Design and Evaluation of a Breast-Specific Collimator Using Response Surface Methodology and Monte Carlo Simulations

Design optimization of collimators for various cameras and scanning procedures has been a challenge for many years. The purpose of this study is to propose a new approach to optimize the collimator for a planar small field-of-view continuous crystal breast-specific camera by making use of validated Monte Carlo simulations and the response surface methodology, such that the lesion contrast-to-noise ratio (CNR) is maximized. CNR was described as a function of collimator parameters (hole diameter, septal thickness and hole length) by fitting a quadratic model and then optimized by using a numerical direct search method. The proposed method can allow the modeling of complex patient and camera configurations and perform a search in the design space for the optimum collimation by considering all the interactions between the parameters. The optimization was performed for lesions of 5 mm diameter lying at 3 cm and 5 cm depths in a 6 cm thick breast phantom in order to be acquainted with the behavior of the optimizer as the source-to-collimator distance changes. The results show that, for a worst-case lesion at equal distance from both sides of the breast, the combination of hole length of 1.137 cm, septal thickness of 0.029 cm and hole diameter of 0.125 cm is the optimal solution that provides the highest CNR of 3.34. As a result, a considerable improvement in CNR up to 73% was obtained with respect to the reference collimators. It was also shown that the critical region for detectability in the contrast detail curves shifted towards the direction of smaller lesion diameter and lower tumor-to-background ratio. This study demonstrated that the pairwise interaction effects of the collimator parameters play a key role in determining the set of design parameter combinations that yield the highest CNR. Another conclusion of the study is that the optimizer adapts itself to the spatial resolution/sensitivity trade-off as the lesion depth changes. We therefore conclude that the advantages obtained with this method may lead to an advancement in the state of the art in collimator design.