In Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy
will defend his dissertation
Methods for Processing 3D Images for Breast Morphology
In this research, two novel algorithms are developed to facilitate quantitative evaluation of breast aesthetics for preoperative planning and postoperative assessment of outcomes in breast reconstruction surgery in cancer patients. First, an algorithm is presented for registering 3D images of individual patients from multiple clinical visits. Registration is performed to eliminate differences in object coordinate systems between the images due to variations in patient positioning and posture, thereby facilitating longitudinal comparison of morphological changes in the reconstructed breasts. Second, an algorithm to detect from 3D images the lowest breast contour, an important attribute for breast aesthetics, is presented. The algorithm allows detection of the lowest breast contour, for ptosis grades of 0, 1, 2, and 3. Most importantly, the algorithm operates independent of the presence of fiducial points such as the nipple, making it robust for applicability to images of breasts at intermediate time points during the reconstructive surgery that are devoid of nipples.
The applicability of the two algorithms is demonstrated in a multi-view 3D data fusion technique for visualization of the inframammary fold (IMF) in upright images from women with ptotic breasts. The upright view image is conventionally used for surgical planning and outcome assessment. However, the IMF, a critical landmark for breast surgery and morphometry, is typically occluded from the upright view in women with ptotic breasts. Multi-view 3D images taken at two different positions (upright (90°) and supine (0°)) are employed in a data fusion approach to superimpose the position of the anatomical IMF, on the 3D images of women with ptotic breasts wherein only the lowest breast contour is visible.
Contributions of this research: (1) The rigid registration algorithm is more effective for multiple-visit images than traditional registration methods. This algorithm outperforms existing ICP algorithms and is robust to variations in body mass index (BMI). (2) The lowest breast contour detection algorithm computes contours in 3D images directly and employs curvature analysis. It effectively identifies contours when compared to current methods, which detect contours in 2D images. (3) Multi-view 3D data fusion technique is a first attempt to visualize the IMF in upright images for women with ptotic breasts. This approach enables the physician to visualize the position of the IMF in the upright images of women with high degrees (≥2) of breast ptosis.
Future work will focus on utilizing the algorithms developed in this study for facilitating quantitative comparison of changes in breast morphology occurring during the time course of several reconstructive procedures.
Date: Friday, April 24, 2015
Time: 3:30 PM
Place: T2 304 (ET Conference Room)
Advisor: Prof. Fatima Merchant, Prof. Shishir Shah
Faculty, students, and the general public are invited.