Trans-Rectal Optical Tomography Reconstruction Using 3-Dimensional Spatial Prior Extracted From Sparse 2-Dimensional Trans-Rectal Ultrasound Imagery

Abstract

Accurate prostate segmentation in trans-rectal ultrasound (TRUS) imagery is an important step in different clinical applications, and it is particularly necessary for providing a 3-dimensional spatial prior to guide the image reconstruction of trans-rectal optical tomography for prostate cancer detection. Utilizing the US prior to guide near infrared tomography reconstruction could be performed by direct segmentation of the US image. Therefore, 2-dimensional segmentation of the axial TRUS images are performed extensively, however, 2-dimensional segmentation of the sagittal TRUS images are challenging, due to more complexities in contrast, morphological features and image artifacts, as well as significant inter-subject variations of the prostate shape and size. We develop a routine of segmenting 2-dimensional TRUS images obtained from canine prostate, based on the combination of a Snake's algorithm and selected manual segmentation. The segmentations obtained from a sparse set of axial and sagittal images are aligned to form the 3-dimensional contour of a prostate. The resulted prostate profile is implemented as the spatial prior to constrain image reconstruction of trans-rectal optical tomography. The trans-rectal optical tomography images reconstructed with the prostate profile prior are compared with those reconstructed without any spatial prior by monitoring oxygen saturation (StO2) and total hemoglobin concentration ([HbT]) in lesions of a canine prostate.