UniToChest

The use of technology in health is clearly a major driver towards more efficient healthcare, from whom both people and national health service budgets can benefit. European national healthcare systems are generating large biomedical imaging datasets because many medical examinations use image-based processes; these datasets are growing and constitute a large database of knowledge because most of their value derives from expert interpretation of those images. With the aim to promote eHealth innovation and improvement in Europe, the <em>Rad4AI</em> project, the Italian branch of the European <em>DeepHealth</em> project, promotes the development of standardized software to manipulate and process images in a more efficient way, thus increasing the productivity of professionals working on biomedical images. The proposed dataset <em><strong>UniToChest </strong></em>is a collection of anonymized<strong> </strong><strong>306440</strong><strong> </strong>chest CT scan slices coupled with the proper lung nodule segmentation map, for a total of <strong>10071</strong> nodules<strong> </strong>from <strong>623</strong> different patients. <em><strong>UniToChest </strong></em>is provided within the <em>DeepHealth </em>Project, by <em>Città della Salute e della Scienza di Torin</em>o in collaboration with the<em> Department of Computer Science </em>at<em> University of Turin</em>. In order to use the dataset as a training resource for AI algorithms, training, validation and test splits are provided. They have been created such that the training set contains CT scans of 80% of the patients, while validation and test set are both 10%.<br> An example of <em><strong>UniToChest </strong></em>usage can be found in DeepHealth GitHub repository. This implementation uses <em>DeepHealth</em> <em>EDDL &amp; ECVL </em>libraries to train a U-Net neural network model to predict nodules segmentation maps automatically.<br> Please refer to <em>"UniToChest: A Lung Image Dataset for Segmentation of Cancerous Nodules on CT Scans"</em> (ICIAP 2021) for more details.