Hong Song

Kidney segmentation in CT sequences using SKFCM and improved GrowCut algorithm

BackgroundOrgan segmentation is an important step in computer-aided diagnosis and pathology detection. Accurate kidney segmentation in abdominal computed tomography (CT) sequences is an essential and crucial task for surgical planning and navigation in kidney tumor ablation. However, kidney segmentation in CT is a substantially challenging work because the intensity values of kidney parenchyma are similar to those of adjacent structures.ResultsIn this paper, a coarse-to-fine method was applied to segment kidney from CT images, which consists two stages including rough segmentation and refined segmentation. The rough segmentation is based on a kernel fuzzy C-means algorithm with spatial information (SKFCM) algorithm and the refined segmentation is implemented with improved GrowCut (IGC) algorithm. The SKFCM algorithm introduces a kernel function and spatial constraint into fuzzy c-means clustering (FCM) algorithm. The IGC algorithm makes good use of the continuity of CT sequences in space which can automatically generate the seed labels and improve the efficiency of segmentation. The experimental results performed on the whole dataset of abdominal CT images have shown that the proposed method is accurate and efficient. The method provides a sensitivity of 95.46% with specificity of 99.82% and performs better than other related methods.ConclusionsOur method achieves high accuracy in kidney segmentation and considerably reduces the time and labor required for contour delineation. In addition, the method can be expanded to 3D segmentation directly without modification.

Liver segmentation based on SKFCM and improved GrowCut for CT images

Accurate liver segmentation is an essential and crucial step for computer-aided liver disease diagnosis and surgical planning. In this paper, a new coarse-to-fine method is proposed to segment liver for abdominal computed tomography (CT) images. This hierarchical framework consists of rough segmentation and refined segmentation. The rough segmentation is implemented based on a kernel fuzzy C-means algorithm with spatial information (SKFCM) algorithm and the refined segmentation is performed based on the proposed improved GrowCut (IGC) algorithm. The SKFCM algorithm introduces a kernel function and spatial constraint based on fuzzy c-means clustering (FCM) algorithm, which can reduce the effect of noise and improve the clustering ability. The IGC algorithm makes good use of the continuity of CT series in space which can automatically generate the seed labels and improve the efficiency of segmentation. The proposed method was applied to segment the liver for the whole dataset of abdominal CT images. The performance evaluation of segmentation results shows that the proposed liver segmentation method is accurate and efficient. Experimental results have been shown visually and achieve reasonable consistency.

Registration and fusion quantification of augmented reality based nasal endoscopic surgery

HighlightsThe impacts of registration and fusion display errors on the navigation accuracy are quantitively studied on a custom ARNES.Accuracy level of a calibrated endoscope is redefined by using a calibration tool with improved structural reliability.A point cloud‐based hybrid tracking method is proposed to rapidly shorten the duration of intraoperative recalibration.The dynamic endoscopic vision expansion, hierarchical rendering, is validated thoroughly in model studies and clinical trials. ABSTRACT This paper quantifies the registration and fusion display errors of augmented reality‐based nasal endoscopic surgery (ARNES). We comparatively investigated the spatial calibration process for front‐end endoscopy and redefined the accuracy level of a calibrated endoscope by using a calibration tool with improved structural reliability. We also studied how registration accuracy was combined with the number and distribution of the deployed fiducial points (FPs) for positioning and the measured registration time. A physically integrated ARNES prototype was customarily configured for performance evaluation in skull base tumor resection surgery with an innovative approach of dynamic endoscopic vision expansion. As advised by surgical experts in otolaryngology, we proposed a hierarchical rendering scheme to properly adapt the fused images with the required visual sensation. By constraining the rendered sight in a known depth and radius, the visual focus of the surgeon can be induced only on the anticipated critical anatomies and vessel structures to avoid misguidance. Furthermore, error analysis was conducted to examine the feasibility of hybrid optical tracking based on point cloud, which was proposed in our previous work as an in‐surgery registration solution. Measured results indicated that the error of target registration for ARNES can be reduced to 0.77 ± 0.07 mm. For initial registration, our results suggest that a trade‐off for a new minimal time of registration can be reached when the distribution of five FPs is considered. For in‐surgery registration, our findings reveal that the intrinsic registration error is a major cause of performance loss. Rigid model and cadaver experiments confirmed that the scenic integration and display fluency of ARNES are smooth, as demonstrated by three clinical trials that surpassed practicality. Graphical abstract Figure. No caption available.

Automatic schizophrenia discrimination on fNIRS by using PCA and SVM

A method is proposed to distinguish patients with schizophrenia from healthy controls based on data measured by functional near-infrared spectroscopy (fNIRS) during a cognitive task, which combines principal component analysis (PCA) and support vector machine (SVM). Firstly, a data reduction technique is applied prior to PCA, and then PCA is used to extract features on oxygenated hemoglobin (oxy-Hb) signals from 52-channel fNIRS data of schizophrenia and healthy subjects. Secondly, a classifier based on SVM is designed to discriminate schizophrenia from healthy controls. We recruited a large sample of 52 schizophrenia patients and 38 healthy controls. The hemoglobin response was measured in the prefrontal cortex during the one-back memory task using a 52-channel fNIRS system. The experimental results indicate that the proposed method can achieve a satisfactory classification with the accuracy of 93.33%, 100% for schizophrenia samples and 84.62% for healthy controls. Also, our results suggested that fNIRS has the potential capacity to be an effective objective biomarker for the diagnosis of schizophrenia.

Global Patch Matching (GPM) for freehand 3D ultrasound reconstruction

Background3D ultrasound volume reconstruction from B-model ultrasound slices can provide more clearly and intuitive structure of tissue and lesion for the clinician.MethodsThis paper proposes a novel Global Path Matching method for the 3D reconstruction of freehand ultrasound images. The proposed method composes of two main steps: bin-filling scheme and hole-filling strategy. For the bin-filling scheme, this study introduces two operators, including the median absolute deviation and the inter-quartile range absolute deviation, to calculate the invariant features of each voxel in the 3D ultrasound volume. And the best contribution range for each voxel is obtained by calculating the Euclidian distance between current voxel and the voxel with the minimum invariant features. Hence, the intensity of the filling vacant voxel can be obtained by weighted combination of the intensity distribution of pixels in the best contribution range. For the hole-filling strategy, three conditions, including the confidence term, the data term and the gradient term, are designed to calculate the weighting coefficient of the matching patch of the vacant voxel. While the matching patch is obtained by finding patches with the best similarity measure that defined by the three conditions in the whole 3D volume data.ResultsCompared with VNN, PNN, DW, FMM, BI and KR methods, the proposed Global Path Matching method can restore the 3D ultrasound volume with minimum difference.ConclusionsExperimental results on phantom and clinical data sets demonstrate the effectiveness and robustness of the proposed method for the reconstruction of ultrasound volume.

Sparse intervertebral fence composition for 3D cervical vertebra segmentation

Statistical shape models are capable of extracting shape prior information, and are usually utilized to assist the task of segmentation of medical images. However, such models require large training datasets in the case of multi-object structures, and it also is difficult to achieve satisfactory results for complex shapes. This study proposed a novel statistical model for cervical vertebra segmentation, called sparse intervertebral fence composition (SiFC), which can reconstruct the boundary between adjacent vertebrae by modeling intervertebral fences. The complex shape of the cervical spine is replaced by a simple intervertebral fence, which considerably reduces the difficulty of cervical segmentation. The final segmentation results are obtained by using a 3D active contour deformation model without shape constraint, which substantially enhances the recognition capability of the proposed method for objects with complex shapes. The proposed segmentation framework is tested on a dataset with CT images from 20 patients. A quantitative comparison against corresponding reference vertebral segmentation yields an overall mean absolute surface distance of 0.70 mm and a dice similarity index of 95.47% for cervical vertebral segmentation. The experimental results show that the SiFC method achieves competitive cervical vertebral segmentation performances, and completely eliminates inter-process overlap.

Local statistical deformation models for deformable image registration

Abstract A fast and robust image registration algorithm for high-dimensional brain Magnetic Resonance images was developed based on the statistical deformation models (SDMs). This model learns deformation fields and achieves fast and robust registration by greatly reducing transformation dimensionality. This model is trained via principal component analysis (PCA), which suffers from large transformation dimensionality and small samples. For the high-dimensional image registration, the dimensions of the deformation fields are huge, the basic functions computed from PCA cannot represent deformation fields well. Therefore, we proposed a local SDM (LSDM) in this paper to solve the aforementioned problems. We divided the images into several small parts, in which the dimensions of the deformation fields are greatly reduced. Then, we trained the LSDM using the deformation fields between sample images and a selected template by applying PCA in each small part. Given that the dimension of eigenvectors of LSDM decreases much more than that of SDM, the orthonormal basis functions of LSDM represent the deformation fields more accurately than those of SDM. We obtained the total deformation fields for warping the image by integrating the deformation fields of all LSDMs. Using the manually labeled MR images of different people, we demonstrated that LSDM could greatly reduce the image registration time while maintaining favorable registration accuracy.

Automatic retinal vessel segmentation using multi-scale superpixel chain tracking

Abstract The segmentation of retinal vessel and its structure information are important for computer-aided diagnosis and treatment of many diseases. This work proposes a superpixel-based chain tracking method for segmentation of retinal vessels. First, a multi-scale superpixel segmentation framework is developed to split the image into patches, which are utilized as the basic unit of the vessel-tracking procedure. Second, a vessel chain model which consists of a series of superpixel nodes is proposed for accurately segmenting small vessels. Third, vessel tracking is achieved by a two-stage procedure where vessel regions with good and bad imaging quality are handled differently. Finally, a maximum gradient method is proposed to estimate the vessel centerline and boundary. The proposed method was validated on synthetic data and public retinal image datasets. Experimental results demonstrate that the proposed method can accurately track the vascular skeletons, and the tracking accuracy can reach 0.9636.

Hybrid constraint optimization for 3D subcutaneous vein reconstruction by near-infrared images

BACKGROUND AND OBJECTIVE The development of biometric identification technology and intelligent medication has enabled researchers to analyze subcutaneous veins from near-infrared images. However, the stereo reconstruction of subcutaneous veins has not been well studied, and the results are difficult to utilize in clinical practice. METHODS We present a hybrid constraint optimization (HCO) matching algorithm for vein reconstruction to solve the matching failure problems caused by the incomplete segmentation of vein structures captured from different views. This algorithm initially introduces the existence of the epipolar and homography constraints in the subcutaneous vein matching. Then, the HCO matching algorithm of the vascular centerline is established by homography point-to-point matching, homography matrix optimization, and vascular section matching. Finally, the 3D subcutaneous vein is reconstructed on the basis of the principle of triangulation and system calibration parameters. RESULTS To validate the performance of the proposed matching method, we designed a series of experiments to evaluate the effectiveness of the hybrid constraint optimization method. The experiments were performed on simulated and real datasets. 42 real vascular images were analyzed on the basis of different matching strategies. Experimental result shows that the matching accuracy increased significantly with the proposed optimization matching method. To calculate the reconstruction accuracy, we reconstructed seven simulated cardboards and measured 10 vascular distances in each simulated cardboard. The average vascular distance error of each simulated image was within 1.0 mm, and the distance errors of 75% feature points were less than 1.5 mm. Also, we printed a 3D simulated vein model to improve the illustration of this system. The reconstruction error extends from -3.58 mm to 1.94 mm with a standard deviation of 0.68 mm and a mean of 0.07 mm. CONCLUSIONS The algorithm is validated in terms of homography optimization, matching efficiency, and simulated vascular reconstruction error. The experimental results demonstrate that the veins captured from the left and right views can be accurately matched through the proposed algorithm.

Sparse deformation prediction using Markove Decision Processes (MDP) for Non-rigid registration of MR image

BACKGROUND AND OBJECTIVE A framework of sparse deformation prediction using Markove Decision Processes is proposed for achieving a rapid and accurate registration by providing a suitable initial deformation. METHODS In the proposed framework, the tree is built based on the training set for each patch from the template image. The template patch is considered as the root. The node is the patch group in which multiple similar patches are extracted around a key point on the training image. Given the linkages between patch groups in the tree, MDP is introduced to select the optimal path with highest registration accuracy from each training patch to the template patch. The deformation between them is estimated along the selected path by patch-wise registration which can be realized by a non-learning-based method. Given the patches on a testing image, their best matching patches are fast chosen from the training patches and the corresponding deformations constitute a sparse deformation. A dense deformation for the entire test image is subsequently interpolated and used as an initial deformation for further registration. RESULTS With the non-learning-based registration as the baseline method, the proposed framework is evaluated using three datasets of inter-subject brain MR images with three learning-based methods. Experimental results of the non-learning-based method using the proposed framework reveal that the computation time is reduced by fivefold after using the proposed framework. And, with the same baseline method, the proposed framework demonstrates the higher accuracy than three learning-based methods which predicts the initial deformation at image scale. The mean Dice of three datasets for the tissues of the brain are 73.52%, 70.73% and 64.82%, respectively. CONCLUSIONS The proposed framework rapidly registers the inter-subject brains and achieves the high mean Dice for the tissues of the brain.