Oliver Weinheimer

Extraction of Airways From CT (EXACT'09)

This paper describes a framework for establishing a reference airway tree segmentation, which was used to quantitatively evaluate fifteen different airway tree extraction algorithms in a standardized manner. Because of the sheer difficulty involved in manually constructing a complete reference standard from scratch, we propose to construct the reference using results from all algorithms that are to be evaluated. We start by subdividing each segmented airway tree into its individual branch segments. Each branch segment is then visually scored by trained observers to determine whether or not it is a correctly segmented part of the airway tree. Finally, the reference airway trees are constructed by taking the union of all correctly extracted branch segments. Fifteen airway tree extraction algorithms from different research groups are evaluated on a diverse set of twenty chest computed tomography (CT) scans of subjects ranging from healthy volunteers to patients with severe pathologies, scanned at different sites, with different CT scanner brands, models, and scanning protocols. Three performance measures covering different aspects of segmentation quality were computed for all participating algorithms. Results from the evaluation showed that no single algorithm could extract more than an average of 74% of the total length of all branches in the reference standard, indicating substantial differences between the algorithms. A fusion scheme that obtained superior results is presented, demonstrating that there is complementary information provided by the different algorithms and there is still room for further improvements in airway segmentation algorithms.

Fully automatic quantitative assessment of emphysema in computed tomography: comparison with pulmonary function testing and normal values

Characterisation and quantification of emphysema are necessary for planning of local treatment and monitoring. Sensitive, easy to measure, and stable parameters have to be established and their relation to the well-known pulmonary function testing (PFT) has to be investigated. A retrospective analysis of 221 nonenhanced thin-section MDCT with a corresponding PFT was carried out, with a subgroup analysis in 102 COPD stage III+IV, 44 COPD stage 0, and 33 investigations into interstitial lung disease (ILD). The in-house YACTA software was used for automatic quantification of lung and emphysema volume [l], emphysema index, mean lung density (MLD [HU]) and 15th percentile [HU]. CT-derived lung volume is significantly smaller in ILD (3.8) and larger in COPD (7.2) than in controls (5.9, p < 0.0001). Emphysema volume and index are significantly higher in COPD than in controls (3.2 vs. 0.5, p < 0.0001, 45% vs. 8%, p < 0.0001). MLD and 15th percentile are significantly smaller in COPD (−877/−985, p < 0.0001) and significantly higher in ILD (−777, p < 0.0006/−914, p < 0.0001) than in controls (−829/−935). A relevant amount of COPD patients apparently do not suffer from emphysema, while controls who do not fulfil PFT criteria for COPD also demonstrate CT features of emphysema. Automatic quantification of thin-section CT delivers convincing parameters and ranges that are able to differentiate among emphysema, control and ILD. An emphysema index of lower 20%, MLD higher than −850, and 15th percentile lower than −950 might be regarded as normal (thin-section, nonenhanced, B40, YACTA). These ranges might be helpful in the judgement of individual measures.

Multi-detector CT of the chest: influence of dose onto quantitative evaluation of severe emphysema: a simulation study.

Purpose: Quantitative evaluation of the lung parenchyma might be impaired or unreliable by use of reduced-dose CT protocols. Aim of the study was to define the threshold where reduced dose has significant impact on quantitative emphysema parameters. Materials and Methods: Thirty patients with severe centrilobular emphysema underwent multidetector computed tomography (120 kV, 150 mAs). Original CT raw data were simulated using 10 mAs settings (10-100 SIMmAs). Quantitative analysis provided lung volume, emphysema volume, emphysema index, mean lung density, and 4 emphysema volume classes. Simulated low-dose results were compared with original acquisition. Results: Emphysema index showed no clinical relevant variation down to 30 SIMmAs. The large emphysema volume class was significantly different below 50 SIMmAs. The intermediate and small classes showed an overproportional variation below 50 SIMmAs. Conclusions: Dose reduction down to 30 SIMmAs is possible for clinical routine. Settings below 50 SIMmAs significantly alter the in-detailed 3-dimensional emphysema quantification.

About Objective 3-D Analysis of Airway Geometry in Computerized Tomography

The technology of multislice X-ray computed tomography (MSCT) provides volume data sets with approximately isotropic resolution, which permits a noninvasive 3-D measurement and quantification of airway geometry. In different diseases, like emphysema, chronic obstructive pulmonary disease (COPD), or cystic fribrosis, changes in lung parenchyma are associated with an increase in airway wall thickness. In this paper, we describe an objective measuring method of the airway geometry in the 3-D space. The limited spatial resolution of clinical CT scanners in comparison to thin structures like airway walls causes difficulties in the measurement of the density and the thickness of these structures. Initially, these difficulties will be addressed and then a new method is introduced to circumvent the problems. Therefore the wall thickness is approximated by an integral based closed-form solution, based on the volume conservation property of convolution. We evaluated the method with a phantom containing 10 silicone tubes and proved the repeatability in datasets of eight pigs scanned twice. Furthermore, a comparison of CT datasets of 16 smokers and 15 nonsmokers was done. Further medical studies are ongoing.

Contrast enhanced CT-scans are not comparable to non-enhanced scans in emphysema quantification

Systemic, interventional and surgical treatments have gone new ways in treatment of emphysema. For longitudinal therapy monitoring and as end-points for clinical trials, quantification of the disease is necessary. Sensitive, easy to measure, as well as stable and reproducible parameters have to be characterized. One parameter that might affect emphysema quantification is IV contrast enhancement, which might also be indicated. Whether or not the contrast enhanced scan is also suited for emphysema quantification or an additional scan is necessary, a retrospective analysis of 12 adult patients undergoing clinically indicated both, a non-enhanced and enhanced thin section MSCT within a week (median 0 days, range 0-4 days) was done. The in-house YACTA software was used for automatic quantification of lung and emphysema volume, emphysema index, mean lung density, and 5th, 10th, 15th percentile. After IV contrast administration, the median CT derived lung volume decreased mild by 1.1%, while median emphysema volume decreased by relevant 11%. This results in a decrease of median emphysema index by 9%. The median lung density (15th percentile) increased after contrast application by 18 HU (9 HU). CT quantification delivers emphysema values that are clearly affected by IV contrast application. The detected changes after contrast application show the results of higher density in the lung parenchyma. Therefore the amount of quantified emphysema is reduced and the lung density increased after contrast enhancement. In longitudinal analyses, non-enhanced scans should be the reference, while enhanced scans cannot be used.

Automatic airway analysis on multidetector computed tomography in cystic fibrosis: correlation with pulmonary function testing.

Purpose: To evaluate the fully automatic quantification of airway dimensions on chest multidetector computed tomography (MDCT) performed in cystic fibrosis (CF) patients. Airflow indices including predicted forced expiratory volume in 1 second (FEV1%) were used to study the impact on regional lung function. Materials and Methods: MDCT data of patients with CF (14 children and 23 adults) and of control patients (11 children and 22 adults) were used to compute total diameter (TD), lumen area (LA), and wall thickness (WT) using dedicated software. Pulmonary function testing including FEV1% was performed in parallel and correlated with MDCT parameters in a generation-based analysis. Results: TD was largely increased in CF patients (third-generation to fourth-generation airways in children, first to ninth in adults; P<0.05). LA remained unchanged, but WT was also larger in CF compared with controls (third generation to sixth generation in children, first to eleventh in adults; P<0.05). In adult CF patients significant negative correlations for TD, LA, and WT with FEV1% were found for intermediate airways (fifth to seventh generation; r=−0.7 to −0.9) but not in pediatric CF patients and controls. Conclusions: Automatic airway analysis succeeded in quantifying specific pathologies such as airway dilatation and wall thickening in CF patients at different ages. Moreover, our results indicate a shift in main airflow resistance to intermediate airways in cases of chronic CF. The objective computational parameters TD, LA, and WT should be considered for assessment and follow-up of CF airway disease.

Pulmonary emphysema in cystic fibrosis detected by densitometry on chest multidetector computed tomography.

Background Histopathological studies on lung specimens from patients with cystic fibrosis (CF) and recent results from a mouse model indicate that emphysema may contribute to CF lung disease. However, little is known about the relevance of emphysema in patients with CF. In the present study, we used computationally generated density masks based on multidetector computed tomography (MDCT) of the chest for non-invasive characterization and quantification of emphysema in CF. Methods Volumetric MDCT scans were acquired in parallel to pulmonary function testing in 41 patients with CF (median age 20.1 years; range 7-66 years) and 21 non-CF controls (median age 30.4 years; range 4-68 years), and subjected to dedicated software. The lung was segmented, low attenuation volumes below a threshold of -950 Hounsfield units were assigned to emphysema volume (EV), and the emphysema index was computed (EI). Results were correlated with forced expiratory volume in 1 s percent predicted (FEV1%), residual volume (RV), and RV/total lung capacity (RV/TLC). Results We show that EV was increased in CF (457±530 ml) compared to non-CF controls (78±90 ml) (P<0.01). EI was also increased in CF (7.7±7.5%) compared to the control group (1.2±1.4%) (P<0.05). EI correlated inversely with FEV1% (rs=-0.66), and directly with RV (rs=0.69) and RV/TLC (rs=0.47) in patients with CF (P<0.007), but not in non-CF controls. Emphysema in CF was detected from early adolescence (~13 years) and increased with age (rs=0.67, P<0.001). Conclusions Our results indicate that early onset emphysema detected by densitometry on chest MDCT is a characteristic pathology that contributes to airflow limitation and may serve as a novel endpoint for monitoring lung disease in CF.

Quantification of lung volume at different tidal volumes and positive end-expiratory pressures in a porcine model by using retrospective respiratory gated 4D-computed tomography.

Purpose:This feasibility study in healthy animals should prove the concept that it is possible to quantitatively assess the effects of different ventilatory settings on the lung parenchyma during ongoing ventilation in respiratory gated 4-dimensional (D)-computed tomography (CT). For this purpose, the influence of different tidal volumes and positive end-expiratory pressure (PEEP) on quantitative assessment of lung volumes (LVs) and lung compartments was analyzed. Methods:Five anesthetized and ventilated (20 breaths/min, inspiratory/expiratory ratio of 1:2) healthy pigs underwent 16-row multidetector CT with retrospective respiratory gating using a noncontact charge-coupled device camera as a gating device. The device was connected to the scanner instead of the electrocardiogram gating unit. Parameters for retrospective scans were collimation 1 mm, 120 kV, 300 mA, gantry rotation time 0.5 seconds, helical pitch 2.4. Two tidal volumes (300 mL and 450 mL) and 3 PEEP levels (0, 5, and 10 cm H2O) were applied resulting in 6 scans per animal. Images were reconstructed throughout the respiratory cycle in increments of 10%. Semiautomatic segmentation provided LV, mean lung density (MLD), and different lung compartments (atelectasis, hypoventilated, normal ventilated, hyperventilated). Results:At tidal volume 300 mL the inspiratory LV were 1.05, 1.26, and 1.5 L and expiratory LV 0.75, 0.99, and 1.24 L (PEEP 0, 5, 10 cm H2O, respectively). Differences of MLD between inspiration and expiration were 86, 65, and 46 HU. At tidal volume 450 mL, the inspiratory LV were 1.21, 1.43, and 1.72 L, and expiratory LV were 0.78, 1.01, and 1.34 L (for PEEP 0, 5, 10 cm H2O). Differences of MLD between inspiration and expiration were 109, 86, and 59 HU. A clear oscillatory wave of the normal and hypoventilated volumes was found at PEEP 0, with increase in PEEP the hypoventilated areas became increasingly normal ventilated, the amplitude of the curves decreased, and hyperventilated areas increased. Conclusion:Using a new 4D-CT technique we were able to demonstrate the effect of different ventilation settings on the whole lung during the whole respiratory cycle. The disadvantages of static lung imaging or dynamic 2D-CT can be overcome. The possibility of quantitative evaluation of the whole lung and direct visualization and measurement of recruitment during different ventilation settings might be a great benefit for patients suffering from inhomogeneous lung injury and failure.

Visual vs Fully Automatic Histogram-Based Assessment of Idiopathic Pulmonary Fibrosis (IPF) Progression Using Sequential Multidetector Computed Tomography (MDCT).

Objectives To describe changes over time in extent of idiopathic pulmonary fibrosis (IPF) at multidetector computed tomography (MDCT) assessed by semi-quantitative visual scores (VSs) and fully automatic histogram-based quantitative evaluation and to test the relationship between these two methods of quantification. Methods Forty IPF patients (median age: 70 y, interquartile: 62-75 years; M:F, 33:7) that underwent 2 MDCT at different time points with a median interval of 13 months (interquartile: 10-17 months) were retrospectively evaluated. In-house software YACTA quantified automatically lung density histogram (10th-90th percentile in 5th percentile steps). Longitudinal changes in VSs and in the percentiles of attenuation histogram were obtained in 20 untreated patients and 20 patients treated with pirfenidone. Pearson correlation analysis was used to test the relationship between VSs and selected percentiles. Results In follow-up MDCT, visual overall extent of parenchymal abnormalities (OE) increased in median by 5 %/year (interquartile: 0 %/y; +11 %/y). Substantial difference was found between treated and untreated patients in HU changes of the 40th and of the 80th percentiles of density histogram. Correlation analysis between VSs and selected percentiles showed higher correlation between the changes (Δ) in OE and Δ 40th percentile (r=0.69; p<0.001) as compared to Δ 80th percentile (r=0.58; p<0.001); closer correlation was found between Δ ground-glass extent and Δ 40th percentile (r=0.66, p<0.001) as compared to Δ 80th percentile (r=0.47, p=0.002), while the Δ reticulations correlated better with the Δ 80th percentile (r=0.56, p<0.001) in comparison to Δ 40th percentile (r=0.43, p=0.003). Conclusions There is a relevant and fully automatically measurable difference at MDCT in VSs and in histogram analysis at one year follow-up of IPF patients, whether treated or untreated: Δ 40th percentile might reflect the change in overall extent of lung abnormalities, notably of ground-glass pattern; furthermore Δ 80th percentile might reveal the course of reticular opacities.

Influence of pixel size on quantification of airway wall thickness in computed tomography.

Objectives: The purpose of this study was to determine the point where a further decrease in voxel size does not result in better automatic quantification of the bronchial wall thickness by using 2 different assessment techniques. Materials and Methods: The results from the commonly used full-width-at-half-maximum (FWHM) principle and a new technique (integral-based method [IBM]) were compared for thin-section multidetector computed tomography (MDCT) data sets from an airway phantom containing 10 different tubular airway phantoms and in a human subsegmental bronchus in vivo. Correlation with the actual wall thickness and comparison of the wall thicknesses assessed for different voxel sizes were performed, and the image resolutions were also compared subjectively. Results: The relative error ranged from 0% (biggest phantom) to 330% (smallest phantom, biggest field of view, smaller matrix, and FWHM). Using IBM, the maximum relative error was 10% in the same setting. For FWHM, the improvement was marginal for most settings with a pixel spacing less than 0.195 × 0.195 × 0.8 mm; however, it still decreases the relative error from 290% to 273.6% for a wall thickness of 0.3 mm and a pixel spacing of 0.076 × 0.076 × 0.8 mm. Conclusions: (1) Using a special technique such as IBM to account for computed tomographys blurring effect in assessing airway wall thickness had the greatest impact on correct quantification. (2) The visual impression and the automatic quantification using the FWHM technique improved marginally by decreasing the voxel size to less than 0.195 × 0.195 × 0.8 mm. (3) The FWHM technique as a model for visual quantification is not reliable for airway wall thicknesses less than 1.5 mm.