Volume Segmentation Method Research Articles

Evaluation of a Fully Automated Method for Ventricular Volume Segmentation Before and After Shunt Surgery in Idiopathic Normal Pressure Hydrocephalus

Determination of the ventricle size in idiopathic normal pressure hydrocephalus (iNPH) is essential for diagnosis and follow-up of shunt results. Fully automated segmentation methods are anticipated to optimize the accuracy and time efficiency of ventricular volume measurements. We evaluated the accuracy of preoperative and postoperative ventricular volume measurements in iNPH by a magnetic resonance imaging (MRI)-based licensed software for fully automated quantitative assessment. Forty-eight patients diagnosed with iNPH were retrospectively analyzed. All patients received a ventriculoperitoneal shunt and had symptom grading and routine MRI preoperatively and 3-6 months postoperatively. Ventricular volumes, generated by fully automated T1-weighted imaging volume sequence segmentation, were compared with semiautomatic measurements and routine radiologic reports. The relation of postoperative ventricular size change to clinical response was evaluated. Fully automated segmentation was achieved in 95% of the MRIs, but showed various rates of 8 minor segmentation errors. The correlation between both segmentation methods was very strong (r >0.9) and the agreement very good using Bland-Altman analyses. The ventricular volumes differed significantly between semiautomated and fully automated segmentations and between preoperative and postoperative MRI. The fully automated method systematically overestimated the ventricles by a median 15 mL preoperatively and 14 mL postoperatively; hence, the magnitudes of volume changes were equivalent. Routine radiologic reports of ventricular size changes were inaccurate in 51% and lacked association with treatment response. Objectively measured ventricular volume changes correlated moderately with postoperative clinical improvement. A fully automated volumetric method permits reliable evaluation of preoperative ventriculomegaly and postoperative ventricular volume change in idiopathic normal pressure hydrocephalus.

Deep learning-based pectoralis muscle volume segmentation method from chest computed tomography image using sagittal range detection and axial slice-based segmentation.

The pectoralis muscle is an important indicator of respiratory muscle function and has been linked to various parenchymal biomarkers, such as airflow limitation severity and diffusing capacity for carbon monoxide, which are widely used in diagnosing parenchymal diseases, including asthma and chronic obstructive pulmonary disease. Pectoralis muscle segmentation is a method for measuring muscle volume and mass for various applications. The segmentation method is based on deep-learning techniques that combine a muscle area detection model and a segmentation model. The training dataset for the detection model comprised multichannel images of patients, whereas the segmentation model was trained on 7,796 cases of the computed tomography (CT) image dataset of 1,841 patients. The dataset was expanded incrementally through an active learning process. The performance of the model was evaluated by comparing the segmentation results with manual annotations by radiologists and the volumetric differences between the CT image datasets of the same patients. The results indicated that the machine learning model is promising in segmenting the pectoralis major muscle, with good agreement between the automatic segmentation and manual annotations by radiologists. The training accuracy and loss values of the validation set were 0.9954 and 0.0725, respectively, and for segmentation, the loss value was 0.0579. This study shows the potential clinical usefulness of the machine learning model for pectoralis major muscle segmentation as a quantitative biomarker for various parenchymal and muscular diseases.

Methods and Reproducibility of Liver Fat Measurement Using 3-Dimensional Liver Segmentation From Noncontrast Computed Tomography in EVAPORATE Cohort.

Nonalcoholic fatty liver disease not only shares multiple risk factors with cardiovascular disease but also independently predicts its increased risk and related outcomes. Here, we evaluate reproducibility of 3-dimensional (3D) liver volume segmentation method to identify fatty liver on noncontrast cardiac computed tomography (CT) and compare measures with previously validated 2-dimensional (2D) segmentation CT criteria for the measurement of liver fat. The study included 68 participants enrolled in the EVAPORATE trial and underwent serial noncontrast cardiac CT. Liver attenuation < 40 Hounsfield units (HU) was used for diagnosing fatty liver, as done in the MESA study. Two-dimensional and 3D segmentation of the liver were performed by Philips software. Bland-Altman plot analysis was used to assess reproducibility. Interreader reproducibility of 3D liver mean HU measurements was 96% in a sample of 111 scans. Reproducibility of 2D and 3D liver mean HU measurements was 93% in a sample of 111 scans. Reproducibility of change in 2D and 3D liver mean HU was 94% in 68 scans. Kappa, a measure of agreement in which the 2D and 3D measures both identified fatty liver, was excellent at 96.4% in 111 scans. Fatty liver can be reliably diagnosed and measured serially in a stable and reproducible way by 3D liver segmentation of noncontrast cardiac CT scans. Future studies need to explore the sensitivity and stability of measures for low liver fat content by 3D segmentation, over the current 2D methodology. This measure can serve as an imaging biomarker to understand mechanistic correlations between atherosclerosis, fatty liver, and cardiovascular disease risk.

H-Seg: a horizontal reconstruction volume segmentation method for accurate depth estimation in a computer-generated hologram.

In this work, we introduce a novel approach for depth estimation in a computer-generated hologram by employing horizontal segmentation of the reconstruction volume instead of conventional vertical segmentation. The reconstruction volume is divided into horizontal slices and each slice is processed using a residual U-net architecture to identify in-focus lines, enabling determination of the slice's intersection with the 3D scene. The individual slice results are then combined to generate a dense depth map of the scene. Our experiments demonstrate the effectiveness of our method, with improved accuracy, faster processing times, lower graphics processing unit (GPU) utilization, and smoother predicted depth maps than existing state-of-the-art models.

Head and neck tumor segmentation convolutional neural network robust to missing PET/CT modalities using channel dropout

Objective. Radiation therapy for head and neck (H&N) cancer relies on accurate segmentation of the primary tumor. A robust, accurate, and automated gross tumor volume segmentation method is warranted for H&N cancer therapeutic management. The purpose of this study is to develop a novel deep learning segmentation model for H&N cancer based on independent and combined CT and FDG-PET modalities. Approach. In this study, we developed a robust deep learning-based model leveraging information from both CT and PET. We implemented a 3D U-Net architecture with 5 levels of encoding and decoding, computing model loss through deep supervision. We used a channel dropout technique to emulate different combinations of input modalities. This technique prevents potential performance issues when only one modality is available, increasing model robustness. We implemented ensemble modeling by combining two types of convolutions with differing receptive fields, conventional and dilated, to improve capture of both fine details and global information. Main Results. Our proposed methods yielded promising results, with a Dice similarity coefficient (DSC) of 0.802 when deployed on combined CT and PET, DSC of 0.610 when deployed on CT, and DSC of 0.750 when deployed on PET. Significance. Application of a channel dropout method allowed for a single model to achieve high performance when deployed on either single modality images (CT or PET) or combined modality images (CT and PET). The presented segmentation techniques are clinically relevant to applications where images from a certain modality might not always be available.

Assessment of Prostate Volume and Prostate-specific Antigen Density With the Segmentation Method on Magnetic Resonance Imaging.

This study aimed to compare the prostate volume (PV) and prostate-specific antigen density (PSAD) obtained using the ellipsoid volume formula or segmentation methods on magnetic resonance imaging (MRI) and further predict prostate cancer (PCa). Retrospectively, the enrolled patients underwent prostate MRI and had PSA levels between 4 and 10 ng/ml. The PV was measured with both the ellipsoid volume formula (PVe) and the segmentation method (PVs). The transitional zone volume (TZV) was measured with the segmentation method. The PSADe, PSADs, and PSAD_TZV were calculated. Bland-Altman plots were used to compare the agreements. ROC curve analysis was used to compare the diagnostic accuracies to predict PCa. The results were also compared between the PCa and the no-PCa groups, and among tumors with different locations and different Gleason scores (GS). Seventy-six of the 117 enrolled patients were classified into the PCa group. There were high agreements between PVs and PVe as well as between PSADs and PSADe, while several outliers were mainly due to post-transurethral resection of the prostate changes and irregular hyperplastic nodules. The diagnostic accuracy of PSADe (AUC: 0.732) was slightly higher than that of PSADs (AUC: 0.729) and PSAD_TZV (AUC: 0.715). The PSADe and PSADs were not different among different tumor locations but were higher in GS ≥7 lesions (both p=0.006). The segmentation method can be an alternative method to measure PV and calculate PSAD before prostate biopsy, particularly in post-transurethral resection of the prostate patients or those with irregular hyperplastic nodules.

Methods to delineate tumour for radiotherapy by fluorodeoxyglucose positron emission tomography

Numerous articles have been published showing the interest to delineate tumours using fluorodeoxyglucose positron emission tomography/computed tomography images in radiotherapy planning. This imaging is used to identify tumour tissues with increased glucose metabolism compared to healthy surrounding tissues. This volume corresponds to the metabolic tumour volume. Despite extensive research on metabolic tumour volume segmentation methods, there is currently no consensus on the optimal segmentation method to use. In this review, the main methods proposed in the literature are presented, as well as their advantages and disadvantages in the context of radiotherapy.

A semi-automatic approach for epicardial adipose tissue segmentation and quantification on cardiac CT scans

Many studies have shown that epicardial fat is associated with a higher risk of heart diseases. Accurate epicardial adipose tissue quantification is still an open research issue. Considering that manual approaches are generally user-dependent and time-consuming, computer-assisted tools can considerably improve the result repeatability as well as reduce the time required for performing an accurate segmentation. Unfortunately, fully automatic strategies might not always identify the Region of Interest (ROI) correctly. Moreover, they could require user interaction for handling unexpected events.This paper proposes a semi-automatic method for Epicardial Fat Volume (EFV) segmentation and quantification. Unlike supervised Machine Learning approaches, the method does not require any initial training or modeling phase to set up the system. As a further key novelty, the method also yields a subdivision into quartiles of the adipose tissue density. Quartile-based analysis conveys information about fat densities distribution, enabling an in-depth study towards a possible correlation between fat amounts, fat distribution, and heart diseases.Experimental tests were performed on 50 Calcium Score (CaSc) series and 95 Coronary Computed Tomography Angiography (CorCTA) series. Area-based and distance-based metrics were used to evaluate the segmentation accuracy, by obtaining Dice Similarity Coefficient (DSC) = 93.74% and Mean Absolute Distance (MAD) = 2.18 for CaSc, as well as DSC = 92.48% and MAD = 2.87 for CorCTA. Moreover, the Pearson and Spearman coefficients were computed for quantifying the correlation between the ground-truth EFV and the corresponding automated measurement, by obtaining 0.9591 and 0.9490 for CaSc, and 0.9513 and 0.9319 for CorCTA, respectively.In conclusion, the proposed EFV quantification and analysis method represents a clinically useable tool assisting the cardiologist to gain insights into a specific clinical scenario and leading towards personalized diagnosis and therapy.

Research on Digital Elevation Model Construction Method Based on TIN

Digital Elevation Model (DEM) is a very important part of geographic information system and a key part of various terrain surface studies. In this paper, we first introduce the classical DEM construction method based on irregular triangulation (TIN). In view of the shortcomings of the classical DEM construction method, this paper proposes a volume segmentation method based on variance criterion, in which the variance criterion can take full account of the elevation information of the original data and select the triangles with good shape in space to ensure that the constructed DEM is closer to the actual surface. Finally, an example is given to demonstrate the advantage of the variance criterion relative to the empty circumscribed circle criterion. The classical DEM construction method and the method of volume segmentation based on variance criterion are realized by MATLAB programming.

Image-based clustering and connected component labeling for rapid automated left and right ventricular endocardial volume extraction and segmentation in full cardiac cycle multi-frame MRI images of cardiac patients

A rapid method for left and right ventricular endocardial volume segmentation and clinical cardiac parameter calculation from MRI images of cardiac patients is presented. The clinical motivation is providing cardiologists a tool for assessing the cardiac function in a patient through the left ventricular endocardial volume's ejection fraction. A new method combining adapted fuzzy membership-based c-means pixel clustering and connected regions component labeling is used for automatic segmentation of the left and right ventricular endocardial volumes. This proposed pixel clustering with labeling approach avoids manual initialization or user intervention and does not require specifying the region of interest. This method fully automatically extracts the left and right ventricular endocardial volumes and avoids manual tracing on all MRI image frames in the complete cardiac cycle from systole to diastole. The average computational processing time per frame is 0.6s, making it much more efficient than deformable methods, which need several iterations for the evolution of the snake or contour. Accuracy of the automated method presented herein was validated against manual tracing-based extraction, performed with the guidance of cardiac experts, on several MRI frames. Dice coefficients between the proposed automatic versus manual traced ventricular endocardial volume segmentations were observed to be 0.9781 ± 0.0070 (for left ventricular endocardial volume) and 0.9819 ± 0.0058 (for right ventricular endocardial volume), and the Pearson correlation coefficients were observed to be 0.9655 ± 0.0206 (for left ventricular endocardial volume) and 0.9870 ± 0.0131 (for right ventricular endocardial volume). Graphical abstract The left ventricular endocardial volume segmentation methodology illustrated as a series of algorithms.

Accurate Segmentation of Heart Volume in CTA With Landmark-Based Registration and Fully Convolutional Network

Accurate delineation of cardiac structures in CTA images remains challenging, due to the similar appearance of different components, the highly variable shape and intensity among individuals, and especially the vicinity of heart volumes and backgrounds (including fat tissues, great vessels, and livers). Therefore, we proposed an accurate heart volume segmentation method with landmark-based registration (LMR) and 3D fully convolutional network (3D-FCN). First, we defined uniformity constrained landmarks in the training images and then trained a regression forest model (RFM) to detect these landmarks in the testing image. Second, the registration between the landmarks in each training image and the test image was performed by a shallow neural network, which guided the label propagation from atlases to the test image. After the label fusion with majority voting, we finally constructed a 3D-FCN to further refine the boundary voxels with low voting values. In 22 cardiac CTA images, we compared our method with multi-atlas segmentation, active shape model, DeepMedic, LMR + DeepMedic, and the separately implemented LMR and 3D-FCN. The results demonstrated the superiority of our method for the segmentation of heart volumes, with the average accuracy, dice coefficient, and mean Hausdorff distance as 96.25%, 93.98%, and 2.12 voxels, respectively. Furthermore, the training efficiency of the constructed 3D-FCN was higher than that of the DeepMedic through the comparison of training time and convergence of loss. The proposed heart segmentation could provide an accurate region of interest (ROI) for not only the four heart chambers but also the major trunks of vessels and coronary arteries.

253. An accurate and operator independent method for biological tumour volume segmentation

Purpose The aim of this paper is to develop an operator independent method for biological tumour volume (BTV) delineation from Positron Emission Tomography (PET) images. BTV delineation is challenging because of the low spatial resolution and high noise level in PET images. In addition, BTV varies substantially depending on the method used to segment. Manual delineation is widely-used, but it is strongly user dependent. Methods The proposed method starts with the automatic identification of the PET slice with maximum Standardized Uptake Value (SUV). Then, a user- independent mask is obtained by a rough pre-segmentation step and it is used to perform the local active contour segmentation on the next slices. When the proposed method finds a slice where the mean SUV on the interior of the delineation contour is greater or equal than the mean SUV on the exterior of the delineation contour, the segmentation is automatically stopped. The algorithm is evaluated on four datasets of synthetic lesions considering different ratios between lesion and background radioactivity concentrations. In this way, the actual lesion volumes are known and the segmentation algorithm is evaluated under different contrast ratio scenarios. In addition, BTVs of 25 patient studies have been manually delineated and compared to the proposed method in order to assess its applicability in a clinical environment. Results In phantom experiments, dice similarity coefficient (DSC) rate and true positive volume fraction (TPVF) rate greater than 90% were observed in synthetic lesions with a diameter greater than 17 mm. In clinical cases, TPVF and DSC were 91.00 ± 7.33% and 85.98 ± 3.40%, respectively. Conclusion Our method produces accurate segmentation results in phantom studies. In addition, it is feasible in clinical context and shows good accuracy in realistic conditions, reducing any user interaction.

Parameters Influencing PET Imaging Features: A Phantom Study with Irregular and Heterogeneous Synthetic Lesions.

Aim To evaluate reproducibility and stability of radiomic features as effects of the use of different volume segmentation methods and reconstruction settings. The potential of radiomics in really capturing the presence of heterogeneous tumor uptake and irregular shape was also investigated. Materials and Methods An anthropomorphic phantom miming real clinical situations including synthetic lesions with irregular shape and nonuniform radiotracer uptake was used. 18F-FDG PET/CT measurements of the phantom were performed including 38 lesions of different shape, size, lesion-to-background ratio, and radiotracer uptake distribution. Different reconstruction parameters and segmentation methods were considered. COVs were calculated to quantify feature variations over the different reconstruction settings. Friedman test was applied to the values of the radiomic features obtained for the considered segmentation approaches. Two sets of test-retest measurement were acquired and the pairwise intraclass correlation coefficient was calculated. Fifty-eight morphological and statistical features were extracted from the segmented lesion volumes. A Mann–Whitney test was used to evaluate significant differences among each feature when calculated from heterogeneous versus homogeneous uptake. The significance of each radiomic feature in terms of capturing heterogeneity was evaluated also by testing correlation with gold standard indexes of heterogeneity and sphericity. Results The choice of the segmentation method has a strong impact on the stability of radiomic features (less than 20% can be considered stable features). Reconstruction affects the estimate of radiomic features (only 26% are stable). Thirty-one radiomic features (53%) resulted to be reproducible, 11 of them are able to discriminate heterogeneity. Among these, we found a subset of 3 radiomic features strongly correlated with GS heterogeneity index that can be suggested as good features for retrospective evaluations.

Do 3D-calculated volume distribution of a stone in pelvicalyceal system affect complications of percutaneous nephrolithotomy?

In our study, we examined the effect of the three-dimensional (3D) stone segmentation volume and its ratio to the renal collecting system on complication rates. Data from141 patients who underwent PCNL surgery were analyzed retrospectively. Volume segmentation of both the renal collecting system and stones was obtained from 3D segmentation software with the images on CT data. After creation of a 3D surface volume rendering of renal stones and the collecting system, segmentation of the renal collecting system volume (RCSV) and analyzed stone volume (ASV) was analyzed and the ASV-to-RCSV ratio was calculated. Univariate analysis and multivariate logistic regression model were used to determine factors that affected complication status. Diagnostic value for the prediction of complication rates was analyzed using receiver operating characteristic (ROC) incline. Overall, there were 141 (92 male and 49 female) eligible patients included in the current study. The overall complication rate for PCNL monotherapy was 31.9%. Multivariate regression analysis (forward stepwise) revealed that the ASV-to-RCSV ratio and number of tracts were independent risk factors for developing complications (OR 1.17, p < 0.001; OR 7.87, p = 0.002; respectively). The ROC analysis revealed a cut-off value of 16.23% (AUC 0.869, p < 0.001, sensitivity 93.3%, specificity 78.1%) for the ASV-to-RCSV ratio. The distribution of stone burden volume in the pelvicalyceal system, which is calculated as a numerical value using the 3D volume segmentation method, is an important predictor of the complication rate before PCNL. The ASV-to-RCSV ratio as a quantitative value may be an instrument for urologists before surgery to help preoperative planning.

Impact of PET and MRI threshold-based tumor volume segmentation on patient-specific targeted radionuclide therapy dosimetry using CLR1404

Variations in tumor volume segmentation methods in targeted radionuclide therapy (TRT) may lead to dosimetric uncertainties. This work investigates the impact of PET and MRI threshold-based tumor segmentation on TRT dosimetry in patients with primary and metastatic brain tumors. In this study, PET/CT images of five brain cancer patients were acquired at 6, 24, and 48 h post-injection of 124I-CLR1404. The tumor volume was segmented using two standardized uptake value (SUV) threshold levels, two tumor-to-background ratio (TBR) threshold levels, and a T1 Gadolinium-enhanced MRI threshold. The dice similarity coefficient (DSC), jaccard similarity coefficient (JSC), and overlap volume (OV) metrics were calculated to compare differences in the MRI and PET contours. The therapeutic 131I-CLR1404 voxel-level dose distribution was calculated from the 124I-CLR1404 activity distribution using RAPID, a Geant4 Monte Carlo internal dosimetry platform. The TBR, SUV, and MRI tumor volumes ranged from 2.3–63.9 cc, 0.1–34.7 cc, and 0.4–11.8 cc, respectively. The average ± standard deviation (range) was 0.19 ± 0.13 (0.01–0.51), 0.30 ± 0.17 (0.03–0.67), and 0.75 ± 0.29 (0.05–1.00) for the JSC, DSC, and OV, respectively. The DSC and JSC values were small and the OV values were large for both the MRI-SUV and MRI-TBR combinations because the regions of PET uptake were generally larger than the MRI enhancement. Notable differences in the tumor dose volume histograms were observed for each patient. The mean (standard deviation) 131I-CLR1404 tumor doses ranged from 0.28–1.75 Gy GBq−1 (0.07–0.37 Gy GBq−1). The ratio of maximum-to-minimum mean doses for each patient ranged from 1.4–2.0. The tumor volume and the interpretation of the tumor dose is highly sensitive to the imaging modality, PET enhancement metric, and threshold level used for tumor volume segmentation. The large variations in tumor doses clearly demonstrate the need for standard protocols for multimodality tumor segmentation in TRT dosimetry.

Evaluation of stone volume distribution in renal collecting system as a predictor of stone-free rate after percutaneous nephrolithotomy: a retrospective single-center study.

We analyzed our stone-free rates of PNL with regard to stone burden and its ratio to the renal collecting system volume. Data of 164 patients who underwent PNL were analyzed retrospectively. Volume segmentation of renal collecting system and stones were done using 3D segmentation software with the images obtained from CT data. Analyzed stone volume (ASV) and renal collecting system volume (RCSV) were measured and the ASV-to-RCSV ratio was calculated after the creation of a 3D surface volume rendering of renal stones and the collecting system. Univariate and multivariate statistical analyses were performed to determine factors affecting stone-free rates; also we assessed the predictive accuracy of the ASV-to-RCSV ratio using the receiving operating curve (ROC) and AUC. The stone-free rate of PNL monotherapy was 53% (164 procedures).The ASV-to-RCSV ratio and calyx number with stones were the most influential predictors of stone-free status (OR 4.15, 95% CI 2.24-7.24, <0.001, OR 2.62, 95% CI 1.38-4.97, p<0.001, respectively). Other factors associated with the stone-free rate were maximum stone size (p<0.029), stone surface area (p<0.010), and stone burden volume (p<0.001). Predictive accuracy of the ASV-to-RCSV ratio was AUC 0.76. Stone burden volume distribution in the renal collecting system, which is calculated using the 3D volume segmentation method, is a significant determinant of the stone-free rate before PCNL surgery. It could be used as a single guide variable by the clinician before renal stone surgery to predict extra requirements for stone clearance.

Dynamic Volume Assessment of Hepatocellular Carcinoma in Rat Livers Using a Clinical 3T MRI and Novel Segmentation

ABSTRACTPurpose: In vivo liver cancer research commonly uses rodent models. One of the limitations of such models is the lack of accurate and reproducible endpoints for a dynamic assessment of growing tumor nodules. The aim of this study was to validate a noninvasive, true volume segmentation method using two rat hepatocellular carcinoma (HCC) models, correlating magnetic resonance imaging (MRI) with histological volume measurement, and with blood levels of α-fetoprotein. Materials and methods: We used 3T clinical MRI to quantify tumor volume with follow-up over time. Using two distinct rat HCC models, calculated MRI tumor volumes were correlated with volumes from histological sections, or with blood levels of α-fetoprotein. Eleven rats, comprising six Buffalo rats (n = 9 scans) and five Fischer rats (n = 14 tumors), were injected in the portal vein with 2.5 × 105 and 2.0 × 106 syngeneic HCC cells, respectively. Longitudinal (T1) relaxation time- and transverse (T2) relaxation time-weighted MR images were acquired. Results: The three-dimensional (3D) T1-weighted gradient echo had 0.35-mm isotropic resolution allowing accurate semi-automatic volume segmentation. 2D T2-weighted imaging provided high tumor contrast. Segmentation of combined 3D gradient echo T1-weighted images and 2D turbo spin echo T2-weighted images provided excellent correlation with histology (y = 0.866x + 0.034, R² = 0.997 p < .0001) and with α-fetoprotein (y = 0.736x + 1.077, R² = 0.976, p < .0001). There was robust inter- and intra-observer reproducibility (intra-class correlation coefficient > 0.998, p < .0001). Conclusions: We have developed a novel, noninvasive contrast imaging protocol which enables semi-automatic 3D volume quantification to analyze nonspherical tumor nodules and to follow up the growth of tumor nodules over time.

Five multiresolution-based calcium volume measurement techniques from coronary IVUS videos: A comparative approach

Background and objectiveFast intravascular ultrasound (IVUS) video processing is required for calcium volume computation during the planning phase of percutaneous coronary interventional (PCI) procedures. Nonlinear multiresolution techniques are generally applied to improve the processing time by down-sampling the video frames. MethodsThis paper presents four different segmentation methods for calcium volume measurement, namely Threshold-based, Fuzzy c-Means (FCM), K-means, and Hidden Markov Random Field (HMRF) embedded with five different kinds of multiresolution techniques (bilinear, bicubic, wavelet, Lanczos, and Gaussian pyramid). This leads to 20 different kinds of combinations. IVUS image data sets consisting of 38,760 IVUS frames taken from 19 patients were collected using 40 MHz IVUS catheter (Atlantis® SR Pro, Boston Scientific®, pullback speed of 0.5 mm/sec.). The performance of these 20 systems is compared with and without multiresolution using the following metrics: (a) computational time; (b) calcium volume; (c) image quality degradation ratio; and (d) quality assessment ratio. ResultsAmong the four segmentation methods embedded with five kinds of multiresolution techniques, FCM segmentation combined with wavelet-based multiresolution gave the best performance. FCM and wavelet experienced the highest percentage mean improvement in computational time of 77.15% and 74.07%, respectively. Wavelet interpolation experiences the highest mean precision-of-merit (PoM) of 94.06 ± 3.64% and 81.34 ± 16.29% as compared to other multiresolution techniques for volume level and frame level respectively. Wavelet multiresolution technique also experiences the highest Jaccard Index and Dice Similarity of 0.7 and 0.8, respectively. Multiresolution is a nonlinear operation which introduces bias and thus degrades the image. The proposed system also provides a bias correction approach to enrich the system, giving a better mean calcium volume similarity for all the multiresolution-based segmentation methods. After including the bias correction, bicubic interpolation gives the largest increase in mean calcium volume similarity of 4.13% compared to the rest of the multiresolution techniques. The system is automated and can be adapted in clinical settings. ConclusionsWe demonstrated the time improvement in calcium volume computation without compromising the quality of IVUS image. Among the 20 different combinations of multiresolution with calcium volume segmentation methods, the FCM embedded with wavelet-based multiresolution gave the best performance.

A fully automatic method for biological target volume segmentation of brain metastases

ABSTRACTLeksell Gamma Knife is a mini‐invasive technique to obtain a complete destruction of cerebral lesions delivering a single high dose radiation beam. Positron Emission Tomography (PET) imaging is increasingly utilized for radiation treatment planning. Nevertheless, lesion volume delineation in PET datasets is challenging because of the low spatial resolution and high noise level of PET images. Nowadays, the biological target volume (BTV) is manually contoured on PET studies. This procedure is time expensive and operator‐dependent. In this article, a fully automatic algorithm for the BTV delineation based on random walks (RW) on graphs is proposed. The results are compared with the outcomes of the original RW method, 40% thresholding method, region growing method, and fuzzy c‐means clustering method. To validate the effectiveness of the proposed approach in a clinical environment, BTV segmentation on 18 patients with cerebral metastases is performed. Experimental results show that the segmentation algorithm is accurate and has real‐time performance satisfying the physician requirements in a radiotherapy environment.

FDG PET SUV-Based Segmentation Method for Biological GTV (BTV) Delineation in Pharyngeal Cancer Dose-Painted IMRT: Preliminary Retrospective Evaluation of Patterns of Local Relapse

To evaluate preliminary data on accuracy of FDG PET SUV-based volume segmentation method for primary BTV delineation in pharyngeal cancer, through the analysis of FDG PET-detected early local recurrences in a prospective series of 18 patients (pts) treated with dose-painted (DP) IMRT.