Physics-informed data augmentation to simulate low dose CT scans: Application to lung nodule detection

The authors investigate how to improve the performance of convolutional neural networks (CNNs) in detecting lung nodules on low-dose CT scans, which are often affected by noise characteristics from different imaging systems. They propose a Physics-Informed Data Augmentation (PIDA) method that simulates low-dose noise by leveraging data from higher-dose scans, ultimately enhancing the training dataset's variability. Their results demonstrate that incorporating PIDA significantly improves the CNN's detection performance, addressing the challenges posed by differences in training and testing data acquisition.