Early detection of lung cancer is essential in reducing mortality. Recent studies have demonstrated the clinical utility oflow-dose computed tomography (CT) to detect lung cancer among individuals selected based on very limited clinicalinformation. However, this strategy yields high false positive rates, which can lead to unnecessary and potentially harmfulprocedures. To address such challenges, we established a pipeline that co-learns from detailed clinical demographics and3D CT images. Toward this end, we leveraged data from the Consortium for Molecular and Cellular Characterization ofScreen-Detected Lesions (MCL), which focuses on early detection of lung cancer. A 3D attention-based deepconvolutional neural net (DCNN) is proposed to identify lung cancer from the chest CT scan without prior anatomicallocation of the suspicious nodule. To improve upon the non-invasive discrimination between benign and malignant, weapplied a random forest classifier to a dataset integrating clinical information to imaging data. The results show that theAUC obtained from clinical demographics alone was 0.635 while the attention network alone reached an accuracy of0.687. In contrast when applying our proposed pipeline integrating clinical and imaging variables, we reached an AUC of0.787 on the testing dataset. The proposed network both efficiently captures anatomical information for classification andalso generates attention maps that explain the features that drive performance.
展开▼
