Spot the Difference: Can ChatGPT4-Vision Transform Radiology Artificial Intelligence?
Bibliographic record
Abstract
Abstract OpenAI’s flagship Large Language Model ChatGPT can now accept image input (GPT4V). “Spot the Difference” and “Medical” have been suggested as emerging applications. The interpretation of medical images is a dynamic process not a static task. Diagnosis and treatment of Multiple Sclerosis is dependent on identification of radiologic change. We aimed to compare the zero-shot performance of GPT4V to a trained U-Net and Vision Transformer (ViT) for the identification of progression of MS on MRI. 170 patients were included. 100 unseen paired images were randomly used for testing. Both U-Net and ViT had 94% accuracy while GPT4V had 85%. GPT4V gave overly cautious non-answers in 6 cases. GPT4V had a precision, recall and F1 score of 0.896, 0.915, 0.905 compared to 1.0, 0.88 and 0.936 for U-Net and 0.94, 0.94, 0.94 for ViT. The impressive performance compared to trained models and a no-code drag and drop interface suggest GPT4V has the potential to disrupt AI radiology research. However misclassified cases, hallucinations and overly cautious non-answers confirm that it is not ready for clinical use. GPT4V’s widespread availability and relatively high error rate highlight the need for caution and education for lay-users, especially those with limited access to expert healthcare. Key points Even without fine tuning and without the need for prior coding experience or additional hardware, GPT4V can perform a zero-shot radiologic change detection task with reasonable accuracy. We find GPT4V does not match the performance of established state of the art computer vision models. GPT4V’s performance metrics are more similar to the vision transformers than the convolutional neural networks, giving some possible insight into its underlying architecture. This is an exploratory experimental study and GPT4V is not intended for use as a medical device. Summary statement GPT4V can identify radiologic progression of Multiple Sclerosis in a simplified experimental setting. However GPT4V is not a medical device and its widespread availability and relatively high error rate highlight the need for caution and education for lay-users, especially those with limited access to expert healthcare.
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 imitationNot 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.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.004 | 0.017 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.002 | 0.004 |
| Open science | 0.003 | 0.002 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.018 | 0.008 |
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 source (direct Gemma or distilled Codex), 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".