For the lung nodule screening, one of the commonly testing methods is the chest radiograph. However, it is difficult to judge with the naked eye with the initial nodule size is usually less than one centimeter. It is known that skilled pulmonary radiologists have a high degree of accuracy in diagnosis, but there remain problems in disease diagnosis. These problems include the miss rate for diagnosis of small nodules and the diagnosis of change in preexisting interstitial lung disease. The recent studies have found that 68% lung cancer nodules in radiographs can be detected by one reader and 82% by two readers. In order to solve this problem, we proposed a 3D-CNN predicting model to differ malignant nodules from all nodules in computed tomography scan. In the experiment results, the model was able to achieve a training accuracy of 100% and a testing accuracy of 94.52%. It shows the proposed model is able to be used for improving the accuracy of detecting nodules.