Abstract

This dataset contains images of dermatological diseases. Data corresponds to patients collected from 2010 to 2020 (about 300 pictures of child dermatological diseases). Data was collected by the Azienda Ospedaliero-Universitaria di Bologna (IRCCS). Data are manually annotated from the electronic medical records of the Pediatric Emergency Department and Pediatric Dermatology Outpatients’ Service and anonymized. The purpose of the dataset is to allow the creation of Machine Learning models for the classification of skin diseases in children. Additionally, it is possible to investigate how the bias present in the data (e.g., the presence of underrepresented groups) can affect the fairness of the resulting classification models. Images corresponding to 9 different diseases have been collected. The diseases are reported in Italian language (as the collection of the data has been done in an Italian hospital). However, the usage of the images is not dependent on the Italian language. The diseases are the following: esantema iatrogeno farmaco indotto (exanthem induced by drugs), esantema maculo papuloso (exanthem spotted/papulose), esantema morbilliforme (exanthem measles-like), esantema polimorfo like (exanthem polymorphous-like), esantema virale (exanthem of viral origin), orticaria (hives), pediculosi (pediculosis), scabbia (scabies), varicella (chickenpox). These are disease very common in children. This activity is part of the HORIZON-CL4-2021-HUMAN-01-24-AEQUITAS project (g.a. 101070363). The aim of AEQUITAS to address and tackle the multiple manifestations of bias and unfairness in Artificial Intelligence (AI) from a variety of dimensions, such as the development of AI tools, the data used to train, test and validate them or the interpretation practices developed around them.