Supplementary Materials

The PDF file includes:

  • Materials and Methods
  • Fig. S1. STARD diagram of data aggregation.
  • Fig. S2. Baseline DCNN model architecture.
  • Fig. S3. Preprocessing and splitting of the base database into training, validation, and testing sets.
  • Fig. S4. Data augmentation strategy.
  • Fig. S5. Transfer learning DCNN model architecture based on VGG16.
  • Fig. S6. Training, validation, and testing of fine-tuned VGG16 DCNN model.
  • Fig. S7. Transfer learning DCNN model architecture based on Xception.
  • Fig. S8. Training, validation, and testing of fine-tuned Xception DCNN model.
  • Fig. S9. Blob detection and naïve saliency calculation.
  • Fig. S10. Selected samples of DCNN ugly duckling outputs as compared with naïve saliency and dermatological consensus.
  • Table S1. Taxonomy of pigmented lesions included in our study’s baseline dataset.
  • Table S2. Distribution of Fitzpatrick skin tones along all skin-relevant classes in the base dataset.
  • Legends for data files S1 to S4

[Download PDF]

Other Supplementary Material for this manuscript includes the following:

  • Data file S1 (.jpg format). Montage of analysis outputs for wide-field images, numbers 1 to 35.
  • Data file S2 (.jpg format). Montage of analysis outputs for wide-field images, numbers 36 to 70.
  • Data file S3 (.jpg format). Montage of analysis outputs for wide-field images, numbers 71 to 105.
  • Data file S4 (.jpg format). Montage of analysis outputs for wide-field images, numbers 106 to 135.