A CAD system for the 3D location of lesions in mammograms
Collaborators：R. Highnam、M. Brady(Oxford Univ.）、E. Tohno（Tsukuba Univ.）
A CAD system for estimating the 3D (three-dimensional) positions of lesions found in two mammographic views is poposed. The system is an extension of our previous method [1)] which finds corresponding 2D positions in different mammographic views. The method calculates curved epipolar lines by developing a simulation of breast deformation into stereo camera geometry. Using such curved epipolar lines, not only can we determine point correspondences, but can estimate the 3D location of a lesion.
This figure shows an overview of our method [1)] , which enables us to calculate the the system. First, suppose that a point in one image is pointed at by a radiologist. The method calculates the epipolar curve, that is the locus of possible corresponding positions of the point in the other image, by simulating the five steps of the following process, A: back projection → B:uncompression → C: rotation → D: compression → E:projection, as shown by the solid arrows. Next, the corresponding position is searched for along the epipolar curve. Once the correspondence is found along the curve, the corresponding 3D position in the uncompressed breast can be determined by retracing the movement of the point during the simulation, as shown by the dashed arrows.
The 3D locations obtained by the system have been examined from a clinical viewpoint. We tentatively conclude that the system achieves errors within 10?20 mm in estimating the 3D locations of lesions. Most error seems to arise in the depth direction. We believe that the accuracy in the depth can be improved by replacing the simplified compression model used for the system with a richer compression model, and indeed we are continuing to work in this area.
1) Y. Kita、E. Tohno, R. Highnam、M. Brady: “A CAD system for the 3D location of lesions in mammograms”, Medical Image Analysis，Vol. 6,No.3, pp.267-273、2002
2) Y. Kita, R. Highnam and M. Brady: "Correspondence between different view breast X rays using curved epipolar lines", Computer Vision and Image Understanding, pp. 38-55,2001