Segmenting lung fields in serial chest radiographs using both population and patient-specific shape statistics / 中国医疗器械杂志

This paper presents a new deformable model using both population-based and patient-specific shape statistics to segment lung fields from serial chest radiographs. First, a modified scale-invariant feature transform (SIFT) local descriptor is used to characterize the image features in the vicinity of each pixel, so that the deformable model deforms in a way that seeks for the region with similar SIFT local descriptors; second, the deformable model is constrained by both population-based and patient-specific shape statistics. At first, population-based shape statistics plays an leading role when the number of serial images is small, and gradually, patient-specific shape statistics plays a more and more important role after a sufficient number of segmentation results on the same patient have been obtained. The proposed deformable model can adapt to the shape variability of different patients, and obtain more robust and accurate segmentation results.

Non-linear registration of MR brain images integrated with machine learning / 中国医疗器械杂志

This paper presents a machine learning method to select best geometric features for deformable brain registration for each brain location. By incorporating those learned best attribute vector into the framework of HAMMER registration algorithm, The accuracy has increased by about 10% in estimating the simulated deformation fields. At the same time, on real MR brain images, we have found a great deal of improvement of registration in cortical regions.

An approach to adaptive stereo brain image's segmentation / 中国医疗器械杂志

This paper presents a new method for automatically segmenting brain parenchyma and cerebrospinal fluid in routine single-echo MR images. This method is based on the coupled Markov models. They can model intensity measurement at each voxel site to implement piecewise smoothness constraint, and at the same time, model discontinuities to control the interaction between each pair of the neighboring voxel. The method is to derive the maximum a posteriori estimate of the regions and the boundaries by using Bayesian inference and neighborhood constraints based on Markov random fields (MRFs) models. This method has the following desirable properties: (1) the brain image can be well classified into white matter, grey matter and cerebrospinal fluid (CSF), and (2) it has a better robustness to noise and intensity inhomogeneity.

Multi-target extraction algorithm based on edge restriction and attraction field regularization / 中国医疗器械杂志

The topologically-adaptable model is an effective method for the contour detection of multiple objects on an image. However, it meets many problems when we apply it to MR brain images, such as poor convergence to boundary concavities, resulting from the broken boundary, and miserable anti-noise ability. In this paper, we proposes a new algorithm, named multi-target extraction algorithm based on edge restriction and attraction field regularization, to overcome these shortcomings. This new algorithm uses prior knowledge about target to perform edge restriction to get the only edge of the object of interest and to regularize attraction field to enlarge attraction field. Results show that the new algorithm can extract the target contour quickly and accurately when we apply it in MR brain images.