Deep Learning Skin Disease Classifiers: Current Status and Future Prospects
Bibliographic record
Abstract
Background Most studies on deep learning skin disease classifiers are done with binary classifications (ie, classifying lesions into malignant and benign). However, dermatology practice involves a large number of inflammatory and infective conditions that are not easily diagnosed by nondermatologist physicians. Objective The aim of this study is to develop a machine learning–based smartphone app for multiclass skin disease classification and evaluate its performance in different levels of dermatology practice. We will also explore similar studies in the literature. Methods We developed an artificial intelligence–driven smartphone app for 40 common skin diseases and tested it in primary care, tertiary care, and private practice settings. Results In the clinical study, the overall top-1 accuracy was 75.07% (95% CI 73.75%-76.36%), top-3 accuracy was 89.62% (95% CI 88.67%-90.52%), and the mean area under the curve was 0.90 (SD 0.07). Multimedia Appendix 1 shows the top-1 positive predictive values and negative predictive values from a clinical study of 35 diseases using the developed mobile health app on patients. In the literature, there are very few studies on image-based deep learning multiclass classification of common skin diseases and none of them included evaluations in actual clinical settings. Conclusions An artificial intelligence–driven smartphone app has the potential to improve the diagnosis and management of skin diseases in patients with skin of color. Nondermatologist, primary care physicians are likely to benefit from having access to such an app. Acknowledgments Nurithm Labs collaborated with the All India Institute Of Medical Science to develop this smartphone app. Other collaborators in this work include Rashi Pangti, Jyoti Mathur, Vikas Chouhan, Sharad Kumar, Lavina Rajput, Sandesh Shah, Atula Gupta, Ambika Dixit, Dhwani Dholakia, Sanjeev Gupta, Savera Gupta, Mariam George, and Vinod Kumar Sharma. Conflicts of Interest None declared.
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 | 0.006 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.002 | 0.003 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.002 | 0.003 |
| Insufficient payload (model declined to judge) | 0.004 | 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".