Evaluating the Diagnostic and Treatment Capabilities of GPT-4 Vision in Dermatology: A Pilot Study
Bibliographic record
Abstract
Background: The integration of generative artificial intelligence within dermatology presents a new frontier for enhancing diagnostic accuracy and treatment planning. Objective: This research evaluates Generative Pre-trained Transformer-4 Vision’s (GPT-4V) performance in accurately diagnosing and generating treatment plans for common dermatological conditions, comparing its assessment of textual versus image data and its performance with multimodal inputs. Methods: A dataset of 102 images representing 9 common dermatological conditions was compiled from dermatlas.org and dermnet.nz . Images were screened by 2 board-certified dermatologists and were excluded if they did not represent a classic presentation of the respective conditions. Fifty-four images were included in the final analysis. In addition, 9 text-based clinical scenarios corresponding to each condition were developed. GPT-4V’s diagnostic capabilities were assessed across 3 setups: Image Prompt, Scenario Prompt, and Image + Scenario Prompt. Results: In the Image Prompt setup, GPT-4V correctly identified the primary diagnosis for 54% of the images. The Scenario Prompt and the Image + Scenario Prompt setups, respectively, both achieved an 89% accuracy rate in identifying the primary diagnosis. Treatment recommendations were evaluated using a modified Entrustment Scale, showing competent but not expert-level performance. A Wilcoxon signed-rank test demonstrated a statistically significant difference in treatment recommendations based on the Entrustment Score, with the model performing better in the Image + Scenario setup ( P < .01). Conclusion: GPT-4V demonstrates the potential to augment dermatological diagnosis and treatment recommendations, particularly in text-based scenarios. However, its underwhelming performance in image-based diagnosis and integration of multimodal data highlights important areas for improvement.
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 | 0.003 | 0.009 |
| 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 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.003 | 0.001 |
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.
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".