Comparison of central and local serial CT assessments of metastatic renal cell carcinoma patients in a clinical phase IIB study.

Background Clinical oncological studies attempt to improve precision of data by central radiological assessments. However, it is unclear, to which extent local and central assessments diverge. Purpose To quantify inter-reader variability and the deviation of local from central radiological assessments of computed tomography (CT) scans. Material and Methods This was a sub-study of a randomized clinical phase IIb trial in metastatic renal cell carcinoma (RCC), comparing first-line sorafenib with interferon-alpha-2a (IFN-α-2a). It analyzed agreements of local with central RECIST CT assessments by Cohen's kappa (κ), symmetry tests, deviations in waterfall plots, Bland-Altman plots, and parametric survival analyses. Results The concordance between local and central radiologic review was quantified by κ = 0.53. While local assessment yielded progressive disease (PD) in 18.6%, central assessment classified 22.5% of patient time points as PD exhibiting only a partial overlap with the 18.6% The tumor shrinkage rates in waterfall plots were 68.1% in local and 55.8% in central review (57.8% and 59% by Reader 1 and Reader 2). Bland-Altman plots identified a systematic shift of tumor change rates by -7.5% in local compared to central assessments, that may reflect a systematic tendency of more favorable results in local assessments. The discordance between local and central review was reflected by a time to progression (TTP) hazard ratio (HR) of 1.73 ( P = 0.0003). Conclusion These data suggest that central radiologic review may reduce technical measurement variability in clinical trials, which should be scrutinized in future studies compared to a volumetric reference.

A Statistical Model to Assess Risk for Supporting COVID-19 Quarantine Decisions.

In Germany, local health departments are responsible for surveillance of the current pandemic situation. One of their major tasks is to monitor infected persons. For instance, the direct contacts of infectious persons at group meetings have to be traced and potentially quarantined. Such quarantine requirements may be revoked, when all contact persons obtain a negative polymerase chain reaction (PCR) test result. However, contact tracing and testing is time-consuming, costly and not always feasible. In this work, we present a statistical model for the probability that no transmission of COVID-19 occurred given an arbitrary number of negative test results among contact persons. Hereby, the time-dependent sensitivity and specificity of the PCR test are taken into account. We employ a parametric Bayesian model which combines an adaptable Beta-Binomial prior and two likelihood components in a novel fashion. This is illustrated for group events in German school classes. The first evaluation on a real-world dataset showed that our approach can support important quarantine decisions with the goal to achieve a better balance between necessary containment of the pandemic and preservation of social and economic life. Future work will focus on further refinement and evaluation of quarantine decisions based on our statistical model.

Variability of semiautomated lung nodule volumetry on ultralow-dose CT: comparison with nodule volumetry on standard-dose CT.

The study investigates the effect of a substantial dose reduction on the variability of lung nodule volume measurements by assessing and comparing nodule volumes using a dedicated semiautomated segmentation software on ultralow-dose computed tomography (ULD-CT) and standard-dose computed tomography (SD-CT) data. In 20 patients, thin-slice chest CT datasets (1 mm slice thickness; 20% reconstruction overlap) were acquired at ultralow-dose (120 kV, 5 mAs) and at standard-dose (120 kV, 75 mAs), respectively, and analyzed using the segmentation software OncoTREAT (MeVis, Bremen, Germany; version 1.3). Interobserver variability of volume measurements of 202 solid pulmonary nodules (mean diameter 11 mm, range 3.2-44.5 mm) was calculated for SD-CT and ULD-CT. With respect to interobserver variability, the 95% confidence interval for the relative differences in nodule volume in the intrascan analysis was measured with -9.7% to 8.3% (mean difference -0.7%) for SD-CT and with -12.6% to 12.4% (mean difference -0.2%) for ULD-CT. In the interscan analysis, the 95% confidence intervals for the differences in nodule volume ranged with -25.1% to -23.4% and 26.2% to 28.9% (mean difference 1.4% to 2.1%) dependent on the combination of readers and scans. Intrascan interobserver variability of volume measurements was comparable for ULD-CT and SD-CT data. The calculated variability of volume measurements in the interscan analysis was similar to the data reported in the literature for CT data acquired with equal radiation dose. Thus, the evaluated segmentation software provides nodule volumetry that appears to be independent of the dose level with which the CT source dataset is acquired.

Evaluation of deep learning methods for parotid gland segmentation from CT images.

The segmentation of organs at risk is a crucial and time-consuming step in radiotherapy planning. Good automatic methods can significantly reduce the time clinicians have to spend on this task. Due to its variability in shape and low contrast to surrounding structures, segmenting the parotid gland is challenging. Motivated by the recent success of deep learning, we study the use of two-dimensional (2-D), 2-D ensemble, and three-dimensional (3-D) U-Nets for segmentation. The mean Dice similarity to ground truth is ∼ 0.83 for all three models. A patch-based approach for class balancing seems promising for false-positive reduction. The 2-D ensemble and 3-D U-Net are applied to the test data of the 2015 MICCAI challenge on head and neck autosegmentation. Both deep learning methods generalize well onto independent data (Dice 0.865 and 0.88) and are superior to a selection of model- and atlas-based methods with respect to the Dice coefficient. Since appropriate reference annotations are essential for training but often difficult and expensive to obtain, it is important to know how many samples are needed for training. We evaluate the performance after training with different-sized training sets and observe no significant increase in the Dice coefficient for more than 250 training cases.

Localization and quantification of regional and segmental air trapping in asthma.

OBJECTIVE: To assess the feasibility of volumetric and densitometric software to localize and quantify signs of regional air trapping after methacholine bronchoprovocations in asthma. METHODS: Eight atopic subjects with mild-to-moderate asthma using short-acting beta2-agonists only, with hyperresponsiveness to methacholine, were evaluated. Low-dose baseline expiratory 16-slice multidetector computed tomography scans before and after a methacholine bronchoprovocation were acquired. MeVisPULMO3D software (Bremen, Germany) was applied to the scans, providing quantitative information on volume and density measures of the total lung and each lobe. RESULTS: After methacholine, the expiratory scan showed a median (interquartile range) increase in volume of 534 mL (357-1279 mL), a decrease in lung density (mean and 15th percentile) of 52 Hounsfield Units (HU) (116-39 HU) and 34 HU (78-25 HU), respectively, and an increase in percentage low attenuation areas of 3% (2%-6%) for the total lung, with similar patterns in individual lung lobes. The right and left lower lung lobes showed the largest increases in air trapping, 211 mL (117-363 mL) and 229 mL (155-315 mL), respectively, versus a volume increase of 70 mL (20-249 mL), 26 mL (-16-92 mL), and 91 mL (-28-241 mL) for the right upper, middle, and left upper lobes, respectively. Volume changes in the lower lobes were associated with baseline forced expiratory flow between 25% and 75% of forced vital capacity, whereas low attenuation areas changes in the lower lobes were not. CONCLUSIONS: This study suggests that multidetector computed tomography scans are able to localize and quantify regional air trapping in asthma after methacholine bronchoprovocations. Volumetric measurements of the lobes as compared to densitometric measurements are superior in detecting local air trapping in gravity-dependent areas of the lung.

Homogeneous application of imaging criteria in a multicenter trial supported by investigator training: A report from the WAKE-UP study.

BACKGROUND AND PURPOSE: WAKE-UP is a randomized, placebo-controlled trial of thrombolysis in stroke with unknown time of symptom onset using magnetic resonance imaging criteria to determine patients' eligibility. As it is a multicenter trial, homogeneous interpretation of criteria is an important contributor to the trial's success. We describe the investigator image training as well as results of the quality control done by the central image reading board (CIRB). METHODS: Investigators at local centers were given an imaging manual and passed a software-based image training prior to being allowed to judge images in the trial. Throughout the trial, the CIRB gave feedback to recruiting centers in cases of disagreement regarding a patient's randomization. We evaluated the investigators performance in the image training and analyzed results of this quality control from the first 1069 screened patients. Additionally, we obtained feedback from investigators regarding their experiences with the trial. RESULTS: Four-hundred-and-sixty physicians from eight European countries took part in the image training, of whom 436 (95%) successfully completed it. In the trial, agreement rates between the local investigators and members of the CIRB were high for the presence of an acute ischemic lesion (94%, κ = 0.87) as well as for the judgment of infarct extent (93%, κ = 0.87). Agreement for the criterion of DWI-FLAIR mismatch was 74%, κ = 0.60. The majority of investigators reported that the DWI-FLAIR mismatch was the hardest imaging criterion to evaluate. Ninety-one percent of investigators who responded to our survey stated that the image training specifically increased their confidence when assessing the DWI-FLAIR mismatch. CONCLUSIONS: Despite its multicenter design, the WAKE-UP study has demonstrated a high level of homogeneity amongst raters in interpreting the various imaging criteria for patient randomization, including the novel criterion of DWI-FLAIR mismatch. Systematic image training increased the confidence of investigators in applying imaging criteria.

A novel iterative method for lesion delineation and volumetric quantification with FDG PET.

OBJECTIVES: The determination of lesion boundaries on FDG PET is difficult due to the point-spread blurring and unknown uptake of activity within a lesion. Standard threshold-based methods for volumetric quantification on PET usually neglect any size dependence and are biased by dependence on the signal-to-background ratio (SBR). A novel, model-based method is hypothesized to provide threshold levels independent f the SBR and to allow accurate measurement of volumes down to the resolution of the PET scanner. METHODS: A background-subtracted relative-threshold level (RTL) method was derived, based on a convolution of the point-spread function and a sphere with diameter D. Validation of the RTL method was performed using PET imaging of a Jaszczak phantom with seven hollow spheres (D=10-60 mm). Activity concentrations for the background and spheres (signal) were varied to obtain SBRs of 1.5-10. An iterative procedure was introduced for volumetric quantification, as the optimal RTL depends on a priori knowledge of the volume. The feasibility of the RTL method was tested in two patients with liver metastases and compared to a standard method using a fixed percentage of the signal. RESULTS: Phantom data validated that the theoretically optimal RTL depends on the sphere size, but not on the SBR. Typically, RTL=40% (D=15-60 mm), and RTL>50% for small spheres (D<12 mm). The RTL method is better applicable to patient data than the standard method. CONCLUSIONS: Based on an iterative procedure, the RTL method has been shown to provide optimal threshold levels independent of the SBR and to be applicable in phantom and in patient studies. It is a promising tool for lesion delineation and volumetric quantification of PET lesions.

Morphological segmentation and partial volume analysis for volumetry of solid pulmonary lesions in thoracic CT scans.

Volumetric growth assessment of pulmonary lesions is crucial to both lung cancer screening and oncological therapy monitoring. While several methods for small pulmonary nodules have previously been presented, the segmentation of larger tumors that appear frequently in oncological patients and are more likely to be complexly interconnected with lung morphology has not yet received much attention. We present a fast, automated segmentation method that is based on morphological processing and is suitable for both small and large lesions. In addition, the proposed approach addresses clinical challenges to volume assessment such as variations in imaging protocol or inspiration state by introducing a method of segmentation-based partial volume analysis (SPVA) that follows on the segmentation procedure. Accuracy and reproducibility studies were performed to evaluate the new algorithms. In vivo interobserver and interscan studies on low-dose data from eight clinical metastasis patients revealed that clinically significant volume change can be detected reliably and with negligible computation time by the presented methods. In addition, phantom studies were conducted. Based on the segmentation performed with the proposed method, the performance of the SPVA volumetry method was compared with the conventional technique on a phantom that was scanned with different dosages and reconstructed with varying parameters. Both systematic and absolute errors were shown to be reduced substantially by the SPVA method. The method was especially successful in accounting for slice thickness and reconstruction kernel variations, where the median error was more than halved in comparison to the conventional approach.

Real-time illustration of vascular structures.

We present real-time vascular visualization methods, which extend on illustrative rendering techniques to particularly accentuate spatial depth and to improve the perceptive separation of important vascular properties such as branching level and supply area. The resulting visualization can and has already been used for direct projection on a patient's organ in the operation theater where the varying absorption and reflection characteristics of the surface limit the use of color. The important contributions of our work are a GPU-based hatching algorithm for complex tubular structures that emphasizes shape and depth as well as GPU-accelerated shadow-like depth indicators, which enable reliable comparisons of depth distances in a static monoscopic 3D visualization. In addition, we verify the expressiveness of our illustration methods in a large, quantitative study with 160 subjects.

Informatics in radiology (infoRAD): new tools for computer assistance in thoracic CT part 2. Therapy monitoring of pulmonary metastases.

Owing to the rapid development of scanner technology, thoracic computed tomography (CT) offers new possibilities but also faces enormous challenges with respect to the quality of computer-assisted diagnosis and therapy planning. In the framework of the Virtual Institute for Computer Assistance in Clinical Radiology cooperative research project, a software application was developed to assist the radiologist in the analysis of thoracic CT data for the purpose of evaluating the response to tumor therapy. The application provides follow-up support for monitoring of tumor therapy by means of volumetric quantification of tumors and temporal registration. In addition, anatomically adequate three-dimensional visualization techniques for convenient examination of large data sets are included. With close cooperation between computer scientists and radiologists, the application was tested and optimized to achieve a high degree of usability. Several clinical studies were carried out, the results of which indicated that the application improves therapy monitoring with respect to accuracy and time required.

Informatics in radiology (infoRAD): new tools for computer assistance in thoracic CT. Part 1. Functional analysis of lungs, lung lobes, and bronchopulmonary segments.

Owing to the rapid development of scanner technology, thoracic computed tomography (CT) offers new possibilities but also faces enormous challenges with respect to the quality of computer-assisted diagnosis and therapy planning. In the framework of the Virtual Institute for Computer Assistance in Clinical Radiology cooperative research project, a prototypical software application was developed to assist the radiologist in functional analysis of thoracic CT data. By identifying the anatomic compartments of the lungs, the software application enables assessment of established functional CT parameters for each individual lung, pulmonary lobe, and pulmonary segment. Such region-based assessment allows a more localized diagnosis of lung diseases such as emphysema and more accurate estimation of regional lung function from CT data. With close cooperation between computer scientists and radiologists, the software application was tested and optimized to achieve a high degree of usability. Several clinical studies were carried out, the results of which indicated that the software application improves quantification in diagnosis, therapy planning, and therapy monitoring with respect to accuracy and time required.

Comparison of volumetric and linear serial CT assessments of lung metastases in renal cell carcinoma patients in a clinical phase IIB study.

RATIONALE AND OBJECTIVES: Accuracy of radiologic assessment may have a crucial impact on clinical studies and therapeutic decisions. We compared the variability of a central radiologic assessment (RECIST) and computer-aided volume-based assessment of lung lesions in patients with metastatic renal cell carcinoma (RCC). MATERIALS AND METHODS: The investigation was prospectively planned as a substudy of a clinical randomized phase IIB therapeutic trial in patients with RCC. Starting with the manual study diameter (SDM) of the central readers using RECIST in the clinical study, we performed computer-aided volume measurements. We compared SDM to an automated RECIST diameter (aRDM) and the diameter of a volume-equivalent sphere (effective diameter [EDM]), both for the individual size measurements and for the change rate (CR) between consecutive time points. One hundred thirty diameter pairs of 30 lung lesions from 14 patients were evaluable, forming 55 change pairs over two consecutive time points each. RESULTS: The SDMs of two different readers showed a correlation of 95.6%, whereas the EDMs exhibited an excellent correlation of 99.4%. Evaluation of CRs showed an SDM-CR correlation of 63.9%, which is substantially weaker than the EDM-CR correlation of 87.6%. The variability of SDM-CR is characterized by a median absolute difference of 11.4% points versus the significantly lower 1.8% points EDM-CRs variability (aRDM: 3.2% points). The limits of agreement between readers suggest that an EDM change of 10% or 1 mm can already be significant. CONCLUSIONS: Computer-aided volume-based assessments result in markedly reduced variability of parameters describing size and change, which may offer an advantage of earlier response evaluations and treatment decisions for patients.

Comparative study with new accuracy metrics for target volume contouring in PET image guided radiation therapy.

The impact of PET on radiation therapy is held back by poor methods of defining functional volumes of interest. Many new software tools are being proposed for contouring target volumes but the different approaches are not adequately compared and their accuracy is poorly evaluated due to the illdefinition of ground truth. This paper compares the largest cohort to date of established, emerging and proposed PET contouring methods, in terms of accuracy and variability. We emphasise spatial accuracy and present a new metric that addresses the lack of unique ground truth. 30 methods are used at 13 different institutions to contour functional VOIs in clinical PET/CT and a custom-built PET phantom representing typical problems in image guided radiotherapy. Contouring methods are grouped according to algorithmic type, level of interactivity and how they exploit structural information in hybrid images. Experiments reveal benefits of high levels of user interaction, as well as simultaneous visualisation of CT images and PET gradients to guide interactive procedures. Method-wise evaluation identifies the danger of over-automation and the value of prior knowledge built into an algorithm.

Rapid image recognition of body parts scanned in computed tomography datasets.

AIM: Automatic CT dataset classification is important to efficiently create reliable database annotations, especially when large collections of scans must be analyzed. METHOD: An automated segmentation and labeling algorithm was developed based on a fast patient segmentation and extraction of statistical density class features from the CT data. The method also delivers classifications of image noise level and patient size. The approach is based on image information only and uses an approximate patient contour detection and statistical features of the density distribution. These are obtained from a slice-wise analysis of the areas filled by various materials related to certain density classes and the spatial spread of each class. The resulting families of curves are subsequently classified using rules derived from knowledge about features of the human anatomy. RESULTS: The method was successfully applied to more than 5,000 CT datasets. Evaluation was performed via expert visual inspection of screenshots showing classification results and detected characteristic positions along the main body axis. Accuracy per body region was very satisfactory in the trunk (lung/liver >99.5% detection rate, presence of abdomen >97% or pelvis >95.8%) improvements are required for zoomed scans. CONCLUSION: The method performed very reliably. A test on 1,860 CT datasets collected from an oncological trial showed that the method is feasible, efficient, and is promising as an automated tool for image post-processing.