An Open Dataset and an Algorithm for Niacin Skin-flushing Segmentation with Application in Psychiatric Disorders Classification

<strong>About the Dataset</strong> <strong>Summary:</strong> The dataset includes images associated with the Niacin Skin-Flushing Response (NSR), a physiological response connected with mental illnesses such as Depression (DP), Bipolar Disorder (BP), and Schizophrenia (SZ), in addition to images from a healthy control group (HC). The data was initially part of a larger dataset collected from inpatients diagnosed according to the International Classification of Diseases, 10th Revision (ICD10). The HC group includes individuals with no lifetime history of psychiatric illness. A subset of images was selected from the original dataset to create an AI screening tool. These images had areas of interest cut out, and the NSR areas were annotated by five students. A pixel was marked as a flushing pixel if at least two students marked it as such. While source files for the developing segmentation algorithms cannot be provided due to IP issues, the results can be viewed on the benchmark page. For collaboration interests, feel free to get in touch. <strong>Dataset Details:</strong> The dataset includes: Number of images: 600 (Formatted as 1.png, 2.png, ... 600.png) Format of images: PNG Contents of the original image: A common label for reference size and four flushing areas Marked images: Images with human-determined skin-flushing areas (Formatted as 1-marked.png, ..., 600-marked.png) Reference score (info.csv): A human-determined relative size of the flushing area compared to the label. <strong>Ethical Considerations:</strong> The data collection process followed guidelines set by the local ethics committee, with all participants having signed informed consent forms. Strict inclusion criteria were followed to ensure the health and safety of participants. The data has been anonymized and does not contain any personally identifiable information.