Features extraction of prostate with graph spectral method for prostate cancer detection
Weiwei Du, Yipeng Liu, Shiyang Wang, Yahui Peng, Aytekin Oto · 2016
Prostate cancers were segmented directly in T2-weighted images in some studies of computer-aided detection (CAD). These methods don't consider the differences between lesion and non-lesion region in T2-weighted images, so some lesions are not easy to be detected. In this paper, to consider the differences between lesion and non-lesion region, some features extraction is proposed by using graph spectral method. First, whole prostate is extracted. And then, some statistics are computed in the region of prostate to find differences between cancer and noncancer in prostate. The statistics are mapped into m dimensional Euclidean space by using graph spectral method to detect prostate cancers. Experiments show features with m dimensional Euclidean space by using graph spectral method can find some differences between lesion and non-lesion region.