188 The development of AI for visualizing PPP

Due to chronic course of palmoplantar pustulosis (PPP), patients may visit a medical institution only when it worsens, and exacerbations remissions repeat. Artificial Intelligence (AI) that can quantify visualize such daily habits will help them understand their own condition grasp changes, leading an increase in motivation for treatment. To generate AI, segmentation (Seg1) was performed on the skin images showing hands soles individuals with certain clinical conditions, including PPP. Segmentation (Seg2) done photographs palms The sites were not segmented as lesions treated healthy used learning data set. In addition, Seg2 carried out at five facilities reduce selection bias. For machine learning, Seg1 discriminate fingers by Generative Adversarial Networks. After that, we developed AI estimates using images. assigning colors major changes (erythema, keratinization, white pustules, brown pustules vesicles), which allowed overlap, others (fine dusts, unclassified objects, out-of-focus areas). excluding areas color, mixed color analyzed. compared correct answers identification accuracy (recall rate follows; erythema: 0.49, scale: 0.39, vesicle: 0.30, pustule: 0.55, 0.07). We hope evaluate same level specialists, "visualize" improve quality