LG210: "Prostate Segmentation Based on Variant Scale Patch and Local Independent Projection," "Hierarchical and Symmetric Infant Image Registration by Robust Longitudinal-Example-Guided Correspondence Detection," and "Semi-automatic Segmentation of Brain

Date: 
Friday, November 14, 2014 - 12:00
Speaker: 
Yao Wu, PhD
Abstract: 
Prostate Segmentation Based on Variant Scale Patch and Local Independent Projection Accurate segmentation of the prostate in computed tomography (CT) images is important in image-guided radiotherapy; however, difficulties remain associated with this task. In this study, an automatic framework is designed for prostate segmentation in CT images. We propose a novel image feature extraction method, namely, variant scale patch, which can provide rich image information in a low dimensional feature space. We assume that the samples from different classes lie on different nonlinear submanifolds and design a new segmentation criterion called local independent projection (LIP). In our method, a dictionary containing training samples is constructed. To utilize the latest image information, we use an online updated strategy to construct this dictionary. In the proposed LIP, locality is emphasized rather than sparsity; local anchor embedding is performed to determine the dictionary coefficients. Several morphological operations are performed to improve the achieved results. The proposed method has been evaluated based on 330 3-D images of 24 patients. Results show that the proposed method is robust and effective in segmenting prostate in CT images. Hierarchical and Symmetric Infant Image Registration by Robust Longitudinal- Example-Guided Correspondence Detection To investigate anatomical differences across individual subjects, or longitudinal tissue appearance changes that are typically investigated in early brain development studies, an accurate image registration is essential. However, due to fast brain development and dynamic tissue appearance changes, it is very difficult to align infant brain images acquired from birth to 1-year-old. To solve this challenging problem, a novel image registration method is proposed to accurately align two infant brain images, regardless of their ages at acquisition. The main idea is to utilize growth trajectories, or spatial-temporal correspondences, learned from a set of longitudinal training images. In essence, the learned growth trajectory models are able to leverage image registration to tackle the image appearance gap between two images at different time points. Specifically, in the training stage, an intrinsic growth trajectory is first estimated for each training subject with longitudinal images. Then, to register the two new infant images with a possible large age gap, we identify the corresponding image patches between each new image and its respective training images with similar age. Finally, the registration between the two new images can be assisted by the learned growth trajectories from one time point to another time point that have been established in the training stage. To further improve registration accuracy, our infant registration method are combined with a hierarchical and symmetric registration framework that can iteratively add new key points in both images to steer the estimation of deformation pathway between the two infant brain images under registration. To evaluate image registration accuracy, the proposed method was used to align 24 real infant subjects from 2-week-old to 1-year-old, and when compared to the state-of-the-art methods, the proposed method demonstrated superior registration performance. Semi-automatic Segmentation of Brain Tumors Using Population and Individual Information Efficient segmentation of tumors in medical images is of great practical importance in early diagnosis and radiation plan. This paper proposes a novel semi-automatic segmentation method based on population and individual statistical information to segment brain tumors in magnetic resonance (MR) images. First, high-dimensional image features are extracted. Neighborhood components analysis is proposed to learn two optimal distance metrics, which contain population and patient-specific information, respectively. The probability of each pixel belonging to the foreground (tumor) and the background is estimated by the k-nearest neighborhood classifier under the learned optimal distance metrics. A cost function for segmentation is constructed through these probabilities and is optimized using graph cuts. Finally, some morphological operations are performed to improve the achieved segmentation results. Our dataset consists of 137 brain MR images, including 68 for training and 69 for testing. The proposed method overcomes segmentation difficulties caused by the uneven gray level distribution of the tumors and even can get satisfactory results if the tumors have fuzzy edges. Experimental results demonstrate that the proposed method is robust to brain tumor segmentation.