Bibliographic record
Abstract
Background Skin diseases affect 2.3 billion people globally. Due to the scarcity of dermatologists, 2 in 3 cases are seen by primary care physicians (PCPs), who have lower diagnostic accuracy. Published studies have shown that the diagnostic accuracy of a PCP or general practitioner is close to 50%. Objective The aim of this study was to build artificial intelligence (AI) classifiers across 26 and 54 common and urgent adult rashes that present in a primary care setting. Methods We trained our AI models with approximately 50,000 total photos. The number of images within each disease or class ranged from 76 to 5505. Additionally, we further tested narrowing the differential diagnosis by adding body part information to identify how this impacts top-5 accuracy for one condition. Results Overall, we trained an AI model to identify 26 classes on par with the accuracy level of a dermatologist, who is, on average, 75% top-3 accurate across 26 conditions. Additionally, we trained the AI model across 54 conditions and achieved 74.3% top-5 accuracy across common conditions and 79.2% top-5 accuracy across urgent conditions. In evaluating if body part information may increase top-5 accuracy, we saw top-5 accuracy for one condition increase from 67% to 97%. Conclusions Overall, we concluded that including body part information to down-select possible disease matches substantially increased the overall differential diagnosis accuracy for body region–specific conditions. We also concluded that AI may assist PCPs to identify the most likely skin conditions quickly in a clinical encounter, improve overall diagnostic accuracy, and inform the most appropriate next step for the patient. These promising findings highlight the need and potential of AI and clinical decision support to augment the ability of PCPs to accurately and confidently evaluate patients with skin conditions. Conflicts of Interest SC is the cofounder and chief executive officer of Piction Health, a company focused on using AI to help augment frontline providers’ clinical decision making to save time and improve outcomes for patients with skin diseases. SC also holds shares in Piction Health.
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 | 0.001 | 0.005 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.003 | 0.002 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.017 | 0.004 |
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".