Improving dermatology classifiers across populations using images generated by large diffusion models
This addresses the issue of sampling bias in dermatological AI for underrepresented populations, but is incremental as it builds on existing data augmentation methods using a new tool.
The paper tackled the problem of poor generalization of dermatology classifiers across populations due to insufficiently diverse training data, and showed that augmenting training with DALL·E 2-generated synthetic images improved classification overall and for underrepresented groups, as demonstrated on the Fitzpatrick 17k dataset.
Dermatological classification algorithms developed without sufficiently diverse training data may generalize poorly across populations. While intentional data collection and annotation offer the best means for improving representation, new computational approaches for generating training data may also aid in mitigating the effects of sampling bias. In this paper, we show that DALL$\cdot$E 2, a large-scale text-to-image diffusion model, can produce photorealistic images of skin disease across skin types. Using the Fitzpatrick 17k dataset as a benchmark, we demonstrate that augmenting training data with DALL$\cdot$E 2-generated synthetic images improves classification of skin disease overall and especially for underrepresented groups.