Volumetric analysis of pelvic hematomas after blunt trauma using semi-automated seeded region growing segmentation: a method validation study

Objective

Manually segmented traumatic pelvic hematoma volumes are strongly predictive of active bleeding at conventional angiography, but the method is time intensive, limiting its clinical applicability. We compared volumetric analysis using semi-automated region growing segmentation to manual segmentation and diameter-based size estimates in patients with pelvic hematomas after blunt pelvic trauma.

Materials and methods

A 14-patient cohort was selected in an anonymous randomized fashion from a dataset of patients with pelvic binders at MDCT, collected retrospectively as part of a HIPAA-compliant IRB-approved study from January 2008 to December 2013. To evaluate intermethod differences, one reader (R1) performed three volume measurements using the manual technique and three volume measurements using the semi-automated technique. To evaluate interobserver differences for semi-automated segmentation, a second reader (R2) performed three semi-automated measurements. One-way analysis of variance was used to compare differences in mean volumes. Time effort was also compared. Correlation between the two methods as well as two shorthand appraisals (greatest diameter, and the ABC/2 method for estimating ellipsoid volumes) was assessed with Spearman’s rho (r).

Results

Intraobserver variability was lower for semi-automated compared to manual segmentation, with standard deviations ranging between ±5–32 mL and ±17–84 mL, respectively (p = 0.0003). There was no significant difference in mean volumes between the two readers’ semi-automated measurements (p = 0.83); however, means were lower for the semi-automated compared with the manual technique (manual: mean and SD 309.6 ± 139 mL; R1 semi-auto: 229.6 ± 88.2 mL, p = 0.004; R2 semi-auto: 243.79 ± 99.7 mL, p = 0.021). Despite differences in means, the correlation between the two methods was very strong and highly significant (r = 0.91, p < 0.001). Correlations with diameter-based methods were only moderate and nonsignificant. Mean semi-automated segmentation time effort was 2 min and 6 s and 2 min and 35 s for R1 and R2, respectively, vs. 22 min and 8 s for manual segmentation.

Conclusion

Semi-automated pelvic hematoma volumes correlate strongly with manually segmented volumes. Since semi-automated segmentation can be performed reliably and efficiently, volumetric analysis of traumatic pelvic hematomas is potentially valuable at the point-of-care.

     

A two-stage rule-constrained seedless region growing approach for mandibular body segmentation in MRI

Purpose Extraction of the mandible from 3D volumetric images is frequently required for surgical planning and evaluation. Image segmentation from MRI is more complex than CT due to lower bony signal-to-noise. An automated method to extract the human mandible body shape from magnetic resonance (MR) images of the head was developed and tested. Methods Anonymous MR images data sets of the head from 12 subjects were subjected to a two-stage rule-constrained region growing approach to derive the shape of the body of the human mandible. An initial thresholding technique was applied followed by a 3D seedless region growing algorithm to detect a large portion of the trabecular bone (TB) regions of the mandible. This stage is followed with a rule-constrained 2D segmentation of each MR axial slice to merge the remaining portions of the TB regions with lower intensity levels. The two-stage approach was replicated to detect the cortical bone (CB) regions of the mandibular body. The TB and CB regions detected from the preceding steps were merged and subjected to a series of morphological processes for completion of the mandibular body region definition. Comparisons of the accuracy of segmentation between the two-stage approach, conventional region growing method, 3D level set method, and manual segmentation were made with Jaccard index, Dice index, and mean surface distance (MSD). Results The mean accuracy of the proposed method is $0.958 \,\pm \, 0.020$ for Jaccard index, $0.979 \,\pm \, 0.011$ for Dice index, and $0.204 \,\pm \, 0.127$  mm for MSD. The mean accuracy of CRG is $0.782 \,\pm \, 0.080$ for Jaccard index, $0.876 \,\pm \, 0.053$ for Dice index, and $0.417 \,\pm \, 0.073$  mm for MSD. The mean accuracy of the 3D level set method is $0.874 \,\pm \, 0.0.051$ for Jaccard index, $0.645 \pm 0.306$ for Dice index, and $0.645 \pm 0.306$  mm for MSD. The proposed method shows improvement in accuracy over CRG and 3D level set. Conclusion Accurate segmentation of the body of the human mandible from MR images is achieved with the proposed two-stage rule-constrained seedless region growing approach. The accuracy achieved with the two-stage approach is higher than CRG and 3D level set.     

Nasopharyngeal carcinoma in young patients

Nasopharyngeal carcinoma (NPC) is rare in under 20-year-olds. Early diagnosis greatly improves survival. A retrospective review of 12 young NPC patients (seven males, five females, mean age 16 years) was performed to identify differing patient and tumour characteristics from adult NPC. Seventy-five per cent presented with neck lumps and 25% with headache. None had a family history of NPC or epistaxis. One patient had early stage NPC, and 11 had late stage NPC. Three late stage patients who received chemoradiotherapy had better clinical outcomes than six late stage patients receiving only radiotherapy. There were six deaths, five bony recurrences and one postnasal space recurrence. The 11 late stage patients' five-year actuarial survival was only 29%. Headache is an important symptom for young NPC. Late stage presentation and distant recurrences are also more common, supporting an increased role of chemotherapy.     

Consistent segmentation using a Rician classifier

Several popular classification algorithms used to segment magnetic resonance brain images assume that the image intensities, or log-transformed intensities, satisfy a finite Gaussian mixture model. In these methods, the parameters of the mixture model are estimated and the posterior probabilities for each tissue class are used directly as soft segmentations or combined to form a hard segmentation. It is suggested and shown in this paper that a Rician mixture model fits the observed data better than a Gaussian model. Accordingly, a Rician mixture model is formulated and used within an expectation maximization (EM) framework to yield a new tissue classification algorithm called Rician Classifier using EM (RiCE). It is shown using both simulated and real data that RiCE yields comparable or better performance to that of algorithms based on the finite Gaussian mixture model. As well, we show that RiCE yields more consistent segmentation results when used on images of the same individual acquired with different T1-weighted pulse sequences. Therefore, RiCE has the potential to stabilize segmentation results in brain studies involving heterogeneous acquisition sources as is typically found in both multi-center and longitudinal studies.     

Pancreas segmentation using a dual-input v-mesh network

Accurate segmentation of the pancreas from abdomen scans is crucial for the diagnosis and treatment of pancreatic diseases. However, the pancreas is a small, soft and elastic abdominal organ with high anatomical variability and has a low tissue contrast in computed tomography (CT) scans, which makes segmentation tasks challenging. To address this challenge, we propose a dual-input v-mesh fully convolutional network (FCN) to segment the pancreas in abdominal CT images. Specifically, dual inputs, i.e., original CT scans and images processed by a contrast-specific graph-based visual saliency (GBVS) algorithm, are simultaneously sent to the network to improve the contrast of the pancreas and other soft tissues. To further enhance the ability to learn context information and extract distinct features, a v-mesh FCN with an attention mechanism is initially utilized. In addition, we propose a spatial transformation and fusion (SF) module to better capture the geometric information of the pancreas and facilitate feature map fusion. We compare the performance of our method with several baseline and state-of-the-art methods on the publicly available NIH dataset. The comparison results show that our proposed dual-input v-mesh FCN model outperforms previous methods in terms of the Dice similarity coefficient (DSC), positive predictive value (PPV), sensitivity (SEN), average surface distance (ASD) and Hausdorff distance (HD). Moreover, ablation studies show that our proposed modules/structures are critical for effective pancreas segmentation.     

Screening serum biomarkers for early primary hepatocellular carcinoma using a phage display technique

Hepatocellular carcinoma (HCC) occurs mainly in chronically diseased livers following hepatitis B virus (HBV) or hepatitis C virus (HCV) infection. Early detection and diagnosis of HCC would be of great clinical benefit. In this study, we used a random phage display peptide library and sera from early-stage primary HCC patients (n = 30) to screen potential serum biomarkers for early primary HCC. Age- and sex-matched patients with HBV and/or HCV infection were used as controls. In the screening phase, 19 out of 20 randomly selected phage clones exhibited specific reaction with purified sera IgG from early primary HCC patients, among them 14 coming from the same phage clone with inserted peptidesequence RGWCRPLPKGEG (named HC1). In the validation phase, phage ELISA results showed that the positive reaction rate of the HC1 phage clone was 91.4% with the early HCC group (n = 70), significantly higher than that with the HBV infection group (20.0%) (n = 70), the HCV infection group (12.9%) (n = 70), the HBV + HCV infection group (24.3%) (n = 70), the cirrhosis group (17.1%) (n = 70), and the healthy control group (10.0%) (n = 70). In conclusion, the HC1 mimic peptide showed high diagnostic validity for early primary HCC, and thereby could be a candidate serum biomarker for early primary HCC.     

Unsupervised identification of white matter tracts in a mouse brain using a directional correlation-based region growing (DCRG) algorithm

The inner product of the major eigenvectors of adjacent pixels, known as the directional correlation (DC), has been used previously as a quantitative index to investigate directional similarity in white matter (WM) tracts. A high degree of directional similarity (i.e., high DC) among pixels within individual WM tracts was observed. Based on this observation, a region growing algorithm was employed to propagate an area from a seed point as a function of the DC threshold (DCt) to manually identify WM tracts in two-dimensional (2D) slices from diffusion tensor imaging (DTI). In the present study, an improved unsupervised DC based region growing (DCRG) method was implemented to reduce the operator-dependent variance and to improve the ease of use of the technique. By employing improved method, a multi-slice DTI data set of an in vivo mouse brain was used to identify the external capsule, visual pathway, and corpus callosum. The resultant WM tracts are computer rendered in three-dimensional (3D) images with anatomical images as structural references. In addition, three sets of ex vivo mouse brain data were used to examine the effects of different slice thickness and the signal-to-noise ratio (SNR) to the outcome of DCRG.     

Segmentation of building roofs from airborne LiDAR point clouds using robust voxel-based region growing

The research herein presents a new approach for extracting building roofs using a robust voxel-based region growing segmentation method. The proposed approach exploits the fact that the roof of the building consists of planar surfaces and has distinctive geometric features than other kinds of objects. Based on this assumption, we present a method using voxel structure and region growing strategy with robust principal component analysis (RPCA). The voxels is clustered by a region growing process, utilizing the smoothness, continuity, and convexity as geometric cues. RPCA is introduced to estimate the attribute of voxels. Roofs are recognized from the segments by using the object-based spectral clustering. Our approach has been validated by different airborne laser scanning (ALS) point clouds. Qualitative and quantitative results reveal that our method outperforms some representative algorithms in segmentation using our testing datasets under a complex situation, with overall quality measure better than 0.7 and 0.6.     

Speckle reduction with edge preservation in medical ultrasonic images using a homogeneous region growing mean filter (HRGMF)

In this paper, we propose a new post formation adaptive image filtering technique, to be called the homogeneous region growing mean filter, in order to reduce speckle noise with edge preservation in medical ultrasonic images. First, to find a proper seed region, an initially assumed seed region, which is larger than the average speckle size around a given filtering point, is successively contracted according to a certain local homogeneity criterion. Once the seed region is determined, the next step is to grow the homogeneous region successively based on some measures of local homogeneity and similarity of the neighboring region. The output of the proposed filter for each filtering point is obtained from the arithmetic mean of the grown locally homogeneous region. Several simulation results are presented to illustrate the performance of the proposed filter. They show that the proposed technique effectively smoothes ultrasonic speckle and completely suppresses isolated impulsive noise over the entire texture in addition to preserving the edge information.     

Peer review using a paired-comparison technique

A paired-comparison technique is used to enable surgical residents and attending surgeons to make peer judgments of each other. Comparative peer judgments were made in three areas: the ability of the surgeon to make a diagnosis and to decide on a plan of active medical care, the operating ability of the surgeon, and the postoperative care of the patient. This method of peer judgment ensures the confidentiality of those making the judgments, and the analysis results in a final ranking of the surgeons and surgical residents with indicated significant differences (P less than 0.05) among them. The technique also includes an assessment of how consistent each judge is in making comparisons of fellow surgeons.     

Region segmentation using information divergence measures

Image segmentations based on maximum likelihood or maximum a posteriori analyses of object textures usually assume parametric models (e.g., Gaussian) for distributions of these features. For real images, parameter accuracy and model stationarity may be elusive, so that model-free inference methods ought to have an advantage over those that are model-dependent. Functions of the relative entropy (RE) from information theory can produce minimum error, model-free inferences, and can detect the boundary of an image object by maximizing the RE between the pixel distributions inside and outside a flexible curve contour. A generalization of the RE -- the Jensen-Rényi divergence (JRD) -- computes optimal n-way decisions and can contour multiple objects in an image simultaneously. Seed regions expand naturally and multiple contours tend not to overlap. An edge detector based on the JRD, combined with multivariate pixel segmentation, generally improved the error of the segmentation. We apply these functions to contour patient anatomy in X-ray computed tomography for radiotherapy treatment planning.     

Ultrasound image segmentation using spectral clustering

Segmentation of ultrasound images is necessary in a variety of clinical applications, but the development of automatic techniques is still an open problem. Spectral clustering techniques have recently become popular for data and image analysis. In particular, image segmentation has been proposed via the normalized cut (NCut) criterion. This article describes an initial investigation to determine the suitability of such segmentation techniques for ultrasound images. The adaptation of the NCut technique to ultrasound is described first. Segmentation is then performed on simulated ultrasound images. Tests are also performed on abdominal and fetal images with the segmentation results compared to manual segmentation. The success of the segmentation on these test cases warrants further research into NCut-based segmentation of ultrasound images.     

BEaST: brain extraction based on nonlocal segmentation technique

Brain extraction is an important step in the analysis of brain images. The variability in brain morphology and the difference in intensity characteristics due to imaging sequences make the development of a general purpose brain extraction algorithm challenging. To address this issue, we propose a new robust method (BEaST) dedicated to produce consistent and accurate brain extraction. This method is based on nonlocal segmentation embedded in a multi-resolution framework. A library of 80 priors is semi-automatically constructed from the NIH-sponsored MRI study of normal brain development, the International Consortium for Brain Mapping, and the Alzheimer's Disease Neuroimaging Initiative databases. In testing, a mean Dice similarity coefficient of 0.9834±0.0053 was obtained when performing leave-one-out cross validation selecting only 20 priors from the library. Validation using the online Segmentation Validation Engine resulted in a top ranking position with a mean Dice coefficient of 0.9781±0.0047. Robustness of BEaST is demonstrated on all baseline ADNI data, resulting in a very low failure rate. The segmentation accuracy of the method is better than two widely used publicly available methods and recent state-of-the-art hybrid approaches. BEaST provides results comparable to a recent label fusion approach, while being 40 times faster and requiring a much smaller library of priors.     

鼻咽癌囊性肺转移2例

病例1:男,51岁,因活动时呼吸困难,咳嗽2月余,伴声嘶、咯血丝痰1个月入院.     

Intraductal growing acinar cell carcinoma of the pancreas

Acinar cell carcinoma (ACC) is a rare pancreatic exocrine neoplasm characterized by a huge, exophytic well-circumscribed hypovascular mass. There has been several reports describing intraductal and papillary variant of ACC and they showed different radiologic features from usual ACC. We present histologically confirmed cases of intraductal and papillary variant of ACC that had been found in two patients, who underwent CT and MRI. This report provides CT and MRI features of intraductal and papillary variant of ACC in pancreas with pathologic correlation after surgical excision.