Volumetric Features Research Articles

Automated OAR Anomaly and Error Detection Tool in Radiation Therapy

Organ at risk (OAR) segmentation anomalies can negatively impact treatment planning quality, and potentially increase treatment morbidity. Therefore, we developed and tested the efficacy of a quality assurance (QA) tool to assist in identifying potential OAR segmentation errors. The QA tool computes OAR volumetric features and compares them with respect to a statistical reference developed from historical priors. In contrast to previous approaches, all features were computed in 3-dimensional space. Furthermore, feature comparison was used in a multi-criteria triggering system that warns user of specific errors based on combinations of feature anomalies. The statistical reference contained features derived from 73 prior lung cancer patients. The tool was assessed using 15 independent test sets: 9 clinical manual segmentation (MS), 3 automatic segmentation (AS) and 3 propagated segmentation (PS) from deformable mapping. The test sets contained 107 OAR structures. The test MS sets included 34 OAR manipulations (ditzel, missing slice, expansion/contraction, extra segment and mislabeling) to mimic common segmentation errors. Three experienced radiation oncology residents reviewed the test sets to identify segmentation errors, first without; and then one week after with the help of the QA tool. Each reviewer classified findings as minor or major errors. For the purpose of analysis, a minor or major error was defined if 2 or more reviewers identified the exact same error during either trial. Of the 34 manipulations, reviewers identified 8 as minor and 23 as major errors. Three manipulations were not classified as errors by the reviewers. Reviewers identified an additional 30 minor and 11 major errors in the OAR structures. Overall, the QA tool detected 37% and 85% of the minor and major errors respectively. The tool sensitivity for specific error types was 47% for boundary errors in the axial plane, 25% for slice extent errors, and 100% for ditzel, missing slice, extra segment and mislabeling. The QA tool improved the user error detection sensitivity from 61% to 68% (p = 0.17) for minor errors, and from 78% to 87% (p = 0.02) for major errors. Table 1 summarizes the error detection sensitivity by the QA tool and the reviewers. The proposed QA tool can assist users in detection of segmentation errors. Our experiment demonstrated that it can significantly improve detection of major errors. The QA tool’s ability to reduce the frequency of OAR segmentation error can potentially improve safety and quality of radiation therapy.Abstract 3317; Table 1Error detection sensitivity by the QA tool and reviewersQA tool (%)Reviewers - w/o QA tool (%)Reviewers - w/ QA tool (%)Minor ErrorAutomatic segmentation (AS)676789Propagated segmentation (PS)08164Manual segmentation w/ manipulation (MS)564465Total376168Major ErrorAS809393PS2510092MS967488Total858089 Open table in a new tab

Meeting Report: Tissue-based Image Analysis

Quantitative image analysis (IA) is a rapidly evolving area of digital pathology. Although not a new concept, the quantification of histological features on photomicrographs used to be cumbersome, resource-intensive, and limited to specialists and specialized laboratories. Recent technological advances like highly efficient automated whole slide digitizer (scanner) systems, innovative IA platforms, and the emergence of pathologist-friendly image annotation and analysis systems mean that quantification of features on histological digital images will become increasingly prominent in pathologists' daily professional lives. The added value of quantitative IA in pathology includes confirmation of equivocal findings noted by a pathologist, increasing the sensitivity of feature detection, quantification of signal intensity, and improving efficiency. There is no denying that quantitative IA is part of the future of pathology; however, there are also several potential pitfalls when trying to estimate volumetric features from limited 2-dimensional sections. This continuing education session on quantitative IA offered a broad overview of the field; a hands-on toxicologic pathologist experience with IA principles, tools, and workflows; a discussion on how to apply basic stereology principles in order to minimize bias in IA; and finally, a reflection on the future of IA in the toxicologic pathology field.

Computed-tomography volumetric study of the ethmoid labyrinth

Abstract Morphological variability of paranasal sinuses is well known for endoscopic surgeons and anatomists alike. The ethmoid sinus is the most complex and variable of all paranasal sinuses, due to the fact that its development is not yet well known and is influenced by many factors. Volumetric studies of the sinuses have been made using dried skulls, cadaver heads and imaging studies, but there are still not sufficient data in order to name a standard value for each sinus. Few data can be found especially regarding the ethmoid sinus. In this paper, we measured the volumes of ethmoid lateral masses, and for anterior and posterior groups of cells, using imaging studies and a volumetric feature of our imaging studies. Results showed an average volume between 7.34 cm3 and 8.39 cm3 for the ethmoid lateral mass, between 4.33 cm3 and 4.92 cm3 for the anterior ethmoid and between 3.01 cm3 and 3.47 cm3 for the posterior ethmoid groups. We also found that the average volume of the anterior ethmoid occupies between 58-59% of the whole volume, while the posterior ethmoid occupies only 41-42% of this volume.

Multiple sclerosis exploration based on automatic MRI modalities segmentation approach with advanced volumetric evaluations for essential feature extraction

Abstract Multiple Sclerosis (MS) could be considered as one of the most serious neurological diseases that can cause damage to the central nervous system. Such pathology has increased dramatically during the past few years. Hence, MS exploration has captivated the interest of various research studies in clinical as well as technological fields such as medical imaging. In this context, this paper introduced a new MS exploration approach based on cerebral segmentation and MS lesion identification using the fusion of magnetic resonance (MRI) modalities sequences. The proposed segmentation approach is based on extracted volumetric features that could be deduced from the gray-level co-occurrence matrix (GLCM) and the gray-level run length (GLRLM) matrix. Volumetric features extraction would be performed by using new voxel wise techniques while preserving connectivity, spatial and shape information. In addition, our segmentation approach includes an optimized feature selection process combining the genetic algorithm (GA) and the support vector machine (SVM) tool in order to preserve only the essential features that could distinguish the main brain tissues and the MS lesions within both white matter and gray matter. The evaluation was carried out on four clinical databases. The results revealed an acceptable conformity with the ground truths compared to those of the usual methods Besides, our approach has proved its ability to select the most discriminative features, ensuring an acceptable cerebral segmentation (averages: Dice = 0.62 ± 0.11, true positive rate ‘TPR’ = 0.64 ± 0.12 and positive predictive value ‘PPV’ = 0.64 ± 0.14) and MS lesions identification with an acceptable accuracy rate (averages: Dice = 0.66 ± 0.07, TPR = 0.70 ± 0.12 and PPV = 0.67 ± 0.03). Based on these promising results, a computer aided diagnosis (CAD) system was henceforth conceived and could be useful for clinicians in order to carefully facilitate MS exploration. Such a helpful CAD system was really highly needed for clinical explorations and could be extended to other neurological pathologies such as Alzheimer’s and Parkinson’s diseases.

Random cutting plane approach for identifying volumetric features in a CAD mesh model

This paper presents a method to identify regions that make up features like holes, slots, pockets as well as interacting features in a three-dimensional mesh of a computer-aided design (CAD)/Engineering model. Feature recognition is an important area in the field of CAD/Engineering with applications in model retrieval, creating an analysis model by defeaturing of the designed model for finite element applications, etc. Most feature recognition methods use either a cluster-based decomposition or feature line extraction through solid angles or curvature values, followed by graph-based heuristics. Such approaches require a user parameter for clustering or a threshold value for angle/curvature, neither of which is an easily deterministic one. The proposed algorithm identifies the features using contours generated by random cutting planes, followed by graph traversals (and not using heuristics) and without using parameter/threshold values. The algorithm can identify blind holes, through holes, slots and pockets. The geometry of most of the extracted features has also been identified using Gauss map. Interacting features have also been extracted and separated, which normally pose difficulty for most algorithms. Extensive experiments on CAD models from various benchmarks show that the algorithm is robust. Comparison with different algorithms (of which code was available) shows that our approach performs admirably and in the case of interacting features, the algorithm performs better than the existing ones.

Three Dimensional Image Processing in Hexagonal Prism Lattice of Z3 Grid

A 2D hexagonal image is an array of what are called pixels which are the coordinates of the hexagonal lattice points decided by the linear horizontal rows and the nonlinear vertical zig-zags. A 2D hexagonal image could also be informally called as image slice or a matrix of pixel values arranged in a hexagonal array. A 3D hexagonal image is viewed as an ordered sequence of 2D hexagonal image slices arranged in the z-direction and the 3D arrangement of voxel values is called as a prism of voxel values. Most of the 3D hexagonal image processing operations are similar to those of 2D hexagonal image processing. 3D hexagonal images are processed with the help of 3D hexagonal scanning windows, whereas 2D hexagonal images are processed with the help of 2D hexagonal scanning windows. For instance, a 3D hexagonal image processing operation like 3D surface detection is carried out using analogous 2D edge detection algorithm on every image slice and the processed slices assembled to visualize 3D surface detected image. In fact, 2D contours of an image slice are called superficial features and closed surfaces of a 3D image are called volumetric features. One can always obtain surface detected version of a 3D hexagonal image by processing the 2D hexagonal image slices using 2D edge detection operation, and consequently the 3D surface detection operation is termed as 2.5D hexagonal image processing. One could also process the 3D hexagonal image data using a 3D surface detection algorithm, in which case it is termed as 3D hexagonal image processing. This is not the case with the operation of skeltonization. One cannot make use of 2.5D skeletonization operation of 2D image slices in order to get skeltonized version of the corresponding 3D image. In fact, one would come across discrepancies and differences when 2.5D skeletonization of 2D hexagonal image slices of a 3D hexagonal image is carried out instead of the direct 3D skeletonization of the 3D hexagonal image. This paper highlights certain 3D algorithms for processing 3D hexagonal images.

CASPER: computer-aided segmentation of imperceptible motion-a learning-based tracking of an invisible needle in ultrasound.

This paper presents a new micro-motion-based approach to track a needle in ultrasound images captured by a handheld transducer. We propose a novel learning-based framework to track a handheld needle by detecting microscale variations of motion dynamics over time. The current state of the art on using motion analysis for needle detection uses absolute motion and hence work well only when the transducer is static. We have introduced and evaluated novel spatiotemporal and spectral features, obtained from the phase image, in a self-supervised tracking framework to improve the detection accuracy in the subsequent frames using incremental training. Our proposed tracking method involves volumetric feature selection and differential flow analysis to incorporate the neighboring pixels and mitigate the effects of the subtle tremor motion of a handheld transducer. To evaluate the detection accuracy, the method is tested on porcine tissue in-vivo, during the needle insertion in the biceps femoris muscle. Experimental results show the mean, standard deviation and root-mean-square errors of [Formula: see text], [Formula: see text] and [Formula: see text] in the insertion angle, and 0.82, 1.21, 1.47mm, in the needle tip, respectively. Compared to the appearance-based detection approaches, the proposed method is especially suitable for needles with ultrasonic characteristics that are imperceptible in the static image and to the naked eye.

Alzheimer disease classification using KPCA, LDA, and multi‐kernel learning SVM

AbstractEarly diagnosis of Alzheimer disease (AD) and mild cognitive impairment (MCI) is always useful. Preventive measures might have an impact on reducing AD risk factors. Structural magnetic resonance (MR) imaging, one of the vital sensitive biomarkers for cerebral atrophy in the brain, is used to extract volumetric feature by FreeSurfer and the CIVET toolbox. All of the structural magnetic resonance imaging (s‐MRI) data that we used were downloaded from the Alzheimer's disease Neuroimaging Initiative (ADNI) database (adni.loni.usc.edu) of imaging data. This novel approach is applied for the diagnosis of AD and MCI from healthy controls (HCs) combining extracted features with the MMSE (mini‐mental state examination) scores, applying a two sample t‐test to select a subset of features. The subset of features is fed to kernel principal component analysis (KPCA) module to project data onto the reduced principal component coefficients at higher dimensional kernel space to increase the linear separability. Then, the kernel PCA coefficients are projected into the more efficient linear discriminant space using linear discriminant analysis. A multi‐kernel learning support vector machine (SVM) is used on newly projected data for stratification of AD and MCI from HCs. Using this approach, we obtain 93.85% classification accuracy when detecting AD from HCs for segmented volumetric features (using FreeSurfer) with high sensitivity and specificity. When distinguishing MCI from HCs and AD using volumetric features after subcortical segmentation, the detection rate reaches 86.54% and 75.12%, respectively.

Bergmann's and Allen's Rules in Native European and Mediterranean Phasmatodea

Bergmann’s rule states that organisms at higher latitudes should be larger and thicker than those closer to the equator to better conserve heat, and Allen’s rule states that they will have shorter and thicker limbs at higher latitudes. Alternative explanations for latitudinal size clines include plant productivity and seasonality. The rules generally hold in endotherms, but in insects different species within the same genus can respond to latitude in unpredictable ways. We present the first biogeographical analysis of these rules in stick insects (order Phasmatodea), using four European species. Their long and thin bauplan makes the Phasmatodea ideal for ecomorphological studies of body length, which could identify the evolutionary drivers of their remarkable size range (including the world’s longest insects). Using preserved specimens from collections across Europe; body segment and limb measurements were taken for both genders of the species Bacillus rossius, Clonopsis gallica, Leptynia attenuata, and Pijnackeria hispanica. Lengths and volumetric features were compared to latitude as well as annual mean temperature, net primary productivity, and annual growing degree days, using weighted linear regressions and ANOVA analyses. At lower latitudes / higher temperatures, B. rossius and L. attenuata had longer limbs [Allen clines] and were larger bodied and/or longer [converse-Bergmann clines], while the other species did not show latitudinal clines per se. This matches what was predicted based on closely related insects and the presence of large Phasmatodea in the tropics, but violates the temperature-size rule. Most variation in size could be attributed to temperature, but untested factors could also play a role Whether these ecogeographic rules hold true for tropical Phasmatodea and whether genetics or environment play are more important in determining adult length are topics for future research.

A design tool for direct and non-stochastic calculations of near-field radiative transfer in complex structures: The NF-RT-FDTD algorithm

Advances in nanotechnology and nanophotonics are inextricably linked with the need for reliable computational algorithms to be adapted as design tools for the development of new concepts in energy harvesting, radiative cooling, nanolithography and nano-scale manufacturing, among others. In this paper, we provide an outline for such a computational tool, named NF-RT-FDTD, to determine the near-field radiative transfer between structured surfaces using Finite Difference Time Domain method. NF-RT-FDTD is a direct and non-stochastic algorithm, which accounts for the statistical nature of the thermal radiation and is easily applicable to any arbitrary geometry at thermal equilibrium. We present a review of the fundamental relations for far- and near-field radiative transfer between different geometries with nano-scale surface and volumetric features and gaps, and then we discuss the details of the NF-RT-FDTD formulation, its application to sample geometries and outline its future expansion to more complex geometries. In addition, we briefly discuss some of the recent numerical works for direct and indirect calculations of near-field thermal radiation transfer, including Scattering Matrix method, Finite Difference Time Domain method (FDTD), Wiener Chaos Expansion, Fluctuating Surface Current (FSC), Fluctuating Volume Current (FVC) and Thermal Discrete Dipole Approximations (TDDA).

Identifying relations between imaging phenotypes and molecular subtypes of breast cancer: Model discovery and external validation.

To determine whether dynamic contrast enhancement magnetic resonance imaging (DCE-MRI) characteristics of the breast tumor and background parenchyma can distinguish molecular subtypes (ie, luminal A/B or basal) of breast cancer. In all, 84 patients from one institution and 126 patients from The Cancer Genome Atlas (TCGA) were used for discovery and external validation, respectively. Thirty-five quantitative image features were extracted from DCE-MRI (1.5 or 3T) including morphology, texture, and volumetric features, which capture both tumor and background parenchymal enhancement (BPE) characteristics. Multiple testing was corrected using the Benjamini-Hochberg method to control the false-discovery rate (FDR). Sparse logistic regression models were built using the discovery cohort to distinguish each of the three studied molecular subtypes versus the rest, and the models were evaluated in the validation cohort. On univariate analysis in discovery and validation cohorts, two features characterizing tumor and two characterizing BPE were statistically significant in separating luminal A versus nonluminal A cancers; two features characterizing tumor were statistically significant for separating luminal B; one feature characterizing tumor and one characterizing BPE reached statistical significance for distinguishing basal (Wilcoxon P < 0.05, FDR < 0.25). In discovery and validation cohorts, multivariate logistic regression models achieved an area under the receiver operator characteristic curve (AUC) of 0.71 and 0.73 for luminal A cancer, 0.67 and 0.69 for luminal B cancer, and 0.66 and 0.79 for basal cancer, respectively. DCE-MRI characteristics of breast cancer and BPE may potentially be used to distinguish among molecular subtypes of breast cancer. 3 Technical Efficacy: Stage 3 J. Magn. Reson. Imaging 2017;46:1017-1027.

Extraction of volumetric features using 2.5D and 3D algorithms

A 3D image processing operation like 3D surface detection is carried out using 2D edge detection algorithm on every slice and the processed slices assembled to visualise 3D surface detected image. One can obtain surface detected version of a 3D image by processing the 2D slices using 2D edge detection operation, and consequently the 3D surface detection operation is termed as 2.5D image processing. One could also process the 3D image data using a 3D surface detection algorithm, which is termed as 3D processing. This is not the case with the operation of skeletonisation. One cannot make use of 2.5D skeletonisation operation of 2D slices in order to get skeltonised version of the corresponding 3D image. This paper discusses the discrepancies faced when 2.5D skeletonisation and edge detection of 2D slices of a 3D image is carried out instead of the direct 3D operations of the 3D image.

Extraction of Volumetric Features Using 2.5D and 3D Algorithms

A 3D image processing operation like 3D surface detection is carried out using 2D edge detection algorithm on every slice and the processed slices assembled to visualise 3D surface detected image. One can obtain surface detected version of a 3D image by processing the 2D slices using 2D edge detection operation, and consequently the 3D surface detection operation is termed as 2.5D image processing. One could also process the 3D image data using a 3D surface detection algorithm, which is termed as 3D processing. This is not the case with the operation of skeletonisation. One cannot make use of 2.5D skeletonisation operation of 2D slices in order to get skeltonised version of the corresponding 3D image. This paper discusses the discrepancies faced when 2.5D skeletonisation and edge detection of 2D slices of a 3D image is carried out instead of the direct 3D operations of the 3D image.

Towards a Holistic Cortical Thickness Descriptor: Heat Kernel-Based Grey Matter Morphology Signatures

In this paper, we propose a heat kernel based regional shape descriptor that may be capable of better exploiting volumetric morphological information than other available methods, thereby improving statistical power on brain magnetic resonance imaging (MRI) analysis. The mechanism of our analysis is driven by the graph spectrum and the heat kernel theory, to capture the volumetric geometry information in the constructed tetrahedral meshes. In order to capture profound brain grey matter shape changes, we first use the volumetric Laplace-Beltrami operator to determine the point pair correspondence between white-grey matter and CSF-grey matter boundary surfaces by computing the streamlines in a tetrahedral mesh. Secondly, we propose multi-scale grey matter morphology signatures to describe the transition probability by random walk between the point pairs, which reflects the inherent geometric characteristics. Thirdly, a point distribution model is applied to reduce the dimensionality of the grey matter morphology signatures and generate the internal structure features. With the sparse linear discriminant analysis, we select a concise morphology feature set with improved classification accuracies. In our experiments, the proposed work outperformed the cortical thickness features computed by FreeSurfer software in the classification of Alzheimer's disease and its prodromal stage, i.e., mild cognitive impairment, on publicly available data from the Alzheimer's Disease Neuroimaging Initiative. The multi-scale and physics based volumetric structure feature may bring stronger statistical power than some traditional methods for MRI-based grey matter morphology analysis.

Auto-recognition and generation of material removal volume for regular form surface and its volumetric features using volume decomposition method

Computer-aided process planning (CAPP) plays a vital role in manufacturing of a product by computer-aided techniques. Auto-recognition of a surface’s volumetric features and segregation of auto-generated material removal volume into feature volumes is necessary to achieve successful downstream manufacturing activities like automated process planning and scheduling. Only few studies generated material removal volume for a product but resulted in drawbacks such as material removal volume discontinuity between two adjacent material removal volumes generated from two adjacent faces that form convex geometry, and material removal volume overlap between two adjacent material removal volumes generated from two adjacent faces that form concave geometry. An effort has been made to newly classify the regular form surface features and develop an algorithm that automatically recognizes regular form computer-aided design (CAD) model’s surface and their volumetric features, and also auto-generate material removal volume for the recognized features. The surface and volumetric features of a CAD model are successfully recognized by the developed algorithm, and estimated material removal volume of these features is obtained from stock model. The material removal volume discontinuity and overlap limitation that occurred in few studies are eliminated. The percentage difference in overall delta volume (ODV) obtained by algorithm and by manual calculations is less than 0.01 %. Exploded view of auto-generated material removal volumes is obtained for subsequent process planning applications.

Identification of volumetric laser endomicroscopy features predictive for early neoplasia in Barrett’s esophagus using high-quality histological correlation

Volumetric laser endomicroscopy (VLE) provides a circumferential scan that enables visualization of the subsurface layers of the esophageal wall at 7 μm resolution. The aims of this study were to identify VLE features of Barrett's esophagus (BE) neoplasia and to develop a VLE prediction score. A database of VLE images from endoscopic resection specimens, precisely correlated with histology, from patients with BE with and without neoplasia was used. Features potentially predictive for early BE neoplasia were identified by unblinded evaluation of 25 VLE-histology images. In a learning phase, 20 VLE images with or without BE neoplasia were scored by 2 VLE experts, blinded to histology. A prediction score was created by using multivariable logistic regression analyses and validated by scoring 40 VLE images (50% neoplastic) by using area under receiver operating characteristic (ROC) curve (AUC) analysis. Three VLE features independently predictive for BE neoplasia were identified: (1) lack of layering; (2) higher surface than subsurface signal; (3) presence of irregular, dilated glands/ducts. A VLE neoplasia prediction score was developed with the following: (1) 6 points; (2) 6 or 8 points for equal or higher surface signal; and (3) 5 points. The ROC curve of this prediction score showed an AUC of 0.81 (95% confidence interval, 0.71-0.90). A cut-off value of≥8 was associated with sensitivity and specificity of 83% and 71%, respectively. When high-quality exvivo VLE-histology correlation was used, the VLE features of layering, surface signal, and irregular glands/ducts were independently and significantly associated with BE neoplasia. A VLE prediction score for BE neoplasia was developed and validated, with promising accuracy. (Clinical trial registration number: NCT01862666.).

Imaging-genomics reveals driving pathways of MRI derived volumetric tumor phenotype features in Glioblastoma.

BackgroundGlioblastoma (GBM) tumors exhibit strong phenotypic differences that can be quantified using magnetic resonance imaging (MRI), but the underlying biological drivers of these imaging phenotypes remain largely unknown. An Imaging-Genomics analysis was performed to reveal the mechanistic associations between MRI derived quantitative volumetric tumor phenotype features and molecular pathways.MethodsOne hundred fourty one patients with presurgery MRI and survival data were included in our analysis. Volumetric features were defined, including the necrotic core (NE), contrast-enhancement (CE), abnormal tumor volume assessed by post-contrast T1w (tumor bulk or TB), tumor-associated edema based on T2-FLAIR (ED), and total tumor volume (TV), as well as ratios of these tumor components. Based on gene expression where available (n = 91), pathway associations were assessed using a preranked gene set enrichment analysis. These results were put into context of molecular subtypes in GBM and prognostication.ResultsVolumetric features were significantly associated with diverse sets of biological processes (FDR < 0.05). While NE and TB were enriched for immune response pathways and apoptosis, CE was associated with signal transduction and protein folding processes. ED was mainly enriched for homeostasis and cell cycling pathways. ED was also the strongest predictor of molecular GBM subtypes (AUC = 0.61). CE was the strongest predictor of overall survival (C-index = 0.6; Noether test, p = 4x10−4).ConclusionGBM volumetric features extracted from MRI are significantly enriched for information about the biological state of a tumor that impacts patient outcomes. Clinical decision-support systems could exploit this information to develop personalized treatment strategies on the basis of noninvasive imaging.Electronic supplementary materialThe online version of this article (doi:10.1186/s12885-016-2659-5) contains supplementary material, which is available to authorized users.

MO-DE-207B-08: Radiomic CT Features Complement Semantic Annotations to Predict EGFR Mutations in Lung Adenocarcinomas

Purpose: To compare the complementary value of quantitative radiomic features to that of radiologist-annotated semantic features in predicting EGFR mutations in lung adenocarcinomas. Methods: Pre-operative CT images of 258 lung adenocarcinoma patients were available. Tumors were segmented using the sing-click ensemble segmentation algorithm. A set of radiomic features was extracted using 3D-Slicer. Test-retest reproducibility and unsupervised dimensionality reduction were applied to select a subset of reproducible and independent radiomic features. Twenty semantic annotations were scored by an expert radiologist, describing the tumor, surrounding tissue and associated findings. Minimum-redundancy-maximum-relevance (MRMR) was used to identify the most informative radiomic and semantic features in 172 patients (training-set, temporal split). Radiomic, semantic and combined radiomic-semantic logistic regression models to predict EGFR mutations were evaluated in and independent validation dataset of 86 patients using the area under the receiver operating curve (AUC). Results: EGFR mutations were found in 77/172 (45%) and 39/86 (45%) of the training and validation sets, respectively. Univariate AUCs showed a similar range for both feature types: radiomics median AUC = 0.57 (range: 0.50 – 0.62); semantic median AUC = 0.53 (range: 0.50 – 0.64, Wilcoxon p = 0.55). After MRMR feature selection, the best-performing radiomic, semantic, and radiomic-semantic logistic regression models, for EGFR mutations, showed a validation AUC of 0.56 (p = 0.29), 0.63 (p = 0.063) and 0.67 (p = 0.004), respectively. Conclusion: Quantitative volumetric and textural Radiomic features complement the qualitative and semi-quantitative radiologist annotations. The prognostic value of informative qualitative semantic features such as cavitation and lobulation is increased with the addition of quantitative textural features from the tumor region.

Multimodal imaging patterns predict survival in recurrent glioblastoma patients treated with bevacizumab.

Bevacizumab is a humanized antibody against vascular endothelial growth factor approved for treatment of recurrent glioblastoma. There is a need to discover imaging biomarkers that can aid in the selection of patients who will likely derive the most survival benefit from bevacizumab. The aim of the study was to examine if pre- and posttherapy multimodal MRI features could predict progression-free survival and overall survival (OS) for patients with recurrent glioblastoma treated with bevacizumab. The patient population included 84 patients in a training cohort and 42 patients in a testing cohort, separated based on pretherapy imaging date. Tumor volumes of interest were segmented from contrast-enhanced T1-weighted and fluid attenuated inversion recovery images and were used to derive volumetric, shape, texture, parametric, and histogram features. A total of 2293 pretherapy and 9811 posttherapy features were used to generate the model. Using standard radiographic assessment criteria, the hazard ratio for predicting OS was 3.38 (P < .001). The hazard ratios for pre- and posttherapy features predicting OS were 5.10 (P < .001) and 3.64 (P < .005) for the training and testing cohorts, respectively. With the use of machine learning techniques to analyze imaging features derived from pre- and posttherapy multimodal MRI, we were able to develop a predictive model for patient OS that could potentially assist clinical decision making.

A slice based approach to recognize and extract free-form volumetric features in a CAD mesh model

ABSTRACTThis paper describes an algorithm to recognize and extract volumetric features by directly clustering the triangles constituting a feature in a CAD mesh model. The algorithm involves two steps – isolating features in 2D slices followed by a 3D traversal to cluster all the triangles in the feature. The main advantage of this approach is that processing of the three dimensional mesh occurs only for the triangles associated with its signature identified on the 2D slice. Results from an implementation tested on various mesh models are presented.