Raghu VK, et al. - Using a population comprising 218 subjects with lung cancer or benign nodules, researchers used clinical, demographic and low-dose CT (LDCT) features to develop an efficient early lung cancer predictor. They integrated data from 92 subjects (training cohort) from the Pittsburgh Lung Screening Study cohort by using probabilistic graphical models (PGMs). They developed the Lung Cancer Causal Model (LCCM) by using three variables that were directly (causally) related to malignant nodules and the largest benign nodule. In a separate cohort of 126 subjects, they validated the model. LCCM enabled the detection of 30% of the benign nodules, without risk of misclassifying cancer nodules, as seen in the validation cohort. Findings demonstrated the promising ability of LCCM as a lung cancer predictor. It showed significant improvement over existing models.