Semi-automatic Segmentation Method Research Articles

Accuracy and efficiency of computer-aided anatomical analysis using 3D visualization software based on semi-automated and automated segmentations

ObjectiveWe investigated and compared the functionality of two 3D visualization software provided by a CT vendor and a third-party vendor, respectively. Using surgical anatomical measurement as baseline, we evaluated the accuracy of 3D visualization and verified their utility in computer-aided anatomical analysis. MethodsThe study cohort consisted of 50 adult cadavers fixed with the classical formaldehyde method. The computer-aided anatomical analysis was based on CT images (in DICOM format) acquired by helical scan with contrast enhancement, using a CT vendor provided 3D visualization workstation (Syngo) and a third-party 3D visualization software (Mimics) that was installed on a PC. Automated and semi-automated segmentations were utilized in the 3D visualization workstation and software, respectively. The functionality and efficiency of automated and semi-automated segmentation methods were compared. Using surgical anatomical measurement as a baseline, the accuracy of 3D visualization based on automated and semi-automated segmentations was quantitatively compared. ResultsIn semi-automated segmentation, the Mimics 3D visualization software outperformed the Syngo 3D visualization workstation. No significant difference was observed in anatomical data measurement by the Syngo 3D visualization workstation and the Mimics 3D visualization software (P>0.05). ConclusionsBoth the Syngo 3D visualization workstation provided by a CT vendor and the Mimics 3D visualization software by a third-party vendor possessed the needed functionality, efficiency and accuracy for computer-aided anatomical analysis.

Computer-Assisted Analysis of Bone Volume for Sinus Augmentation Procedure

Sinus floor augmentation | Cone beam computed tomography | Graft volume measurement | Computerassisted image interpretation | Surgical simulation | Greffe sinusienne | Tomodensitométrie à faisceau conique | Simulation de greffe osseuse | Mesure volumétrique

A novel diffusion-tensor MRI approach for skeletal muscle fascicle length measurements.

Musculoskeletal (dys‐)function relies for a large part on muscle architecture which can be obtained using Diffusion‐Tensor MRI (DT‐MRI) and fiber tractography. However, reconstructed tracts often continue along the tendon or aponeurosis when using conventional methods, thus overestimating fascicle lengths. In this study, we propose a new method for semiautomatic segmentation of tendinous tissue using tract density (TD). We investigated the feasibility and repeatability of this method to quantify the mean fascicle length per muscle. Additionally, we examined whether the method facilitates measuring changes in fascicle length of lower leg muscles with different foot positions. Five healthy subjects underwent two DT‐MRI scans of the right lower leg, with the foot in 15° dorsiflexion, neutral, and 30° plantarflexion positions. Repeatability of fascicle length measurements was assessed using Bland–Altman analysis. Changes in fascicle lengths between the foot positions were tested using a repeated multivariate analysis of variance (MANOVA). Bland–Altman analysis showed good agreement between repeated measurements. The coefficients of variation in neutral position were 8.3, 16.7, 11.2, and 10.4% for soleus (SOL), fibularis longus (FL), extensor digitorum longus (EDL), and tibialis anterior (TA), respectively. The plantarflexors (SOL and FL) showed significant increase in fascicle length from plantarflexion to dorsiflexion, whereas the dorsiflexors (EDL and TA) exhibited a significant decrease. The use of a tract density for semiautomatic segmentation of tendinous structures provides more accurate estimates of the mean fascicle length than traditional fiber tractography methods. The method shows moderate to good repeatability and allows for quantification of changes in fascicle lengths due to passive stretch.

Assessment of local pulse wave velocity distribution in mice using k-t BLAST PC-CMR with semi-automatic area segmentation

BackgroundLocal aortic pulse wave velocity (PWV) is a measure for vascular stiffness and has a predictive value for cardiovascular events. Ultra high field CMR scanners allow the quantification of local PWV in mice, however these systems are yet unable to monitor the distribution of local elasticities.MethodsIn the present study we provide a new accelerated method to quantify local aortic PWV in mice with phase-contrast cardiovascular magnetic resonance imaging (PC-CMR) at 17.6 T. Based on a k-t BLAST (Broad-use Linear Acquisition Speed-up Technique) undersampling scheme, total measurement time could be reduced by a factor of 6. The fast data acquisition enables to quantify the local PWV at several locations along the aortic blood vessel based on the evaluation of local temporal changes in blood flow and vessel cross sectional area. To speed up post processing and to eliminate operator bias, we introduce a new semi-automatic segmentation algorithm to quantify cross-sectional areas of the aortic vessel. The new methods were applied in 10 eight-month-old mice (4 C57BL/6J-mice and 6 ApoE (−/−)-mice) at 12 adjacent locations along the abdominal aorta.ResultsAccelerated data acquisition and semi-automatic post-processing delivered reliable measures for the local PWV, similiar to those obtained with full data sampling and manual segmentation. No statistically significant differences of the mean values could be detected for the different measurement approaches. Mean PWV values were elevated for the ApoE (−/−)-group compared to the C57BL/6J-group (3.5 ± 0.7 m/s vs. 2.2 ± 0.4 m/s, p < 0.01). A more heterogeneous PWV-distribution in the ApoE (−/−)-animals could be observed compared to the C57BL/6J-mice, representing the local character of lesion development in atherosclerosis.ConclusionIn the present work, we showed that k-t BLAST PC-MRI enables the measurement of the local PWV distribution in the mouse aorta. The semi-automatic segmentation method based on PC-CMR data allowed rapid determination of local PWV. The findings of this study demonstrate the ability of the proposed methods to non-invasively quantify the spatial variations in local PWV along the aorta of ApoE (−/−)-mice as a relevant model of atherosclerosis.

Intra- and inter-reader reproducibility of blood flow measurements on the ascending aorta and pulmonary artery using cardiac magnetic resonance.

The aim of our study was to estimate the intra- and inter-reader reproducibility of blood flow measurements in the ascending aorta and main pulmonary artery using cardiac magnetic resonance (CMR) and a semi-automated segmentation method. The ethics committee approved this retrospective study. A total of 50 consecutive patients (35 males and 15 females; mean age±standard deviation 27±13 years) affected by congenital heart disease were reviewed. They underwent CMR for flow analysis of the ascending aorta and main pulmonary artery (1.5 T, through-plane phase-contrast sequences). Two independent readers (R1, trained radiology resident; R2, lower-trained technician student) obtained segmented images twice (>10-day interval), using a semi-automated method of segmentation. Peak velocity, forward and backward flows were obtained. Bland-Altman analysis was used and reproducibility was reported as complement to 100% of the ratio between the coefficient of repeatability and the mean. R1 intra-reader reproducibility for the aorta was 99% (peak velocity), 95% (forward flow) and 49% (backward flow); for the pulmonary artery, 99%, 91% and 90%, respectively. R2 intra-reader reproducibility was 92%, 91% and 38%; 98%, 86% and 87%, respectively. Inter-reader reproducibility for the aorta was 91%, 85% and 20%; for the pulmonary artery 96%, 75%, and 82%, respectively. Our results showed a good to excellent reproducibility of blood flow measurements of CMR together with a semiautomated method of segmentation, for all variables except the backward flow of the ascending aorta, with a limited impact of operator's training.

A semiautomatic method for rapid segmentation of velocity-encoded myocardial magnetic resonance imaging data.

To develop a semiautomatic method for rapid segmentation of myocardial tissue phase mapping (TPM) data. Manual segmentation of the myocardium was performed at end-diastole and end-systole. The points in both user-defined masks were then automatically tracked over the entire cardiac cycle using temporal integration of the velocity field. Paths that failed to visit both masks at the expected times were excluded, after which masks for all time points were generated automatically from the accepted paths. Midventricular and basal phase contrast TPM slices from 12 rats were segmented using the proposed method and fully manual segmentation. The results were compared using Dice's metric and Bland-Altman analysis, and interobserver variability was assessed. The semiautomatic method reduced the average user input time from 21 min to 1 min per slice. The Dice metrics between the methods were 0.88 ± 0.03 (midventricular) and 0.83 ± 0.06 (basal), and Bland-Altman limits of agreement of peak systolic and diastolic regional velocities were: midventricular: 0.05 ± 0.65 cm/s, -0.02 ± 0.42 cm/s, and -0.03 ± 0.40 cm/s (radial, tangential, longitudinal); basal: -0.04 ± 0.73 cm/s, 0.03 ± 0.60 cm/s, and -0.04 ± 0.48 cm/s (radial, tangential, longitudinal). Interobserver variability following semiautomatic segmentation was lower than for manual segmentation. The proposed method reduced the segmentation time substantially and exhibited well-preserved data quality and excellent interobserver limits of agreement. Magn Reson Med 78:1199-1207, 2017. © 2016 International Society for Magnetic Resonance in Medicine.

A Local ROI-specific Atlas-based Segmentation of Prostate Gland and Transitional Zone in Diffusion MRI

Segmentation of prostate and related anatomic structure, such as transitional zone, in medical images facilitates prostate cancer detection, as well as a number of other clinical practices. In this paper, we propose a semi-automatic local ROI-specific atlas-based segmentation (LABS) method to segment prostate gland and transitional zone in diffusion magnetic resonance images. Inspired by a sequential registration-based segmentation method, the proposed method further reduces the amount of user intervention and focuses on the vicinity of prostate for atlas matching and atlas-to-target registration by specifying the bounding boxes of prostate gland on key slices of volume images. We evaluated the method on an atlas database with the 100 cases by performing a leave-one-out study. Our proposed method produced favorable outcomes with an average Dice similarity coefficient of 0.85±0.03 for prostate gland and 0.77±0.06 for transitional zone segmentations, which indicates the effectiveness of the proposed method and its potential to be used in relevant clinical applications.

Influence of the volume of soft tissue herniation on clinical symptoms of patients with orbital floor fractures.

One of the most important complications of orbital floor fractures is diplopia and restricted ocular movement. The role of the volume of soft tissue herniation on these clinical symptoms after orbital floor fractures is unclear and potentially may predict development of persistent clinical symptoms. Therefore the aim of this study was to assess pre- and postoperative complications associated to the volume of soft tissue herniation, with special interest to diplopia and bulbus motility impairment. 204 untreated patients with orbital floor fractures from 2009 to 2011 were included in this retrospective study. Contingency tables and χ2-test were performed to analyze associations between two qualitative variables. p-Values p<0.05 were considered as significant. Volume measurement was performed in a semiautomatic segmentation method with the software "ITK-Snap". The volume of soft tissue herniation correlated significantly with pre- and postoperative diplopia (p=0.003; p=0.002), persistent diplopia (p=0.009) as well as pre- and postoperative bulbus motility impairment (both p<0.001). Furthermore we found out significant associations between the volume of soft tissue herniation and pre- and postoperative complications depending on fracture type and reconstruction technique. Volume measurement of soft tissue herniation may help to predict postoperative complications, particularly bulbus motility restriction and persistent diplopia. The risk for these symptoms rises with increasing volume of soft tissue herniation after orbital floor fractures. Therefore we recommend for these patients within indication an early repair and/or closer observation.

Calibrating level set approach by granular computing in computed tomography abdominal organs segmentation

Hybrid methodology that combines granular computing with a level set approach for image segmentation is introduced in this paper. The goal of the study is to provide an adjustable semi-automated method for 3D segmentation of abdominal organs in computed tomography studies. Based on initial guidance in terms of models or seed points the information granule is formulated for each anatomical structure under consideration. The granulation leads via spatial volume resampling and granule-driven image intensity fuzzification to the final segmentation stage employing a hybrid level set approach. Depending on the organ intensity and shape the segmentation runs with selected parameters. The algorithm has been evaluated by 20 computed tomography studies and four various structures delineated by an expert. The experimental part presents a comprehensive analysis of subsequent stages of the algorithm. The voxel-wise segmentation sensitivities reached 94.4% for the right kidney, 93.8% for the left kidney, 94.0% for the spleen, and 94.0% for the liver, with the Dice index at 93.7%, 94.2%, 91.0%, and 92.9%, respectively.

Volumetric Analysis of Cerebral Peduncles and Cerebellar Hemispheres for Predicting Hemiparesis After Hemispherectomy.

In some cases of refractory epilepsy, hemispherectomy is the final invasive treatment option. However, predictors of postoperative hemiparesis in these patients have not been widely studied. To investigate how the volumetric analysis of cerebral peduncles and cerebellar hemispheres in patients who have undergone hemispherectomy may determine prognostic implications for postoperative hemiparesis. Twenty-two patients who underwent hemispherectomy at our institution were retrospectively included. Using iPlan/BrainLAB (BrainLAB, Feldkirchen, Germany) imaging software and a semiautomatic voxel-based segmentation method, we calculated the preoperative cerebral peduncle and cerebellar hemisphere volumes. Cerebral peduncle and cerebellar hemisphere ratios were compared between patients with worsened or unchanged/better hemiparesis postoperatively. The ratios of ipsilateral/contralateral cerebral peduncles (0.570 vs 0.828; P = .02) and contralateral/ipsilateral cerebellar hemispheres (0.885 vs 1.031; P = .009) were significantly lower in patients who had unchanged/improved hemiparesis postoperatively compared with patients who had worsened hemiparesis. Relative risk of worsening hemiparesis was significantly higher in patients with a cerebral peduncle ratio < 0.7 (relative risk, 4.3; P = .03) or a cerebellar ratio < 1.0 (relative risk, 6.4; P = .006). Although patients who undergo hemispherectomy are heterogeneous, we report a method of predicting postoperative hemiparesis using only standard volumetric magnetic resonance imaging. This information could be used in preoperative discussions with patients and families to help better understand that chance of retaining baseline motor function. CST, corticospinal tractfMRI, functional magnetic resonance imagingTMS, transcranial magnetic stimulation.

Issues in accurate PET for target delineation in radiotherapy

Tumor delineation in PET imaging has been one of the main concerns for treatment planning in radiation oncology. However, a standardized way of converting PET signals into target volumes is not yet available. The majority of the literature on delineation methods deals with 18FDG-PET imaging. In this context, semiautomatic and automatic segmentation methods have been developed implying gradient, region growing, clustering, statistical methods, but the majority of these new algorithms are not widely and readily available. Adaptive thresholding methods based on contrast-oriented contouring algorithms have been developed by many groups and validated in patient data for different anatomical regions. Among their strong points easiness of use and low cost must be accounted for, while among the drawbacks heterogeneity in lesion uptake and lesion motion are still open questions. Moreover, quantification and delineation issues become even more troublesome when partial volume effect is considered and corrected for by applying resolution modeling techniques. Literature shows lack of evidence ensuring that resolution modeling algorithms can provide accurate and reproducible data, suggesting that resolution modeling reconstruction algorithms in quantitative 18FDG-PET imaging need to be accurately validated before introduction in clinical routine. Other issues in tumor delineation are concerned with PET imaging performed with radiotracers different from 18FDG, for example 11C-choline and 11C-methionine for prostate and brain tumors imaging, for which the conclusions drawn for 18FDG imaging delineation may not apply and different techniques and strategies must be looked for.

A Semiautomatic Method for Multiple Sclerosis Lesion Segmentation on Dual-Echo MR Imaging: Application in a Multicenter Context

The automatic segmentation of MS lesions could reduce time required for image processing together with inter- and intraoperator variability for research and clinical trials. A multicenter validation of a proposed semiautomatic method for hyperintense MS lesion segmentation on dual-echo MR imaging is presented. The classification technique used is based on a region-growing approach starting from manual lesion identification by an expert observer with a final segmentation-refinement step. The method was validated in a cohort of 52 patients with relapsing-remitting MS, with dual-echo images acquired in 6 different European centers. We found a mathematic expression that made the optimization of the method independent of the need for a training dataset. The automatic segmentation was in good agreement with the manual segmentation (dice similarity coefficient = 0.62 and root mean square error = 2 mL). Assessment of the segmentation errors showed no significant differences in algorithm performance between the different MR scanner manufacturers (P > .05). The method proved to be robust, and no center-specific training of the algorithm was required, offering the possibility for application in a clinical setting. Adoption of the method should lead to improved reliability and less operator time required for image analysis in research and clinical trials in MS.

Automatic Segmentation of Lower Jaw and Mandibular Bone in Digital Dental Panoramic Radiographs

Objectives: Panoramic x-ray is a medical imaging technique used in dentistry. This paper aims to present fully automatic segmentation algorithm to extract lower jaw and mandible from panoramic x-ray. Methods/Analysis: Panoramic x-rays provide features of various diseases like advanced periodontal diseases, cysts, tumors, impacted teeth, fractures in lower or upper jaw, etc. Image processing techniques are applied on these images to detect diseases automatically. Segmentation is an inevitable step in automation to extract disease features. Majority existing literature use semi-automatic method for segmentation of panoramic x-ray. The algorithm proposed in this paper is fully automatic method of segmentation. Findings: Panoramic x-rays give entire view of mouth and facilitate assessment of hard tissue structures of the facial area rather than assessment of dental carries. Segmentation of panoramic x-rays is an inevitable step in automating diagnosis of oral disorders like, tumor, cysts, fracture, etc. In this paper, an algorithm is developed for automatic segmentation of lower jaw and then mandibular bone in dental digital panoramic x-ray. Dental digital panoramic x-ray is a good quality image which does not require noise removal. The segmentation algorithm uses novel strips method where input image is subdivided in number of strips. These strips are further processed to separate lower jaw and mandible from input image. The algorithm is fully automatic and does not involve human intervention at any level. Paper presents analysis of number of strips as 3, 5, 7 and 9. The algorithm presented in this paper gives 93% success for segmentation of lower jaw and 90% success in segmentation of mandible structure when strip count is 5. Applications/Improvement: Strip method focuses ROI and ignores components like teeth, spinal cord shadow, etc. Central strip is useful to segment lower jaw whereas side strips are useful to detect mandible edge points.

Discrimination of Breast Cancer with Microcalcifications on Mammography by Deep Learning

Microcalcification is an effective indicator of early breast cancer. To improve the diagnostic accuracy of microcalcifications, this study evaluates the performance of deep learning-based models on large datasets for its discrimination. A semi-automated segmentation method was used to characterize all microcalcifications. A discrimination classifier model was constructed to assess the accuracies of microcalcifications and breast masses, either in isolation or combination, for classifying breast lesions. Performances were compared to benchmark models. Our deep learning model achieved a discriminative accuracy of 87.3% if microcalcifications were characterized alone, compared to 85.8% with a support vector machine. The accuracies were 61.3% for both methods with masses alone and improved to 89.7% and 85.8% after the combined analysis with microcalcifications. Image segmentation with our deep learning model yielded 15, 26 and 41 features for the three scenarios, respectively. Overall, deep learning based on large datasets was superior to standard methods for the discrimination of microcalcifications. Accuracy was increased by adopting a combinatorial approach to detect microcalcifications and masses simultaneously. This may have clinical value for early detection and treatment of breast cancer.

Semiautomated segmentation of head and neck cancers in 18F-FDG PET scans: A just-enough-interaction approach.

Purpose:The purpose of this work was to develop, validate, and compare a highly computer-aided method for the segmentation of hot lesions in head and neck 18F-FDG PET scans.Methods:A semiautomated segmentation method was developed, which transforms the segmentation problem into a graph-based optimization problem. For this purpose, a graph structure around a user-provided approximate lesion centerpoint is constructed and a suitable cost function is derived based on local image statistics. To handle frequently occurring situations that are ambiguous (e.g., lesions adjacent to each other versus lesion with inhomogeneous uptake), several segmentation modes are introduced that adapt the behavior of the base algorithm accordingly. In addition, the authors present approaches for the efficient interactive local and global refinement of initial segmentations that are based on the “just-enough-interaction” principle. For method validation, 60 PET/CT scans from 59 different subjects with 230 head and neck lesions were utilized. All patients had squamous cell carcinoma of the head and neck. A detailed comparison with the current clinically relevant standard manual segmentation approach was performed based on 2760 segmentations produced by three experts.Results:Segmentation accuracy measured by the Dice coefficient of the proposed semiautomated and standard manual segmentation approach was 0.766 and 0.764, respectively. This difference was not statistically significant (p = 0.2145). However, the intra- and interoperator standard deviations were significantly lower for the semiautomated method. In addition, the proposed method was found to be significantly faster and resulted in significantly higher intra- and interoperator segmentation agreement when compared to the manual segmentation approach.Conclusions:Lack of consistency in tumor definition is a critical barrier for radiation treatment targeting as well as for response assessment in clinical trials and in clinical oncology decision-making. The properties of the authors approach make it well suited for applications in image-guided radiation oncology, response assessment, or treatment outcome prediction.

Use of a semi-automated cardiac segmentation tool improves reproducibility and speed of segmentation of contaminated right heart magnetic resonance angiography.

Three-dimensional printing has an increasing number of clinical applications in pediatric cardiology. Time required for dataset segmentation and conversion to stereolithography (STL) format remains a significant limitation. We investigated the impact of semi-automated cardiovascular-specific segmentation software on time and reproducibility of segmentation. Magnetic resonance angiograms (MRAs) of 19 patients undergoing intervention for right ventricular outflow lesions were segmented to demonstrate the right heart. STLs were created by two independent clinicians using semi-automated cardiovascular segmentation (SAS) and traditional manual segmentation (MS). Time was recorded and geometric STL disagreement was determined (0 % = no disagreement, 100 % = complete disagreement). MRA datasets were categorized as clean when only right heart structures were present in the MRA, or contaminated when left heart structures were also present and required removal. Eighteen (seven clean and 11 contaminated) cases were successfully segmented with both methods. Time to STL for clean datasets was faster with MS than SAS [median 209s (IQR 192-252) vs. 296s (272-317), p = 0.018] while contaminated datasets were faster with SAS [455s (384-561) vs. 866s (310-1429), p = 0.033]. Interobserver STL geometric disagreement was significantly lower using SAS than MS overall (0.70 ± 1.15 % vs. 1.31 ± 1.52 %, p = 0.030), and for the contaminated subset (0.81 ± 1.08 % vs. 1.75 ± 1.57 %, p = 0.036). Most geometric disagreement occurred at areas where left heart contamination was removed. Semi-automated segmentation was faster and more reproducible for contaminated datasets, while MS was faster but equally reproducible for clean datasets. Semi-automated segmentation methods are preferable for contaminated datasets and continued refinement of these tools should be supported.

Fully automatic prostate segmentation from transrectal ultrasound images based on radial bas-relief initialization and slice-based propagation

Prostate segmentation from transrectal ultrasound (TRUS) images plays an important role in the diagnosis and treatment planning of prostate cancer. In this paper, a fully automatic slice-based segmentation method was developed to segment TRUS prostate images. The initial prostate contour was determined using a novel method based on the radial bas-relief (RBR) method, and a false edge removal algorithm proposed here in. 2D slice-based propagation was used in which the contour on each image slice was deformed using a level-set evolution model, which was driven by edge-based and region-based energy fields generated by dyadic wavelet transform. The optimized contour on an image slice propagated to the adjacent slice, and subsequently deformed using the level-set model. The propagation continued until all image slices were segmented. To determine the initial slice where the propagation began, the initial prostate contour was deformed individually on each transverse image. A method was developed to self-assess the accuracy of the deformed contour based on the average image intensity inside and outside of the contour. The transverse image on which highest accuracy was attained was chosen to be the initial slice for the propagation process. Evaluation was performed for 336 transverse images from 15 prostates that include images acquired at mid-gland, base and apex regions of the prostates. The average mean absolute difference (MAD) between algorithm and manual segmentations was 0.79±0.26mm, which is comparable to results produced by previously published semi-automatic segmentation methods. Statistical evaluation shows that accurate segmentation was not only obtained at the mid-gland, but also at the base and apex regions.

Crown Segmentation From Computed Tomography Images With Metal Artifacts

Tooth segmentation from dental computed tomography (CT) images with metal artifacts is challenging as metal artifacts make some of the crown boundaries unrecognizable. This letter proposes a semiautomatic method for crown segmentation from CT images with metal artifacts. A user manually selects a starting slice and initializes this slice. Then crown contours are segmented automatically from volumetric CT images slice by slice. In the segmentation of each slice, the Radon transform is used to extract a line to separate neighboring crowns into independent ones. A statistical shape prior-based level set model is then applied to segment each crown from the mesial or distal side of the line. The proposed method was tested on 15 set of volumetric images. Experimental results validated that it is effective to extract crown contours from CT images with metal artifacts.

Assessment of Glioma Response to Radiotherapy Using Multiple MRI Biomarkers with Manual and Semiautomated Segmentation Algorithms.

Multimodality magnetic resonance imaging (MRI) can provide complementary information in the assessment of brain tumors. We aimed to segment tumor in amide proton transfer-weighted (APTw) images and to investigate multiparametric MRI biomarkers for the assessment of glioma response to radiotherapy. For tumor extraction, we evaluated a semiautomated segmentation method based on region of interest (ROI) results by comparing it with the manual segmentation method. Thirteen nude rats injected with U87 tumor cells were irradiated by an 8-Gy radiation dose. All MRI scans were performed on a 4.7-T animal scanner preradiation, and at day 1, day 4, and day 8 postradiation. Two experts performed manual and semiautomated methods to extract tumor ROIs on APTw images. Multimodality MRI signals of the tumors, including structural (T2 and T1 ), functional (apparent diffusion coefficient and blood flow), and molecular (APTw and magnetization transfer ratio or MTR), were calculated and compared quantitatively. The semiautomated method provided more reliable tumor extraction results on APTw images than the manual segmentation, in less time. A considerable increase in the ADC intensities of the tumor was observed during the postradiation. A steady decrease in the blood flow values and in the APTw signal intensities were found after radiotherapy. The semiautomated method of tumor extraction showed greater efficiency and stability than the manual method. Apparent diffusion coefficient, blood flow, and APTw are all useful biomarkers in assessing glioma response to radiotherapy.

Metastatic liver tumour segmentation with a neural network-guided 3D deformable model.

The segmentation of liver tumours in CT images is useful for the diagnosis and treatment of liver cancer. Furthermore, an accurate assessment of tumour volume aids in the diagnosis and evaluation of treatment response. Currently, segmentation is performed manually by an expert, and because of the time required, a rough estimate of tumour volume is often done instead. We propose a semi-automatic segmentation method that makes use of machine learning within a deformable surface model. Specifically, we propose a deformable model that uses a voxel classifier based on a multilayer perceptron (MLP) to interpret the CT image. The new deformable model considers vertex displacement towards apparent tumour boundaries and regularization that promotes surface smoothness. During operation, a user identifies the target tumour and the mesh then automatically delineates the tumour from the MLP processed image. The method was tested on a dataset of 40 abdominal CT scans with a total of 95 colorectal metastases collected from a variety of scanners with variable spatial resolution. The segmentation results are encouraging with a Dice similarity metric of [Formula: see text] and demonstrates that the proposed method can deal with highly variable data. This work motivates further research into tumour segmentation using machine learning with more data and deeper neural networks.