MétaCan
Menu
← Back to cohort
Record W4409726734 · doi:10.1101/2025.04.22.25326204

Artificial Intelligence in Neuro-Oncology: Assessing ChatGPT’s Accuracy in MRI Interpretation and Treatment Advice

2025· preprint· en· W4409726734 on OpenAlexaff
Abdullah Ishaque, Alexandre Boutet, Shivaprakash B. Hiremath, Matthew P. Mullarkey, Maria Peris‐Celda, Gelareh Zadeh

Bibliographic record

VenuemedRxiv · 2025
Typepreprint
Languageen
FieldMedicine
TopicArtificial Intelligence in Healthcare and Education
Canadian institutionsUniversity of Toronto
Fundersnot available
KeywordsAdvice (programming)Interpretation (philosophy)Medical physicsMedicinePsychologyArtificial intelligenceOncologyComputer science

Abstract

fetched live from OpenAlex

Abstract Purpose Large language models (LLMs) have demonstrated advanced capabilities in interpreting text and visual inputs. Their potential to transform oncological practice is significant, but their accuracy and reliability in interpreting medical imaging and offering management suggestions remain underexplored. This study aimed to evaluate the performance of ChatGPT in interpreting T1-weighted contrast-enhanced MRI images of meningiomas and glioblastomas and providing treatment recommendations based on simulated patient inquiries. Methods This observational cohort study utilized publicly available MRI datasets. Thirty cases of meningiomas and glioblastomas were randomly selected, yielding 90 images (three orthogonal planes per case). ChatGPT-4o was tasked with interpreting these images and responding to six standardized patient-simulated questions. Two neuroradiologists and neurosurgeons assessed ChatGPT’s performance using five-point Likert scales and their inter-rater agreement was evaluated. Results ChatGPT identified MRI sequences with 91.7% accuracy and localized tumors correctly in 66.7% of cases. Tumor size was qualitatively described in 85% of cases, and the median acceptability was rated as 4.0 (IQR 4.0–5.0) by neuroradiologists. ChatGPT included meningioma in the differential diagnosis for 73.3% of meningioma cases and glioma in 83.3% of glioblastoma cases. Inter-rater agreement among neuroradiologists ranged from moderate to good (κ = 0.45–0.72). While surgical treatment was suggested in all symptomatic cases, neurosurgeon acceptability ratings varied, with poor inter-rater reliability. Conclusions ChatGPT demonstrates potential in interpreting neuro-oncological MRI images and offering preliminary management recommendations. However, errors in tumor localization and variability in recommendation acceptability underscore the need for physician oversight and further refinement of LLMs before clinical integration.

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 machine prediction

Teacher imitation

Not calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.

metaresearch head score (Codex)0.026
metaresearch head score (Gemma)0.146
Version: metacan-v3-hybrid-931329e0061cValidation status: machine_predicted_unvalidated
Candidate categoriesnone
Consensus categoriesnone
DomainCandidate signal: none · Consensus signal: none
Study designCandidate signal: Bench or experimental · Consensus signal: none
GenreCandidate signal: Empirical · Consensus signal: Empirical
Teacher disagreement score0.026
Threshold uncertainty score0.138

Distilled classifier scores by category (both heads)

CategoryCodexGemma
Metaresearch0.0260.146
Meta-epidemiology (narrow)0.0010.000
Meta-epidemiology (broad)0.0010.001
Bibliometrics0.0020.001
Science and technology studies0.0010.001
Scholarly communication0.0020.001
Open science0.0010.003
Research integrity0.0010.001
Insufficient payload (model declined to judge)0.0010.001

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.

Opus teacher head0.139
GPT teacher head0.466
Teacher spread0.326 · how far apart the two teachers sit on this one work
Validation statusscore_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it

Classification

machine, unvalidated

Machine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.

The models applied no category: nothing in the taxonomy fit this work.
Study designBench or experimental
Domainnot available
GenreEmpirical

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".

Quick stats

Citations1
Published2025
Admission routes1
Has abstractyes

Explore more

Same venuemedRxiv→Same topicArtificial Intelligence in Healthcare and Education→French-language works237,207→