Medical Education
Conditions
Keywords
Explainable Artificial Intelligence, Medical Student Education, Neuroimaging Interpretation, Mild Cognitive Impairment, SHAP, LIME
Brief summary
This study evaluates whether Explainable Artificial Intelligence (XAI) explanations integrated into medical training improve AI literacy, reduce cognitive workload, and enhance learner trust compared to traditional lecture methods. Third-year medical students participated in a randomized controlled trial assessing the CerViD-MultiModal diagnostic framework during a neuroimaging diagnostic module focused on fornix atrophy in early and late mild cognitive impairment
Detailed description
The study utilized a two-phase sequential explanatory design with mixed methodologies. In Phase 1 (Technical Development), the CerViD-MultiModal model was developed and validated using neuroimaging data from 100 Alzheimer's Disease Neuroimaging Initiative (ADNI) subjects to classify early vs. late mild cognitive impairment via fornix morphometry features. In Phase 2 (Educational Intervention), a randomized controlled trial was conducted with 120 third-year medical students enrolled in the clinical neuroscience rotation at the University of Liberia. Participants were randomized into two equal groups (n=60 per group): Control Group: Completed a 45-minute traditional lecture module using static text and bar charts. XAI-Enhanced Group: Completed an interactive 45-minute module featuring SHAP summary charts, LIME patient-specific explanations, and interactive force graphs. Post-intervention electronic assessments evaluated four primary outcomes: AI Literacy Score (0-100 scale), System Usability Scale (SUS, 0-100 scale), perceived cognitive workload using the NASA Task Load Index (NASA-TLX, 0-100 scale), and Confidence in AI Interpretation (1-5 scale)
Interventions
OTHERXAI-Enhanced Interactive Module (CerViD-MultiModal)
Standard educational instruction delivered via traditional slides and static charts explaining neuroimaging AI outputs.
OTHERTraditional AI Lecture Module
Standard educational instruction delivered via traditional slides and static charts explaining neuroimaging AI outputs.
Sponsors
University of Liberia
National Institute on Aging (NIA)
Study design
Allocation
RANDOMIZED
Intervention model
PARALLEL
Primary purpose
HEALTH_SERVICES_RESEARCH
Masking
NONE
Intervention model description
Participants were randomized into two parallel groups (Control vs. XAI-Enhanced) for a 45-minute educational intervention.
Eligibility
Sex/Gender
ALL
Age
18 Years to No maximum
Healthy volunteers
Yes
Inclusion criteria
1. Enrolled as a third-year medical student in the clinical neuroscience rotation at the University of Liberia. 2. Willing and able to complete the 45-minute educational module and post-intervention evaluations. 3. Provided informed consent to participate in the study.
Exclusion criteria
1. Prior formal coursework, professional training, or specialized technical degree in artificial intelligence, machine learning, or computer science. 2. Inability to complete the post-intervention assessment.
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| AI Literacy Score | Immediately post-intervention (Day 1) | Continuous score (0-100 scale) measuring conceptual knowledge, practical application, ethical awareness, and critical evaluation of AI systems in medicine |
| System Usability Scale (SUS) Score | Immediately post-intervention (Day 1) | Standardized 10-item scale assessing user perception of system usability, converted to a 0-100 overall score |
Countries
Liberia
Outcome results
None listed