Network construction apparatus, method and program

A network construction method that extracts, as a plurality of segments, each of a plurality of predetermined structures, each including a straight line portion intermingled with each other, from medical image data, and that constructs a plurality of network structures by connecting the extracted plurality of segments for each of the plurality of predetermined structures, the method comprising the steps of:

• calculating a likelihood of a cross segment representing a probability that each of the extracted plurality of segments is a cross segment, which is a segment, based on a cross segment characteristic condition defining a geometric characteristic between segments in the neighborhood of the intermingled straight line portion that a portion corresponding to the intermingled straight line portion of each of the plurality of predetermined structures is present as a cross segment, and that at least two pairs of segments connectable to each other in a straight line are present in the neighborhood of the cross segment;
• calculating a likelihood of a straight line representing a probability that each of the segments is connected, in a straight line, to another segment based on a straight line connection condition defining a geometric characteristic between the segments that each of the segments is connected, in a straight line, to another segment in the neighborhood of each of the segments;
• setting a strength of connection between the segments by using the likelihood of a cross segment and the likelihood of a straight line with respect to each of the extracted plurality of segments;
• and
• constructing the plurality of network structures by connecting the segments based on the set strength of connection.

A non-transitory computer-readable recording medium stored therein a network construction program that extracts, as a plurality of segments, each of a plurality of predetermined structures, each including a straight line portion intermingled with each other, from medical image data, and that constructs a plurality of network structures by connecting the extracted plurality of segments for each of the plurality of predetermined structures, the program causing a computer to function as:

• a strength-of-connection setting unit that includes:
• a likelihood-of-cross-segment calculation unit that calculates a likelihood of a cross segment representing a probability that each of the extracted plurality of segments is a cross segment, which is a segment, based on a cross segment characteristic condition defining a geometric characteristic between segments in the neighborhood of the intermingled straight line portion that a portion corresponding to the intermingled straight line portion of each of the plurality of predetermined structures is present as a cross segment, and that at least two pairs of segments connectable to each other in a straight line are present in the neighborhood of the cross segment;
• and
• a likelihood-of-straight-line calculation unit that calculates a likelihood of a straight line representing a probability that each of the segments is connected, in a straight line, to another segment based on a straight line connection condition defining a geometric characteristic between the segments that each of the segments is connected, in a straight line, to another segment in the neighborhood of each of the segments, and
• the strength-of-connection setting unit setting a strength of connection between the segments by using the likelihood of a cross segment and the likelihood of a straight line with respect to each of the extracted plurality of segments;
• and
• a network structure construction unit that constructs the plurality of network structures by connecting the segments based on the set strength of connection.