Figure 1.

The architecture of the machine learning (ML) method. Input x is the patient whose risk of abnormal CT scan is being evaluated. L _t is the training set consisting of K samples (x _k, y _k) where k = 1, 2, …, K and y _k is the class label. By using the training data, a total of T individual classifiers φ _t(x, L _t) are created to form the decision ensemble. Each individual classifier is built based on a subset of the training data. Then the prediction outcomes are combined by means of majority voting scheme to generate a final risk score for patient x.