II.B. Decision trees
II.B.1. Concept
It is both natural and intuitive to classify a suspicious lesion seen in a mammogram using a sequence of questions, in which the next question depends on the answer to the previous question. A typical example would be to classify a suspicious mass in a mammogram by starting with the question “What is the shape of the mass?” and if the answer is “irregular” to proceed with the question “How does the margin of the mass look?” The answer to this second question might be “spiculated,” in which case the classification process might already be finished with the final classification “malignant.” Such sequences of questions can be represented by a directed tree structure called a decision tree. In a decision tree, nodes represent questions and directed links between the nodes represent the possible answers to the questions. Terminal nodes are called leaf nodes and represent