Abstract
Use of population imaging data in combination of image quantification and ideas from big data analytic technique might open an opportunity for biomedical imaging to contribute in tackling the healthcare problems of next generation. Mammographic breast density is well established marker for predicting breast cancer risk which could be obtained at a relatively low cost. We have developed a population-based tissue probability map technique to enable reliable and fully automated segmentation of glandular tissue and thereby providing mammographic breast density. Our technique creates tissue probability map by using local statistics from ROI's drawn by expert readers and incorporates it into a level set scheme. We applied the similar approach to emphysema index quantification in chest CT imaging. We extracted a set of statistical data related to emphysema indices from CT imaging data and EMR of healthy population, and investigated underlying associations. After applying a multivariate regression, we could build a model that effectively normalized those confounding factors and provided precise index for emphysema risk.
