A multi-center chest CT dataset for differential diagnosis of pediatric Mycoplasma pneumoniae pneumonia and subtypes

Children with Mycoplasma pneumoniae pneumonia (MPP) represent a common pediatric respiratory infection, among which severe Mycoplasma pneumoniae pneumonia (SMPP) is associated with a high risk of complications. Early and accurate identification of SMPP is therefore crucial for clinical decision-making. However, existing computer-aided diagnosis (CAD) studies are limited by the lack of high-quality imaging datasets specifically designed for fine-grained MPP severity classification. To address this gap, this study constructed and publicly released a large-scale, multicenter chest CT PNG sequence dataset comprising 1,164 independent pediatric patients. The dataset was collected from two medical institutions and includes four clinical categories: mixed infectious pneumonia (HH), common MPP, severe SMPP, and viral pneumonia. All cases were jointly verified using clinical gold-standard criteria and expert review by senior imaging professionals, ensuring the reliability and authority of the annotations for 89,053 CT slices. This dataset is intended to provide a reliable benchmark for developing deep learning-based diagnostic models for complex pediatric pneumonia and to promote the advancement of precision medicine in pediatrics.