly used to help diagnose and manage chest conditions. Artificial intelligence (AI) tools are increasingly being used to support chest X-ray interpretation. However, it is not yet clear whether the timing of AI information affects how clinicians review images, make decisions, and use AI support.
This study will look at whether showing AI information before or after a clinician first reviews a chest X-ray changes how they look at the image, how long they take, their interpretation decisions, their confidence, and their trust in AI support.
Healthcare professional participants will complete two chest X-ray interpretation sessions in a controlled NHS research setting. During each session, participants will review de-identified chest X-ray images while wearing eye-tracking equipment. Eye-tracking will record where a participant looks on the image and how long they spend looking at different areas.
In one session, AI information will be shown before the participant reviews the chest X-ray. In the other session, AI information will be shown after the participant has first reviewed the chest X-ray. The order of these two sessions will be balanced across participants.
The study uses de-identified chest X-ray images from existing examinations. It does not involve patients directly, does not change clinical care, and no clinical decisions will be made from the study readings. Participants will also complete a short questionnaire about their experience of using AI support. A separate anonymous survey will collect wider views from clinicians, patients, members of the public, and healthcare staff about the use of AI in chest X-ray interpretation.
详细描述
CREAITED is a behavioural diagnostic reader study examining how the timing of artificial intelligence (AI) decision-support information influences chest X-ray interpretation by healthcare professionals. The study focuses on whether showing AI information before or after a participant first reviews a chest X-ray affects visual search behaviour, interpretation time, clinical decision-making, confidence, and trust in AI support.
Chest X-rays are widely used in clinical practice, but interpretation can be challenging because findings may be subtle and normal anatomy can overlap with abnormalities. AI systems are increasingly used to support image interpretation by highlighting suspected abnormalities. However, less is known about how clinicians use AI information during reporting tasks, including whether seeing AI output early changes attention, reliance, confidence, or interpretation strategy. Current clinical guidance generally supports independent image review before consulting decision-support tools, but direct evidence on AI timing in chest X-ray interpretation remains limited. This study is designed to provide controlled evidence on that question.
The main reader study is a single-site study at University Hospitals of Leicester NHS Trust. It uses a within-subject, counterbalanced, multi-reader, multi-case design. The exposure of interest is the timing of AI decision-support presentation during chest X-ray interpretation. The AI information is used only within the research task and is not used to guide real patient care.
Healthcare professional participants will complete two chest X-ray interpretation sessions in a controlled NHS research setting. Participants will include clinicians and healthcare professionals who interpret, review, check, or act on chest X-ray findings as part of their current or recent professional role. The study is not designed to assess individual competence. Individual participant results will not be used for employment, appraisal, training progression, formal assessment, or performance management.
Each participant will complete both AI timing conditions. In one session, AI information will be shown before the participant reviews the chest X-ray. In the other session, the participant will first review the chest X-ray without AI information, and AI information will then be shown afterwards. The order of these two-timing conditions will be balanced across participants using a pre-generated allocation schedule. The two sessions will be separated by at least four weeks to reduce recall and learning effects.
研究设计
- 研究类型
- Interventional
- 分配方式
- Randomized
- 干预模型
- Crossover
- 主要目的
- Diagnostic
- 盲法
- None
盲法说明
No masking is used. It is clear to participants and study staff whether the chest X-ray or the AI information is shown first during each session.
入排标准
- 年龄范围
- 18 Years 至 —（Adult, Older Adult）
- 性别
- All
- 接受健康志愿者
- 是
入选标准
- •Main reader study:
- •Healthcare professionals aged 18 years or over
- •Registered to practise in the UK
- •Employed by the NHS or another healthcare service operating in the UK
- •Current or recent, within the last 3 years, clinical experience involving chest X-ray interpretation, review, or use in clinical practice
- •Able to attend two onsite study sessions at University Hospitals of Leicester NHS Trust at mutually agreed times
- •Able and willing to provide written informed consent
- •Compatible with the eye-tracking equipment
- •Supplementary survey:
- •Adults aged 18 years or over
- •Live in the UK or have used the NHS or another UK healthcare service within the last 5 years
- •Able to provide informed electronic consent
- •Belong to one of the following respondent groups: healthcare professionals or healthcare staff, patients or carers, or members of the public
- •Healthcare professional respondents may include adults involved in requesting, interpreting, checking, or acting on chest X-ray findings in clinical practice
排除标准
- •Main reader study:
- •Inability to attend both onsite study sessions at University Hospitals of Leicester NHS Trust
- •Eye-tracking incompatibility, such as visual, neurological, or physical conditions preventing adequate gaze tracking or participant comfort
- •Direct involvement in selection, adjudication, or preparation of the chest X-rays used in the study
- •Conflicts of interest, including direct involvement in development of the AI system under evaluation
- •Prior participation in a closely related AI chest X-ray study where overlap in image sets or study procedures may compromise validity, assessed on a case-by-case basis
- •Supplementary survey:
- •Aged under 18 years
- •Does not live in the UK and has not used the NHS or another UK healthcare service within the last 5 years
- •Unable to provide informed electronic consent
- •Does not meet one of the eligible respondent groups for the survey
研究组 & 干预措施
Original CXR First Sequence
Participants complete the original chest X-ray first timing condition in Session 1, followed by the AI output first timing condition in Session 2. Sessions are separated by at least four weeks.
干预措施: Original CXR First Timing (Behavioral)
Original CXR First Sequence
Participants complete the original chest X-ray first timing condition in Session 1, followed by the AI output first timing condition in Session 2. Sessions are separated by at least four weeks.
干预措施: AI Output First Timing (Behavioral)
AI Output First Sequence
Participants complete the AI output first timing condition in Session 1, followed by the original chest X-ray first timing condition in Session 2. Sessions are separated by at least four weeks.
干预措施: Original CXR First Timing (Behavioral)
AI Output First Sequence
Participants complete the AI output first timing condition in Session 1, followed by the original chest X-ray first timing condition in Session 2. Sessions are separated by at least four weeks.
干预措施: AI Output First Timing (Behavioral)
结局指标
主要结局
Change in Chest X-ray Case Interpretation Time
时间窗: Session 1 at enrolment and Session 2 at least 4 weeks later, anticipated average 5 weeks
Case interpretation time is the duration, in seconds, from initial case display to final structured report submission. Change in case interpretation time will be measured between Session 1 and Session 2, displaying either original chest X-ray first or AI output first.
Change in Chest X-ray Case Diagnostic Accuracy
时间窗: Session 1 at enrolment and Session 2 at least 4 weeks later, anticipated average 5 weeks
Diagnostic accuracy will be assessed using structured report outcomes compared with clinician-verified chest X-ray ground truth. Each participant's final structured interpretation for each chest X-ray case will be compared with the verified study ground truth for that case. Change in diagnostic accuracy will be measured between Session 1 and Session 2, displaying either original chest X-ray first or AI output first.
Change in Chest X-ray Case Visual Search Behaviour
时间窗: Session 1 at enrolment and Session 2 at least 4 weeks later, anticipated average 5 weeks
Visual search behaviour will be assessed using predefined eye-tracking metrics that capture gaze location and search behaviour during chest X-ray interpretation. Metrics include time to first fixation, hit time, dwell time within abnormal case areas of interest, fixation count, mean fixation duration, revisit count, and fixation dispersion. Change in visual search behaviour will be measured between Session 1 and Session 2, displaying either original chest X-ray first or AI output first.
Trust in AI Support After Reader-Study Completion
时间窗: Immediately after Session 2, anticipated average 5 weeks after enrolment
Trust in AI support will be assessed using structured questionnaire ratings after participants have completed both Session 1 and Session 2. This measure captures participants' overall perceived usefulness, reliance on AI support, and willingness to revise decisions after reviewing AI information.
次要结局
- Change in Chest X-ray Case Clinical Management Recommendations(Session 1 at enrolment and Session 2 at least 4 weeks later, anticipated average 5 weeks)
- Change in Chet X-ray Case Decision Confidence(Session 1 at enrolment and Session 2 at least 4 weeks later, anticipated average 5 weeks)
- Overall Reflective Confidence After Reader-Study Completion(Immediately after Session 2, anticipated average 5 weeks after enrolment)
- Participant Perceptions of AI Support After Reader-Study Completion(Immediately after Session 2, anticipated average 5 weeks after enrolment)
- Broader Attitudes Toward AI in Chest X-ray Interpretation(2 months)