AI speechbots and 3D segmentations in virtual reality improve radiology on-call training in resource-limited settings
Bibliographic record
Abstract
Evaluate the use of large-language model (LLM) speechbot tools and deep learning-assisted generation of 3D reconstructions when integrated in a virtual reality (VR) setting to teach radiology on-call topics to radiology residents. Three first year radiology residents in Guyana were enrolled in an 8-week radiology course that focused on preparation for on-call duties. The course, delivered via VR headsets with custom software integrating LLM-powered speechbots trained on imaging reports and 3D reconstructions segmented with the help of a deep learning model. Each session focused on a specific radiology area, employing a didactic and case-based learning approach, enhanced with 3D reconstructions and an LLM-powered speechbot. Post-session, residents reassessed their knowledge and provided feedback on their VR and LLM-powered speechbot experiences. Residents found that the 3D reconstructions segmented semi-automatically by deep learning algorithms and AI-driven self-learning via speechbot was highly valuable. The 3D reconstructions, especially in the interventional radiology session, were helpful and the benefit is augmented by VR where navigating the models is seamless and perception of depth is pronounced. Residents also found conversing with the AI-speechbot seamless and was valuable in their post session self-learning. The major drawback of VR was motion sickness, which was mild and improved over time. AI-assisted VR radiology education could be used to develop new and accessible ways of teaching a variety of radiology topics in a seamless and cost-effective way. This could be especially useful in supporting radiology education remotely in regions which lack local radiology expertise. • Virtual reality with LLM-powered speechbots addresses radiology training challenges in low-resource settings like Guyana. • Residents' quiz scores improved after training, with positive feedback on 3D reconstructions and the speechbot's usefulness. • The platform bridges gaps in expert-led training, supporting scalable self-directed learning with minimal health effects reported.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.002 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".