This paper investigates the efficacy of automated pattern recognition methods on magnetic resonance data with the objective of assisting radiologists in the clinical diagnosis of brain tissue tumors. In this paper, the sciences of magnetic resonance imaging (MRI) and magnetic resonance spectroscopy (MRS) are combined to improve the accuracy of the classifier, based on the multidimensional co-occurrence matrices to assess the detection of pathological tissues (tumor and edema), normal tissues (white matter — WM and gray matter — GM), and fluid (cerebrospinal fluid — CSF). The results show the ability of the classifier with iterative training to automatically and simultaneously recover tissue-specific spectral and structural patterns and achieve segmentation of tumor and edema and grading of high and low glioma tumor. Here, extreme learning machine – improved particle swarm optimization (ELM-IPSO) neural network classifier is trained with the feature descriptions in brain magnetic resonance (MR) spectra. This has the characteristics of varying the normal spectral pattern associated with tumor patterns along with imaging features. Validation was performed considering 35 clinical studies. The volumetric features extracted from the vectors of this matrix articulate some important elementary structures, which along with spectroscopic metabolite ratios discriminate the tumor grades and tissue classes. The quantitative 3D analysis reveals significant improvement in terms of global accuracy rate for automatic classification in brain tissues and discriminating pathological tumor tissue from structural healthy brain tissue.
Magnetic resonance spectroscopy; Magnetic resonance imaging; Multidimensional co-occurrence matrices; Feature extraction; Extreme learning machine; Particle swarm optimization
To analyze if an iPad-based patient briefing can serve as a digital alternative to conventional documentations prior to radiological examinations. One hundred one patients referred for routine MRI were randomized into two groups, who underwent iPad-based and classic written briefing in opposite order. For each briefing completion time, completeness and correctness were noted. Patient’s knowledge about the content of either briefing modality was subsequently tested. The influence of patient-related factors on the performance of the electronic briefing (EB) was analyzed. Finally, the patient’s subjective impression of the EB was assessed. The mean durations were 4.4 ± 2.2 min for EB and 1.7 ± 1.3 min for the classic briefing (p < 0.01). All iPad briefings were returned entirely filled out, whereas 11 % of the classic forms were returned with missing data. No significant differences in memorization of the briefing’s information were objectified. There was a positive correlation between the duration of EB and age (r = 0.53; p < 0.01), whereas a negative correlation was found between computer skills and patient’s age (r = −0.55; p < 0.01) or duration of EB (r = −0.62; p < 0.01). More than half of the study patients would prefer EB in the future; another 29 % had no preference at all. Patient briefing on iPads transfers the information for the patients equally well compared to the classic written approach. Although iPad briefing took patients longer to perform, the majority would prefer it to written consent briefings in the future. Nevertheless, measures have to be undertaken to improve the overall acceptance and performance.
Clinical application; Acceptance testing; Data collection; Electronic patient consent; Tablet PC
Picture archiving and communication systems (PACS) play a critical role in radiology. This paper presents the criteria important to PACS administrators for selecting a PACS. A set of criteria are identified and organized into an integrative hierarchical framework. Survey responses from 48 administrators are used to identify the relative weights of these criteria through an analytical hierarchy process. The five main dimensions for PACS selection in order of importance are system continuity and functionality, system performance and architecture, user interface for workflow management, user interface for image manipulation, and display quality. Among the subdimensions, the highest weights were assessed for security, backup, and continuity; tools for continuous performance monitoring; support for multispecialty images; and voice recognition/transcription. PACS administrators’ preferences were generally in line with that of previously reported results for radiologists. Both groups assigned the highest priority to ensuring business continuity and preventing loss of data through features such as security, backup, downtime prevention, and tools for continuous PACS performance monitoring. PACS administrators’ next high priorities were support for multispecialty images, image retrieval speeds from short-term and long-term storage, real-time monitoring, and architectural issues of compatibility and integration with other products. Thus, next to ensuring business continuity, administrators’ focus was on issues that impact their ability to deliver services and support. On the other hand, radiologists gave high priorities to voice recognition, transcription, and reporting; structured reporting; and convenience and responsiveness in manipulation of images. Thus, radiologists’ focus appears to be on issues that may impact their productivity, effort, and accuracy.
Picture archiving and communication system; PACS; Analytical hierarchy process; AHP; RIS; Structured reporting; Voice recognition; Transcription; Open systems; Proprietary systems; Display quality; System continuity; Security; Backup; Recovery; Downtime prevention; PACS performance monitoring; Configuration; Upgrade; Cardiology images; Pathology images; System architecture and performance; User interface for image manipulation; User interface workflow management; Worklist management
We tested the accuracy and efficiency of a novel automated program capable of extracting 15 cardiac computed tomography angiography (CTA) parameters from clinical CTA reports. Five hundred cardiac CTA reports were retrospectively collected and processed. All reports were pre-populated with a structured template per guideline. The program extracted 15 parameters with high accuracy (97.3 %) and efficiency (84 s). This program may be used at other institutions with similar accuracy if its report format follows the Society of Cardiovascular Computed Tomography (SCCT) guideline recommendation.
Algorithm; Database management; Data extraction; Efficiency; Radiation dose
This study presents a completely automated method for separating the left and right lungs using free-formed surface fitting on volumetric computed tomography (CT). The left and right lungs are roughly divided using iterative 3-dimensional morphological operator and a Hessian matrix analysis. A point set traversing between the initial left and right lungs is then detected with a Euclidean distance transform to determine the optimal separating surface, which is then modeled from the point set using a free-formed surface-fitting algorithm. Subsequently, the left and right lung volumes are smoothly and directly separated using the separating surface. The performance of the proposed method was estimated by comparison with that of a human expert on 44 CT examinations. For all data sets, averages of the root mean square surface distance, maximum surface distance, and volumetric overlap error between the results of the automatic and the manual methods were 0.032 mm, 2.418 mm, and 0.017 %, respectively. Our study showed the feasibility of automatically separating the left and right lungs by identifying the 3D continuous separating surface on volumetric chest CT images.
Left and right lung separation; Hessian matrix analysis; Euclidean distance transform; Free-formed surface fitting; Lung segmentation
Devising a method that can select cases based on the performance levels of trainees and the characteristics of cases is essential for developing a personalized training program in radiology education. In this paper, we propose a novel hybrid prediction algorithm called content-boosted collaborative filtering (CBCF) to predict the difficulty level of each case for each trainee. The CBCF utilizes a content-based filtering (CBF) method to enhance existing trainee-case ratings data and then provides final predictions through a collaborative filtering (CF) algorithm. The CBCF algorithm incorporates the advantages of both CBF and CF, while not inheriting the disadvantages of either. The CBCF method is compared with the pure CBF and pure CF approaches using three datasets. The experimental data are then evaluated in terms of the MAE metric. Our experimental results show that the CBCF outperforms the pure CBF and CF methods by 13.33 and 12.17 %, respectively, in terms of prediction precision. This also suggests that the CBCF can be used in the development of personalized training systems in radiology education.
Personalized radiology education; Making error predication; Content-based predictor; Collaborative filtering
Paging referring clinicians with imaging results is a frequently repeated “microtask” performed by practicing radiologists. Many institutions use online alpha-paging systems to provide this integral part of safe and efficient patient care. Although sending an alpha-page can often be accomplished within one minute, current tools may disrupt workflow by distracting users with a series of tedious mouse clicks. We describe the development, evaluation, and updates of a portable tool that sends alpha-pages to referring clinicians using two keystrokes. This software integrates study information obtained from the Picture Archiving and Communication System (PACS) with an existing hospital paging system.
Reading Room; Workflow reengineering; PACS integration
Radiologists are adept at recognizing the character and extent of lung parenchymal abnormalities in computed tomography (CT) scans. However, the inconsistent differential diagnosis due to subjective aggregation necessitates the exploration of automated classification based on supervised or unsupervised learning. The robustness of supervised learning depends on the training samples. Towards optimizing emphysema classification, we introduce a physician-in-the-loop feedback approach to minimize ambiguity in the selected training samples. An experienced thoracic radiologist selected 412 regions of interest (ROIs) across 15 datasets to represent 124, 129, 139 and 20 training samples of mild, moderate, severe emphysema and normal appearance, respectively. Using multi-view (multiple metrics to capture complementary features) inductive learning, an ensemble of seven un-optimized support vector models (SVM) each based on a specific metric was constructed in less than 6 s. The training samples were classified using seven SVM models and consensus labels were created using majority voting. In the active relearning phase, the ensemble-expert label conflicts were resolved by the expert. The efficacy and generality of active relearning feedback was assessed in the optimized parameter space of six general purpose classifiers across the seven dissimilarity metrics. The proposed just-in-time active relearning feedback with un-optimized SVMs yielded 15 % increase in classification accuracy and 25 % reduction in the number of support vectors. The average improvement in accuracy of six classifiers in their optimized parameter space was 21 %. The proposed cooperative feedback method enhances the quality of training samples used to construct automated classification of emphysematous CT scans. Such an approach could lead to substantial improvement in quantification of emphysema.
Emphysema; Supervised classification; Support Vector Machines (SVM); Active relearning; Training sample cleaning
Visual content in biomedical academic papers is a growing source of critical information, but it is not always fully readable for people with visual impairments. We aimed to assess current image processing practices, accessibility policies, and submission policies in a sample of 12 highly cited biomedical journals. We manually checked the application of text-based alternative image descriptions for every image in 12 articles (one for each journal). We determined whether the journals claimed to follow an accessibility policy and we reviewed their submission policy and their guidelines related to the visual content. We identified important features concerning the processing of images and the characteristics of the visual and the retrieval options of visual content offered by the publishers. The evaluation shows that the actual practices of textual image description in highly cited biomedical journals do not follow general guidelines on accessibility. The images within the articles analyzed lack alternative descriptions or have uninformative descriptions, even in the case of journals claiming to follow an accessibility policy. Consequently, the visual information of scientific articles is not accessible to people with severe visual disabilities. Instructions on image submission are heterogeneous and a declaration of accessibility guidelines was only found in two thirds of the sample of journals, with one third not explicitly following any accessibility policy, although they are required to by law.
Medical images; Publishing; Biomedical journals; Accessibility policies; Image description; Alternative text; Visual impairment; Disabilities; Publications
Providing patients and clinicians with self-contained PACS viewer on CD format is a common and necessary tool to share vital imaging data. However, to be useful, this tool should be reliable, robust, and convenient. Numerous PACS viewer options are available, often without empirical data to guide in choosing one for routine use. To assist in making a standardized choice for our institution, we chose four common viewers, benchmarked on four different workstations reflecting the variety of environments used by non-radiologist clinicians who would receive a CD. Four CD-based DICOM viewers from eFilm, Philips, Pacsgear Gearview, and iSite were examed on two radiology PACS workstations, a standard desktop computer, and a laptop using a test case consisting of a multi-series CTA with 13 series and 3,035 total images. Multiple objective measures, subjective measures, and presence of key features were evaluated including program time to load, image time to load, cine/movie mode, ability to adequately window and level, pan and zoom functionality, basic measurement tools, and perceived lag when scrolling through a multi-image series. Substantial differences in speed of operation and behavior on multiple systems were documented, which could potentially add several minutes to the time required to open and view a patient’s imaging data. The eFilm and iSite viewers operated consistently and reliably across all tested computer environments. The iSite viewer, having among the quickest load times in the group tested and consistently low subjective scroll lag during series viewing, and also beneficially allowing partial viewing while images load in the background, was found to generate the best overall user experience. Because of these significant differences, we have recommended that our institution standardize all patient imaging CD creation using the iSite viewer.
PACS; Clinical workflow; Clinical image viewing; Computers in medicine; Computer graphics; Data display; PACS system performance
In the transition from paper to electronic workflow, the University of Colorado Health System’s implementation of a new electronic health record system (EHR) forced all clinical groups to reevaluate their practices including the infrastructure surrounding clinical trials. Radiological imaging is an important piece of many clinical trials and requires a high level of consistency and standardization. With EHR implementation, paper orders were manually transcribed into the EHR, digitizing an inefficient work flow. A team of schedulers, radiologists, technologists, research personnel, and EHR analysts worked together to optimize the EHR to accommodate the needs of research imaging protocols. The transition to electronic workflow posed several problems: (1) there needed to be effective communication throughout the imaging process from scheduling to radiologist interpretation. (2) The exam ordering process needed to be automated to allow scheduling of specific research studies on specific equipment. (3) The billing process needed to be controlled to accommodate radiologists already supported by grants. (4) There needed to be functionality allowing exams to finalize automatically skipping the PACS and interpretation process. (5) There needed to be a way to alert radiologists that a specialized research interpretation was needed on a given exam. These issues were resolved through the optimization of the “visit type,” allowing a high-level control of an exam at the time of scheduling. Additionally, we added columns and fields to work queues displaying grant identification numbers. The build solutions we implemented reduced the mistakes made and increased imaging quality and compliance.
Clinical research; Imaging informatics; Workflow; Electronic medical record
Stereology is a volume estimation method, typically applied to diagnostic imaging examinations in population studies where planimetry is too time-consuming (Chapman et al. Kidney Int 64:1035–1045, 2003), to obtain quantitative measurements (Nyengaard J Am Soc Nephrol 10:1100–1123, 1999, Michel and Cruz-Orive J Microsc 150:117–136, 1988) of certain structures or organs. However, true segmentation is required in order to perform advanced analysis of the tissues. This paper describes a novel method for segmentation of region(s) of interest using stereology data as prior information. The result is an efficient segmentation method for structures that cannot be easily segmented using other methods.
3D segmentation; Digital image processing; Biomedical image analysis; Fuzzy logic; Image segmentation; 3D imaging (three-dimensional imaging); Boundary extraction; Segmentation; Magnetic resonance imaging; MR imaging; Data extraction; Image analysis; Polycystic kidney disease; Python; Planimetry
We adapted and evaluated the Microsoft Kinect (touchless interface), Hillcrest Labs Loop Pointer (gyroscopic mouse), and the Apple iPad (multi-touch tablet) for intra-procedural imaging review efficacy in a simulation using MIM Software DICOM viewers. Using each device, 29 radiologists executed five basic interactions to complete the overall task of measuring an 8.1-cm hepatic lesion: scroll, window, zoom, pan, and measure. For each interaction, participants assessed the devices on a 3-point subjective scale (3 = highest usability score). The five individual scores were summed to calculate a subjective composite usability score (max 15 points). Overall task time to completion was recorded. Each user also assessed each device for its potential to jeopardize a sterile field. The composite usability scores were as follows: Kinect 9.9 (out of 15.0; SD = 2.8), Loop Pointer 12.9 (SD = 13.5), and iPad 13.5 (SD = 1.8). Mean task completion times were as follows: Kinect 156.7 s (SD = 86.5), Loop Pointer 51.5 s (SD = 30.6), and iPad 41.1 s (SD = 25.3). The mean hepatic lesion measurements were as follows: Kinect was 7.3 cm (SD = 0.9), Loop Pointer 7.8 cm (SD = 1.1), and iPad 8.2 cm (SD = 1.2). The mean deviations from true hepatic lesion measurement were as follows: Kinect 1.0 cm and for both the Loop Pointer and iPad, 0.9 cm (SD = 0.7). The Kinect had the least and iPad had the most subjective concern for compromising the sterile field. A new intra-operative imaging review interface may be near. Most surveyed foresee these devices as useful in procedures, and most do not anticipate problems with a sterile field. An ideal device would combine iPad’s usability and accuracy with the Kinect’s touchless aspect.
Electronic supplementary material
The online version of this article (doi:10.1007/s10278-014-9687-y) contains supplementary material, which is available to authorized users.
Computer interface; Gesture-based input device; Intra-procedural
The current technologies that trend in digital radiology (DR) are toward systems using portable smart mobile as patient-centered care. We aimed to develop a mini-mobile DR system by using smart devices for wireless connection into medical information systems. We developed a mini-mobile DR system consisting of an X-ray source and a Complementary Metal–Oxide Semiconductor (CMOS) sensor based on a flat panel detector for small-field diagnostics in patients. It is used instead of the systems that are difficult to perform with a fixed traditional device. We also designed a method for embedded systems in the development of portable DR systems. The external interface used the fast and stable IEEE 802.11n wireless protocol, and we adapted the device for connections with Picture Archiving and Communication System (PACS) and smart devices. The smart device could display images on an external monitor other than the monitor in the DR system. The communication modules, main control board, and external interface supporting smart devices were implemented. Further, a smart viewer based on the external interface was developed to display image files on various smart devices. In addition, the advantage of operators is to reduce radiation dose when using remote smart devices. It is integrated with smart devices that can provide X-ray imaging services anywhere. With this technology, it can permit image observation on a smart device from a remote location by connecting to the external interface. We evaluated the response time of the mini-mobile DR system to compare to mobile PACS. The experimental results show that our system outperforms conventional mobile PACS in this regard.
Digital radiography; Web technology; Medical devices; Portable DR; Smart device; Wireless communication
A neural network ensemble (NNE) based computer-aided diagnostic (CAD) system to assist radiologists in differential diagnosis between focal liver lesions (FLLs), including (1) typical and atypical cases of Cyst, hemangioma (HEM) and metastatic carcinoma (MET) lesions, (2) small and large hepatocellular carcinoma (HCC) lesions, along with (3) normal (NOR) liver tissue is proposed in the present work. Expert radiologists, visualize the textural characteristics of regions inside and outside the lesions to differentiate between different FLLs, accordingly texture features computed from inside lesion regions of interest (IROIs) and texture ratio features computed from IROIs and surrounding lesion regions of interests (SROIs) are taken as input. Principal component analysis (PCA) is used for reducing the dimensionality of the feature space before classifier design. The first step of classification module consists of a five class PCA-NN based primary classifier which yields probability outputs for five liver image classes. The second step of classification module consists of ten binary PCA-NN based secondary classifiers for NOR/Cyst, NOR/HEM, NOR/HCC, NOR/MET, Cyst/HEM, Cyst/HCC, Cyst/MET, HEM/HCC, HEM/MET and HCC/MET classes. The probability outputs of five class PCA-NN based primary classifier is used to determine the first two most probable classes for a test instance, based on which it is directed to the corresponding binary PCA-NN based secondary classifier for crisp classification between two classes. By including the second step of the classification module, classification accuracy increases from 88.7 % to 95 %. The promising results obtained by the proposed system indicate its usefulness to assist radiologists in differential diagnosis of FLLs.
Focal liver lesions; B-mode ultrasound; Texture analysis; Neural network ensemble; Computer-aided diagnostic system; Principal component analysis
Owing to large financial investments that go along with the picture archiving and communication system (PACS) deployments and inconsistent PACS performance evaluations, there is a pressing need for a better understanding of the implications of PACS deployment in hospitals. We claim that there is a gap in the research field, both theoretically and empirically, to explain the success of the PACS deployment and maturity in hospitals. Theoretical principles are relevant to the PACS performance; maturity and alignment are reviewed from a system and complexity perspective. A conceptual model to explain the PACS performance and a set of testable hypotheses are then developed. Then, structural equation modeling (SEM), i.e. causal modeling, is applied to validate the model and hypotheses based on a research sample of 64 hospitals that use PACS, i.e. 70 % of all hospitals in the Netherlands. Outcomes of the SEM analyses substantiate that the measurements of all constructs are reliable and valid. The PACS alignment—modeled as a higher-order construct of five complementary organizational dimensions and maturity levels—has a significant positive impact on the PACS performance. This result is robust and stable for various sub-samples and segments. This paper presents a conceptual model that explains how alignment in deploying PACS in hospitals is positively related to the perceived performance of PACS. The conceptual model is extended with tools as checklists to systematically identify the improvement areas for hospitals in the PACS domain. The holistic approach towards PACS alignment and maturity provides a framework for clinical practice.
Picture archiving and communication systems; PACS maturity model; Performance; Complexity theory; Strategic planning; Structural equation modeling
Fusion of CT and MR images allows simultaneous visualization of details of bony anatomy provided by CT image and details of soft tissue anatomy provided by MR image. This helps the radiologist for the precise diagnosis of disease and for more effective interventional treatment procedures. This paper aims at designing an effective CT and MR image fusion method. In the proposed method, first source images are decomposed by using nonsubsampled contourlet transform (NSCT) which is a shift-invariant, multiresolution and multidirection image decomposition transform. Maximum entropy of square of the coefficients with in a local window is used for low-frequency sub-band coefficient selection. Maximum weighted sum-modified Laplacian is used for high-frequency sub-bands coefficient selection. Finally fused image is obtained through inverse NSCT. CT and MR images of different cases have been used to test the proposed method and results are compared with those of the other conventional image fusion methods. Both visual analysis and quantitative evaluation of experimental results shows the superiority of proposed method as compared to other methods.
Image fusion; Nonsubsampled Contourlet Transform; X-ray computed tomography; Magnetic resonance imaging; Image visualization
The naming of imaging procedures is currently not standardized across institutions. As a result, it is a challenge to establish national registries, for instance, a national registry of dose to facilitate comparisons among different types of CT procedures. RSNA’s RadLex Playbook is an effort towards addressing this gap (by introducing a unique Playbook identifier called an RPID for each procedure), and the current research focuses on semi-automatically mapping institution-specific procedure descriptions to Playbook entries to assist with this standardization effort. We discuss an algorithm we have developed to facilitate the mapping process which first extracts RadLex codes from the procedure description and then uses the definition of an RPID to determine the most suitable RPID(s) for the extracted set of RadLex codes. We also developed a tool that has three modes of operations—a single procedure mapping mode that allows a user to map a single institution-specific procedure description to a Playbook entry, a bulk mode to process large number of descriptions, and an exploratory mode that assists a user to better understand how the selection of values for various Playbook attributes affects the resulting RPID. We validate our algorithms using 166 production CT procedure descriptions and discuss how the tool can be used by administrators to map institution-specific procedure descriptions to RPIDs.
Interoperable radiology study descriptions; Mapping study description to Playbook entries; Radiology procedures and orderables; RadLex Playbook; Standardized image acquisition
Workflow is a widely used term to describe the sequence of steps to accomplish a task. The use of workflow technology in medicine and medical imaging in particular is limited. In this article, we describe the application of a workflow engine to improve workflow in a radiology department. We implemented a DICOM-enabled workflow engine system in our department. We designed it in a way to allow for scalability, reliability, and flexibility. We implemented several workflows, including one that replaced an existing manual workflow and measured the number of examinations prepared in time without and with the workflow system. The system significantly increased the number of examinations prepared in time for clinical review compared to human effort. It also met the design goals defined at its outset. Workflow engines appear to have value as ways to efficiently assure that complex workflows are completed in a timely fashion.
Workflow management; System architecture; DICOM; Business process management
Structured reporting, created when a standardized template with organized subheadings is combined with relevant observations of a diagnostic study into a meaningful result, has the potential to raise both the quality and the predictability of the radiologist report, revolutionizing the workflow and its outcomes. These templates contain great value, as they carve a path based on best practice for the radiologist to follow, and thus should be shared, reviewed, and improved. Unfortunately, these templates are often not shareable today due to a lack of standards for describing and transporting templates. This paper outlines and discusses an appropriate and effective electronic method for transporting radiology report templates using of the style of representational state transfer (REST). Enabling a structured radiology report template library with REST enables just-in-time accessibility of templates, achieving efficiencies and effectiveness.
REST; Ontologies; Structured; Reporting
The quantitative, multiparametric assessment of brain lesions requires coregistering different parameters derived from MRI sequences. This will be followed by analysis of the voxel values of the ROI within the sequences and calculated parametric maps, and deriving multiparametric models to classify imaging data. There is a need for an intuitive, automated quantitative processing framework that is generalized and adaptable to different clinical and research questions. As such flexible frameworks have not been previously described, we proceeded to construct a quantitative post-processing framework with commonly available software components. Matlab was chosen as the programming/integration environment, and SPM was chosen as the coregistration component. Matlab routines were created to extract and concatenate the coregistration transforms, take the coregistered MRI sequences as inputs to the process, allow specification of the ROI, and store the voxel values to the database for statistical analysis. The functionality of the framework was validated using brain tumor MRI cases. The implementation of this quantitative post-processing framework enables intuitive creation of multiple parameters for each voxel, facilitating near real-time in-depth voxel-wise analysis. Our initial empirical evaluation of the framework is an increased usage of analysis requiring post-processing and increased number of simultaneous research activities by clinicians and researchers with non-technical backgrounds. We show that common software components can be utilized to implement an intuitive real-time quantitative post-processing framework, resulting in improved scalability and increased adoption of post-processing needed to answer important diagnostic questions.
Brain imaging; Computer-Aided Diagnoses (CAD); User interface; Algorithms; Biomedical Image Analysis; Brain Morphology; Digital Image Processing; Digital Imaging and Communications in Medicine (DICOM); Image analysis; MR imaging; Segmentation; Software design; Systems integration
We present a novel algorithm for the extraction of cavity features on images of human vessels. Fat deposits in the inner wall of such structure introduce artifacts, and regions in the images captured invalidating the usual assumption of an elliptical model which makes the process of extracting the central passage effectively more difficult. Our approach was designed to cope with these challenges and extract the required image features in a fully automated, accurate, and efficient way using two stages: the first allows to determine a bounding segmentation mask to prevent major leakages from pixels of the cavity area by using a circular region fill that operates as a paint brush followed by Principal Component Analysis with auto correction; the second allows to extract a precise cavity enclosure using a micro-dilation filter and an edge-walking scheme. The accuracy of the algorithm has been tested using 30 computed tomography angiography scans of the lower part of the body containing different degrees of inner wall distortion. The results were compared to manual annotations from a specialist resulting in sensitivity around 98 %, false positive rate around 8 %, and positive predictive value around 93 %. The average execution time was 24 and 18 ms on two types of commodity hardware over sections of 15 cm of length (approx. 1 ms per contour) which makes it more than suitable for use in interactive software applications. Reproducibility tests were also carried out with synthetic images showing no variation for the computed diameters against the theoretical measure.
Image segmentation; Image brushing; Principal component analysis; Micro-dilation; Edge-walking; Local and nonlocal method; Dual-pass
In this study, the performance of a recently proposed computer-aided diagnosis (CAD) scheme in detection and 3D quantification of reticular and ground glass pattern extent in chest computed tomography of interstitial lung disease (ILD) patients is evaluated. CAD scheme performance was evaluated on a dataset of 37 volumetric chest scans, considering five representative axial anatomical levels per scan. CAD scheme reliability analysis was performed by estimating agreement (intraclass correlation coefficient, ICC) of automatically derived ILD pattern extent to semi-quantitative disease extent assessment in terms of 29-point rating scale provided by two expert radiologists. Receiver operating characteristic (ROC) analysis was employed to assess CAD scheme accuracy in ILD pattern detection in terms of area under ROC curve (Az). Correlation of reticular and ground glass volumetric pattern extent to pulmonary function tests (PFTs) was also investigated. CAD scheme reliability was substantial for ILD extent (ICC = 0.809) and distinct reticular pattern extent (0.806) and moderate for distinct ground glass pattern extent (0.543), performing within inter-observer agreement. CAD scheme demonstrated high accuracy in detecting total ILD (Az = 0.950 ± 0.018), while accuracy in detecting distinct reticular and ground glass patterns was 0.920 ± 0.023 and 0.883 ± 0.024, respectively. Moderate and statistically significant negative correlation was found between reticular volumetric pattern extent and diffusing capacity, forced expiratory volume in 1 s, forced vital capacity, and total lung capacity (R = −0.581, −0.513, −0.494, and −0.446, respectively), similar to correlations found between radiologists’ semi-quantitative ratings with PFTs. CAD-based quantification of disease extent is in agreement with radiologists’ semi-quantitative assessment and correlates to specific PFTs, suggesting a potential imaging biomarker for ILD staging and management.
Interstitial lung disease; Disease extent assessment; Automated 3D disease extent quantification; Semi-quantitative scoring; Pulmonary function tests