AI-Enhanced Virtual Simulation as an Adjunct to Skills Laboratory and Clinical Training in Radiography Education: A Commentary from Somalia and Other Resource-Constrained Settin

Abstract

Radiography education requires repeated practice in positioning, exposure selection, image evaluation, communication, and radiation protection. Virtual simulation can provide safe rehearsal before clinical exposure, but virtual reality, conventional digital simulation, and artificial intelligence (AI)-enhanced simulation are distinct technologies and their evidence should not be conflated. This commentary examines how AI-enhanced virtual simulation could complement, rather than replace, physical skills laboratories and supervised clinical training in Somalia and other resource-constrained educational settings. Evidence from radiography education supports virtual simulation for positioning, exposure-factor selection, image appraisal, and preparation for clinical practice, while emerging AI-specific studies suggest potential roles for adaptive case generation, automated feedback, and assessment support. However, AI-generated educational content may be inaccurate, biased, overly recall-focused, or clinically unrealistic; expert validation and psychometric evaluation are therefore required before curricular or high-stakes assessment use. Somalia has emerging undergraduate radiography programs and heterogeneous imaging capacity, but national data describing teaching-laboratory infrastructure, digital radiography vs computed radiography availability, faculty capacity, and clinical placement resources remain limited. Accordingly, country-level scarcity claims should be treated as hypotheses for local needs assessment rather than established facts. We propose a three-layer model: AI-enhanced preparation, focused physical laboratory practice, and supervised clinical placement. The educational goal should be development of capability - the ability to adapt knowledge and skills to uncertain clinical situations - rather than completion of isolated technical competencies alone. This model offers a testable framework for context-sensitive implementation and future evaluation.

Plain Language Summary

This article is a commentary, not an original experimental study. It discusses how AI-supported virtual simulation might help radiography students practise safely before using real imaging equipment or working with patients. Research shows that virtual simulation can support positioning, exposure selection, image evaluation, and preparation for clinical practice, but it cannot reproduce every physical, interpersonal, and professional aspect of real radiography. AI adds possible functions such as adaptive feedback and case generation, but AI-generated material must be checked by qualified educators before use. For Somalia, published information confirms that radiography training programs and modern imaging services exist, but national evidence about teaching-laboratory capacity and the distribution of specific technologies is still limited. We therefore propose a blended approach: digital or AI-supported preparation first, focused hands-on laboratory practice second, and supervised clinical placement third. This approach aims to use scarce practical training time more effectively without weakening patient safety or clinical learning.

Introduction

Radiography education integrates scientific knowledge with psychomotor performance, communication, radiation protection, image evaluation, and clinical judgment. Simulation is valuable because it allows students to rehearse selected decisions and procedures without exposing patients to avoidable risk. Recent radiography reviews report benefits of virtual and simulation-based learning for positioning, exposure-factor selection, confidence, and preparation for clinical placement, while also emphasizing heterogeneity in study design and the continued importance of authentic clinical learning.^1,2

This article is a commentary and does not report original experimental data. Its purpose is to develop a context-sensitive educational argument: AI-enhanced virtual simulation may be useful as an adjunct to, but not a substitute for, physical skills laboratory and supervised clinical education. The distinction matters because the evidence base for virtual reality (VR), conventional digital simulation, and AI-enhanced simulation is not interchangeable. The paper therefore separates these modalities, identifies what current evidence supports, and proposes a three-layer model that can be evaluated in Somalia and other settings where practical education resources are constrained.

Distinguishing Virtual Simulation, VR, and AI-Enhanced Simulation

Virtual simulation is a broad term for computer-based representations of clinical tasks or environments. VR is a subset that creates an interactive three-dimensional environment, which may be immersive or non-immersive. AI-enhanced simulation adds algorithmic functions that can modify or generate content, infer learner performance, personalize difficulty, provide adaptive feedback, generate cases or questions, or support analytics. A simulation can therefore be virtual without using AI, and an AI tool can support education without being a VR simulator.

This distinction is important for interpretation of evidence. Radiography studies of VR have shown transfer to selected clinical and objective structured clinical examination outcomes. In one study, first-year students who received immersive VR training performed better in several clinical assessment domains, including positioning, exposure-factor selection, and image appraisal. A randomized comparison of VR with physical simulation also reported fewer positioning and equipment-movement errors in the VR group, while exposure-setting errors did not differ significantly.^3,4

These findings support virtual simulation as a preparatory modality, but they do not by themselves establish an effect attributable to AI. AI-specific evidence is newer. A pilot study of generative AI case-based imaging simulation found educational promise but also limited clinical fidelity, with educators concluding that further training, testing, and validation were required before fuller integration.⁵

Why AI May Add Educational Value

AI may extend conventional simulation in four practical ways. First, generative systems can produce varied case prompts or scenarios, potentially increasing exposure to uncommon presentations. Second, adaptive systems can alter task difficulty or feedback according to learner performance. Third, natural-language interfaces can support conversational virtual patients or guided reflection. Fourth, AI can assist educators with formative item generation and learner analytics. These functions may be useful where faculty time or access to physical simulation is limited, but convenience should not be equated with validity.

Evidence from medical imaging education illustrates both opportunity and caution. In a 2025 study, 72.5% of 80 AI-generated medical imaging multiple-choice questions were judged suitable by faculty, yet most assessed recall rather than higher-order cognition. The authors emphasized the continued need for expert review. Radiography students have also reported using AI for learning and assessment while expressing concerns about reliability, ethics, and responsible use. Surveys of radiographers, radiologists, and students similarly identify gaps in AI education and the need for structured preparation.^6–8

Capability, Competence, and the Limits of Simulation

Radiography programs must assess discrete competencies, such as safe positioning or exposure selection, but the broader educational aim should be capability. In this commentary, competence refers to demonstrated performance against defined tasks or standards, whereas capability includes the ability to adapt knowledge and skills to unfamiliar, ambiguous, or changing clinical situations. Medical radiation education literature has highlighted that professional capability extends beyond task completion to responding to feedback, uncertainty, and new situations.⁹

This distinction limits what any simulator can claim. Virtual rehearsal can support procedural sequencing, reasoning, and deliberate practice, but it cannot fully reproduce tactile equipment handling, patient transfer, pain-limited positioning, infection prevention, unpredictable communication, interprofessional workflow, or responsibility for real clinical decisions. AI-generated feedback can also be confidently wrong or contextually inappropriate. Simulation performance should therefore be interpreted as evidence of preparation for practice, not proof of independent clinical capability.

A Context-Sensitive View of Resource Constraints

The terms low-resource and high-resource can obscure substantial variation within countries and institutions. Here, resource-constrained educational settings are defined operationally as programs in which one or more resources required for repeated practical learning - dedicated simulation space, functional imaging equipment for teaching, qualified faculty time, stable connectivity, standardized clinical placement capacity, or validated educational software - are insufficient relative to learner numbers and curriculum requirements. This definition refers to the educational environment, not simply to national income classification.

For Somalia, the available evidence should be interpreted carefully. A nationwide hospital assessment found that X-ray access was more common than ultrasound or computed tomography across surveyed hospitals, demonstrating heterogeneous diagnostic capacity rather than uniform absence of imaging technology. Public institutional information also confirms active radiography or radiological sciences programs at SIMAD University and the University of Health Sciences Somalia, while the latter describes its four-year program as having been inaugurated in 2021. These sources establish that formal training and modern imaging services exist, but they do not provide national data on the number or type of teaching laboratories, the distribution of digital radiography vs computed radiography systems, faculty-to-student ratios, or standardized clinical placement capacity.^10–12

Accordingly, statements about Somali radiography education should distinguish documented evidence from inference. A plausible need for efficient simulation does not justify assuming that all Somali institutions lack modern equipment. The more defensible approach is local needs assessment at each participating program: inventory available modalities and teaching equipment; document learner numbers and faculty supervision capacity; map clinical placement sites; assess internet and device access; and identify which competencies are currently limited by insufficient opportunities for repeated practice.

A Three-Layer Adjunct Model

We propose a three-layer model in which technology is used to protect, rather than displace, scarce hands-on and clinical learning time.

Layer 1: AI-Enhanced or Conventional Virtual Preparation

Students rehearse anatomy, examination indications, projection selection, positioning sequence, exposure decisions, radiation protection, image critique, and communication. AI may be added where it demonstrably improves adaptation or feedback; otherwise, conventional virtual simulation may be sufficient.

Layer 2: Focused Physical Laboratory Practice

Students translate rehearsed knowledge into psychomotor performance using real or training equipment. Faculty time is concentrated on tactile positioning, equipment movement, spatial awareness, infection control, patient transfer, and correction of misconceptions that a virtual environment cannot reveal.

Layer 3: Supervised Clinical Placement

Students develop capability in authentic care: adapting to patient variation, communicating under uncertainty, prioritizing safety, working within teams, and integrating technical decisions with clinical context. Simulation may inform readiness but should not replace supervised workplace assessment.

Table 1 summarizes an adjunct model rather than a replacement model.

Table 1.

Appropriate Role of AI-Enhanced Simulation Across Selected Radiography Learning Activities

Educational Activity	Appropriate Role of AI-Enhanced Simulation	Rationale and Safeguard
Selected preparatory demonstrations	May partially substitute	Interactive rehearsal may reduce dependence on passive demonstrations, provided content is faculty validated.
Basic positioning sequence rehearsal	May supplement	Supports repeated sequencing and spatial rehearsal before physical equipment use.
Exposure-factor decision exercises	May supplement	Can provide scenario-based practice; local protocols and faculty review remain necessary.
Radiation-protection decision making	May supplement	Useful for reasoning rehearsal, but cannot replace supervised application and local regulatory instruction.
Image critique and error recognition	May supplement	Can expand case exposure; AI-generated or AI-labeled images require expert validation.
Formative question generation	May supplement	Potentially efficient, but questions require expert review and psychometric evaluation before consequential use.
Physical equipment handling	Should not replace	Requires tactile interaction, equipment movement, spatial awareness, and supervised practice.
Patient transfer and tactile positioning	Should not replace	Requires real physical skills, communication, comfort assessment, and safety supervision.
Authentic patient communication	Should not replace	Virtual patients may prepare learners but cannot reproduce the full variability and accountability of real interaction.
Supervised clinical placement	Should not replace	Essential for workplace capability, professional identity, interprofessional practice, and adaptation to uncertainty.
Final clinical capability decisions	Should not replace	AI data may inform formative review, but accountable human assessors should make progression and readiness decisions.

Notes: “May partially substitute” is restricted to selected preparatory activities; it does not imply equivalence to physical laboratory or clinical education.

Abbreviation: AI, artificial intelligence.

Expert Validation and Assessment Quality

AI-generated educational material should not enter a curriculum or assessment simply because it is fluent or plausible. For teaching content, qualified radiography educators should verify factual accuracy, alignment with local protocols, realism, cultural and linguistic appropriateness, and radiation-safety content. For assessment items, expert review should be followed, where feasible, by psychometric evaluation such as item difficulty, discrimination, reliability, and evidence that the assessment samples the intended cognitive level. The medical imaging chatbot study is instructive: faculty deemed many AI-generated items usable, but most remained recall-focused, demonstrating why automated generation cannot substitute for assessment design expertise.⁶

High-stakes decisions require even greater caution. AI feedback can support formative learning, but progression, readiness for unsupervised procedures, and final clinical judgments should remain under accountable human assessors. Educational programs should also document model version, prompts or task specifications where relevant, review procedures, data handling, and mechanisms for identifying erroneous outputs.

Ethical and Implementation Considerations

Implementation should follow principles of transparency, safety, accountability, inclusion, privacy, and human oversight. WHO guidance on AI for health and large multimodal models emphasizes governance, risk management, and continued human responsibility.^13,14

Equity requires particular attention. A digital intervention that assumes personal laptops, expensive headsets, or uninterrupted broadband may widen rather than reduce educational disparities. Programs should consider shared computer laboratories, mobile-compatible platforms, offline functionality, multilingual content, and accessibility for students with disabilities. Procurement decisions should compare total cost of ownership, including licensing, hardware replacement, faculty training, technical support, and local maintenance, with the educational benefit actually demonstrated.

Where patient-derived images or cases are used, institutions should ensure lawful data access, de-identification, appropriate permissions, and secure storage. Synthetic images should be clearly identified as synthetic and validated for educational fidelity before being used to teach pathology or image interpretation.

Implications for Curriculum and Research

The proposed model is intentionally testable. A Somali pilot could begin with one high-volume examination, such as chest or extremity radiography, and compare conventional preparation with virtual preparation plus the existing laboratory and clinical curriculum. Outcomes should move beyond satisfaction and confidence to include transfer into physical performance and clinical capability. Candidate measures include structured practical assessment, positioning and equipment errors, exposure-selection accuracy, image critique, radiation-protection behaviors, communication, time-on-task, faculty workload, access equity, and cost per learner.

Evaluation should also distinguish the contribution of AI from that of simulation itself. A three-arm design - conventional preparation, non-AI virtual simulation, and AI-enhanced virtual simulation - would be more informative than comparing AI simulation only with usual teaching. Such work could clarify whether AI adds meaningful educational value or merely adds complexity to an already effective simulation platform.

Recommendations

AI-enhanced simulation should be implemented as an adjunct to, rather than a replacement for, physical skills laboratory teaching and supervised clinical education, because radiography capability depends on both cognitive preparation and authentic psychomotor and patient-facing experience. Educational institutions should describe the technology used with precision and avoid attributing benefits demonstrated for virtual reality or conventional simulation to artificial intelligence unless supported by AI-specific evidence. Resource constraints should likewise be defined at the level of the educational program or learning environment, including access to equipment, faculty, connectivity, simulation infrastructure, and clinical placement opportunities, rather than assuming that all institutions within a particular country face the same limitations. Curriculum design should extend beyond task competence toward capability, enabling learners to adapt knowledge, technical skills, judgment, and professional behavior to unfamiliar or complex clinical situations; this requires assessment of transfer from simulated learning to authentic clinical performance. AI-generated cases, images, feedback, questions, and assessment materials should undergo subject-matter expert review and, where used for formal assessment, appropriate psychometric evaluation before implementation. Platform selection and deployment should also safeguard equitable access, data privacy, cultural and linguistic relevance, transparency, and continued human accountability for educational and assessment decisions. Finally, institutions should adopt a staged implementation strategy, beginning with locally evaluated pilot programs that examine educational effectiveness, feasibility, acceptability, safety, cost, and transfer to practice before broader institutional or national scale-up.

Conclusion

AI-enhanced virtual simulation has a plausible role in radiography education, but the strongest evidence currently supports a complementary rather than replacement model. VR and conventional simulation can improve selected preparatory skills, and emerging AI-specific studies suggest additional possibilities for adaptive content, case generation, and feedback. Those possibilities remain contingent on clinical fidelity, expert validation, assessment quality, and human oversight. In Somalia, formal radiography education and modern imaging services are developing, but national evidence on teaching-laboratory and modality distribution remains insufficient for broad scarcity claims. A three-layer model - virtual preparation, focused physical laboratory practice, and supervised clinical placement - provides a practical framework that can conserve scarce hands-on teaching time while preserving the experiences required for clinical capability. The next priority is not wider adoption by assumption, but careful local evaluation of what AI adds, for whom, under what conditions, and at what cost.

Funding Statement

This study was supported by SIMAD University, Mogadishu, Somalia.

Abbreviations

AI, artificial intelligence; CT, computed tomography; OSCE, Objective Structured Clinical Examination; VR, virtual reality; WHO, World Health Organization.

Data Sharing Statement

No datasets were generated or analyzed for this commentary.

Ethics Approval and Informed Consent

Ethics approval and informed consent are not required because this commentary does not report research involving human participants, human tissue, animals, or identifiable personal data.

Consent for Publication

This was deemed not applicable.

Disclosure

The author declares no conflicts of interest in this work.