Lung nodule malignancy prediction in chest CT scans based on a CNN model with auxiliary task learning

Convolutional neural networks (CNNs) have been increasingly applied to computer-aided diagnosis (CADx) for lung nodule malignancy prediction, which usually is a binary classification task. However, CNNs were often difficult to capture optimal features, thereby affect the classification performance. This study developed a CADx system based on a CNN model (VI-Net) with auxiliary task learning to predict lung nodule malignancy in chest computed tomography (CT) scans. Our CADx system took CT image cubes containing lung nodules as input and generated one main output and eight auxiliary outputs. The main output predicted lung nodule malignancy; the auxiliary outputs predicted lesion size and some lesion characteristics. The auxiliary tasks offered assistance for predicting the final nodule malignancy. The CNN with auxiliary task learning was trained as a whole by optimizing a global loss function including all tasks. The performance of the developed lung nodule CADx system was verified by use of the Lung Image Database Consortium (LIDC) dataset. The lung nodule malignancy prediction results were quantitatively evaluated by using the area under the ROC curve (AUC), accuracy, sensitivity, and specificity. The evaluation results showed that our CADx system achieved improved performance for lung nodule malignancy prediction. The auxiliary task learning not only helped to predict the lung nodule malignancy, but also contributed to explain the prediction to some extent.