SLEEP-SEE-THROUGH: Explainable Deep Learning for Sleep Event Detection and Quantification From Wearable Somnography.

Evidence is rapidly accumulating that multifactorial nocturnal monitoring, through the coupling of wearable devices and deep learning, may be disruptive for early diagnosis and assessment of sleep disorders. In this work, optical, differential air-pressure and acceleration signals, acquired by a chest-worn sensor, are elaborated into five somnographic-like signals, which are then used to feed a deep network. This addresses a three-fold classification problem to predict the overall signal quality (normal, corrupted), three breathing-related patterns (normal, apnea, irregular) and three sleep-related patterns (normal, snoring, noise). In order to promote explainability, the developed architecture generates additional information in the form of qualitative (saliency maps) and quantitative (confidence indices) data, which helps to improve the interpretation of the predictions. Twenty healthy subjects enrolled in this study were monitored overnight for approximately ten hours during sleep. Somnographic-like signals were manually labeled according to the three class sets to build the training dataset. Both record- and subject-wise analyses were performed to evaluate the prediction performance and the coherence of the results. The network was accurate (0.96) in distinguishing normal from corrupted signals. Breathing patterns were predicted with higher accuracy (0.93) than sleep patterns (0.76). The prediction of irregular breathing was less accurate (0.88) than that of apnea (0.97). In the sleep pattern set, the distinction between snoring (0.73) and noise events (0.61) was less effective. The confidence index associated with the prediction allowed us to elucidate ambiguous predictions better. The saliency map analysis provided useful insights to relate predictions to the input signal content. While preliminary, this work supported the recent perspective on the use of deep learning to detect particular sleep events in multiple somnographic signals, thus representing a step towards bringing the use of AI-based tools for sleep disorder detection incrementally closer to clinical translation.

Ex Vivo Evaluation of a Pressure-Sensitive Device to Aid Big Bubble Intrastromal Dissection in Deep Anterior Lamellar Keratoplasty.

Purpose: To develop and perform ex vivo testing for a device designed for semiquantitative determination of intracorneal dissection depth during big bubble (BB) deep anterior lamellar keratoplasty. Methods: A prototype device connected to a syringe and cannula was designed to determine depth of intrastromal placement based on air rebound pressure emitted by a software controlled generator. Ex vivo testing of the device was conducted on human corneas mounted on an artificial anterior chamber in three experiments: (1) cannula purposely introduced at different depths measured with anterior segment optical coherence tomography, (2) cannula introduced as per the BB technique, and (3) simulation of the BB technique guided by the device. Results: A positive pressure differential and successful BB were observed only when the cannula was positioned within 150 microns from the endothelial plane. In all successful BB cases (21/40), a repeatable increase in tissue rebound pressure was detected, which was not recorded in unsuccessful cases. The device was able to signal to the surgeon correct placement of the cannula (successful BB) in 16 of 17 cases and incorrect placement of the cannula (unsuccessful BB) in 8 of 8 cases (94.1% sensitivity, 100% specificity). Conclusions: In our ex vivo model, this novel medical device could reliably signal cannula positioning in the deep stroma for effective pneumatic dissection and possibly aid technical execution of BB deep anterior lamellar keratoplasty. Translational Relevance: A medical device that standardizes big bubble deep anterior lamellar keratoplasty could increase the overall success rate of the surgical procedure and aid popularization of deep anterior lamellar keratoplasty.

Identification of Characteristic Points in Multivariate Physiological Signals by Sensor Fusion and Multi-Task Deep Networks.

Identification of characteristic points in physiological signals, such as the peak of the R wave in the electrocardiogram and the peak of the systolic wave of the photopletismogram, is a fundamental step for the quantification of clinical parameters, such as the pulse transit time. In this work, we presented a novel neural architecture, called eMTUnet, to automate point identification in multivariate signals acquired with a chest-worn device. The eMTUnet consists of a single deep network capable of performing three tasks simultaneously: (i) localization in time of characteristic points (labeling task), (ii) evaluation of the quality of signals (classification task); (iii) estimation of the reliability of classification (reliability task). Preliminary results in overnight monitoring showcased the ability to detect characteristic points in the four signals with a recall index of about 1.00, 0.90, 0.90, and 0.80, respectively. The accuracy of the signal quality classification was about 0.90, on average over four different classes. The average confidence of the correctly classified signals, against the misclassifications, was 0.93 vs. 0.52, proving the worthiness of the confidence index, which may better qualify the point identification. From the achieved outcomes, we point out that high-quality segmentation and classification are both ensured, which brings the use of a multi-modal framework, composed of wearable sensors and artificial intelligence, incrementally closer to clinical translation.

Hybrid Convolutional Networks for End-to-End Event Detection in Concurrent PPG and PCG Signals Affected by Motion Artifacts.

The accurate detection of physiologically-related events in photopletismographic (PPG) and phonocardiographic (PCG) signals, recorded by wearable sensors, is mandatory to perform the estimation of relevant cardiovascular parameters like the heart rate and the blood pressure. However, the measurement performed in uncontrolled conditions without clinical supervision leaves the detection quality particularly susceptible to noise and motion artifacts. This work proposes a new fully-automatic computational framework, based on convolutional networks, to identify and localize fiducial points in time as the foot, maximum slope and peak in PPG signal and the S1 sound in the PCG signal, both acquired by a custom chest sensor, described recently in the literature by our group. The event detection problem was reframed as a single hybrid regression-classification problem entailing a custom neural architecture to process sequentially the PPG and PCG signals. Tests were performed analysing four different acquisition conditions (rest, cycling, rest recovery and walking). Cross-validation results for the three PPG fiducial points showed identification accuracy greater than 93 % and localization error (RMSE) less than 10 ms. As expected, cycling and walking conditions provided worse results than rest and recovery, however reaching an accuracy greater than 90 % and a localization error less than 15 ms. Likewise, the identification and localization error for S1 sound were greater than 90 % and less than 25 ms. Overall, this study showcased the ability of the proposed technique to detect events with high accuracy not only for steady acquisitions but also during subject movements. We also showed that the proposed network outperformed traditional Shannon-energy-envelope method in the detection of S1 sound, reaching detection performance comparable to state of the art algorithms. Therefore, we argue that coupling chest sensors and deep learning processing techniques may disclose wearable devices to unobtrusively acquire health information, being less affected by noise and motion artifacts.

Quantitative Multivolume Proton-Magnetic Resonance Imaging in Lung Transplant Recipients: Comparison With Computed Tomography and Spirometry.

RATIONALE AND OBJECTIVES: Acute and chronic graft rejection remains the major problem in clinical surveillance of lung-transplanted patients and early detection of complications is of capital importance to allow the optimal therapeutic option. The aim of this study was to investigate the role of quantitative non contrast-enhanced magnetic resonance imaging (MRI) as a non-ionizing imaging modality to assess ventilation impairment in patients who have undergone lung transplantation, in comparison with quantitative computed tomography (CT) and spirometry. MATERIALS AND METHODS: Ten lung-transplanted patients (39 ±12 years, forced-expiratory volume in 1 second (FEV1) = 81 ± 27%, forced vital capacity (FVC) = 87 ± 27%) were acquired in breath-hold at full-expiration and full-inspiration with 1.5T MRI and CT. Maps of expiratory-inspiratory difference in MR signal-intensity and CT-density were computed to estimate regional ventilation. Based on expiratory, inspiratory, and expiratory-inspiratory difference values, each pixel was classified as healthy (H), low ventilation (LV), air trapping (AT), and consolidation (C) and the percent extent of each class was quantified. RESULTS: Overall, expiratory-inspiratory difference in MR signal-intensity correlated to CT-density (r = 0.64, p < 0.0001) and to FEV1 (ρ = 0.71, p = 0.02). The linear correlation between MRI and CT functional maps considering all the four classes is r = 0.93 (p < 0.0001). MRI percent volumes of H, AT, and C correlated to FEV1 %pred, with the highest correlation reported for AT (ρ = -0.82). CONCLUSION: Results demonstrated a good agreement between MRI and CT ventilation imaging and between the corresponding percent volumes of lung damage. Quantitative MRI may represent an accurate non-ionizing imaging technique for longitudinal monitoring of lung transplant recipients.

The effect of primary graft dysfunction after lung transplantation on parenchymal remodeling detected by quantitative computed tomography.

BACKGROUND: Regional analysis by computed tomography (CT) is an attractive technique to interpret lung patterns after transplantation (LTx). We evaluated the application of CT functional mask derived parameters to determine whether development of primary graft dysfunction (PGD) is associated with short and/or long-term postoperative evidences of pulmonary function alterations. METHODS: A total of 38 patients who underwent bilateral LTx were evaluated at 24, 48 and 72 hours after the end of surgery to establish PGD occurrence and grading. CT scans at 3 and 12 months after LTx were analyzed to measure specific gas volume (SVg) changes normalized on expiratory SVgEXP of the whole lung (ΔSVg/SVgEXP) and to obtain functional masks of density variation, namely maps of low ventilation (LV), consolidation (C), air trapping (AT) and healthy parenchyma (H). RESULTS: Our main result was the evidence of a marked decrease in ΔSVg/SVgEXP in all subjects, irrespectively on PGD, at each time point after LTx, indicating a high degree of ventilation defects versus healthy. High percentages of LV were found in all subjects while percentages of AT and C were negligible. CONCLUSIONS: We demonstrate that quantification of ventilation defects by CT functional mask offers insights into the correlation between PGD and pulmonary function after LTx at short and mid-term.

Quantitative multivolume proton-magnetic resonance imaging in patients with cystic fibrosis lung disease: comparison with clinical indicators.

This cross-sectional study aims to verify the relationship between quantitative multivolume proton-magnetic resonance imaging (1H-MRI) and clinical indicators of ventilatory abnormalities in cystic fibrosis (CF) lung disease.Non-enhanced chest MRI, spirometry and multiple breath washout was performed by 28 patients (10-27 years) with CF lung disease. Images acquired at end-inspiration and end-expiration were registered by optical flow to estimate expiratory-inspiratory proton-density change (Δ1H-MRI) as a measure of regional ventilation. Magnetic resonance images were also evaluated using a CF-specific scoring system.Biomarkers of CF ventilation impairment were defined from the Δ1H-MRI as follows: Δ1H-MRI median, Δ1H-MRI quartile coefficient of variation (QCV) and percentage of low-ventilation volume (%LVV). Imaging biomarkers correlated to all the clinical measures of ventilation abnormality, with the strongest correlation between Δ1H-MRI median and forced expiratory volume in 1 s (r2=0.44, p<0.001), Δ1H-MRI QCV and lung clearance index (LCI) (r2=0.51, p<0.001) and %LVV and LCI (r2=0.66, p<0.001). Correlations were also found between imaging biomarkers of ventilation and morphological scoring.The study showed a significant correlation between quantitative multivolume MRI and clinical indicators of CF lung disease. MRI, as a non-ionising imaging technique, may be particularly attractive in CF care for longitudinal evaluation, providing a new imaging biomarker to detect early ventilatory abnormalities.

Mechanisms of exercise limitation in patients with chronic hypersensitivity pneumonitis.

Small airway and interstitial pulmonary involvements are prominent in chronic hypersensitivity pneumonitis (cHP). However, their roles on exercise limitation and the relationship with functional lung tests have not been studied in detail. Our aim was to evaluate exercise performance and its determinants in cHP. We evaluated maximal cardiopulmonary exercise testing performance in 28 cHP patients (forced vital capacity 57±17% pred) and 18 healthy controls during cycling. Patients had reduced exercise performance with lower peak oxygen production (16.6 (12.3-19.98) mL·kg-1·min-1versus 25.1 (16.9-32.0), p=0.003), diminished breathing reserve (% maximal voluntary ventilation) (12 (6.4-34.8)% versus 41 (32.7-50.8)%, p<0.001) and hyperventilation (minute ventilation/carbon dioxide production slope 37±5 versus 31±4, p<0.001). All patients presented oxygen desaturation and augmented Borg dyspnoea scores (8 (5-10) versus 4 (1-7), p=0.004). The prevalence of dynamic hyperinflation was found in only 18% of patients. When comparing cHP patients with normal and low peak oxygen production (<84% pred, lower limit of normal), the latter exhibited a higher minute ventilation/carbon dioxide production slope (39±5.0 versus 34±3.6, p=0.004), lower tidal volume (0.84 (0.78-0.90) L versus 1.15 (0.97-1.67) L, p=0.002), and poorer physical functioning score on the Short form-36 health survey. Receiver operating characteristic curve analysis showed that reduced lung volumes (forced vital capacity %, total lung capacity % and diffusing capacity of the lung for carbon dioxide %) were high predictors of poor exercise capacity. Reduced exercise capacity was prevalent in patients because of ventilatory limitation and not due to dynamic hyperinflation. Reduced lung volumes were reliable predictors of lower performance during exercise.

The effect of exogenous surfactant on alveolar interdependence.

To investigate the nature of alveolar mechanical interdependence, we purposefully disturbed the equilibrium condition by administering exogenous surfactant in physiological non-surfactant deprived conditions. Changes in alveolar morphology induced by intra-tracheal delivery of CUROSURF were evaluated after opening a pleural window allowing in-vivo microscopic imaging of sub-pleural alveoli in 6 male anesthetized, tracheotomized and mechanically ventilated rabbits. Surfactant instillation increased the surface area of alveoli smaller than 20,000 µm(2) up to ∼ 50% at 15 min after instillation, reflecting a lowering of surface tension due to local surfactant enrichment. Conversely, for alveoli greater than 20,000 µm(2), surface area decreased by ∼ 5%. Opposite changes in alveolar surface are interpreted as reflecting a new inter-alveolar mechanical equilibrium modified by local surfactant distribution and by a decrease in lung distending pressure. We propose that smaller alveoli, representing the majority of alveolar population, might mostly contribute to improve the oxygenation index following surfactant replacement therapy in case of surfactant deficiency.

Heterogeneity of specific gas volume changes: a new tool to plan lung volume reduction in COPD.

OBJECTIVE: The aim of this work was to investigate if regional differences of specific gas volume (SVg) in the different regions (lobes and bronchopulmonary segments) in healthy volunteers and patients with severe emphysema can be used as a tool for planning lung volume reduction (LVR) in emphysema. METHODS: CT scans of 10 healthy subjects and 10 subjects with severe COPD were obtained at end-inspiration (total lung capacity [TLC]) and end-expiration (residual volume [RV]). For each subject, ΔSVg (ΔSVg = SVg,TLC - SVg,RV, where SVg,TLC and SVg,RV are specific gas volume at TLC and RV, respectively) vs ΔV (ΔV = V,TLC-V,RV, where V,TLC and V,RV are lung volume at TLC and RV, respectively) was plotted for the entire lung, each lobe, and all bronchopulmonary segments. For each subject, a heterogeneity index (HI) was defined to quantify the range of variability of ΔSVg/ΔV in all bronchopulmonary regions. RESULTS: In patients with COPD, SVg,TLC and SVg,RV were significantly higher and ΔSVg variations lower than in healthy subjects (P < .001). In COPD, ΔSVg/ΔV slopes were lower in upper lobes than in lower lobes. In healthy subjects, the entire lung, lobes, and bronchopulmonary segments all showed similar ΔSVg/ΔV slopes, whereas in COPD a high variance was found. As a consequence, HI was significantly higher in subjects with COPD than in healthy subjects (0.80 ± 0.34 vs 0.15 ± 0.10, respectively; P < .001). CONCLUSIONS: SVg variations within the lung are highly homogeneous in healthy subjects. Regions with low ΔSVg/ΔV (ie, more pronounced gas trapping) should be considered as target areas for LVR. Regions with negative values of ΔSVg/ΔV identify where collateral ventilation is present. HI is helpful to assess the patient in the different stages of disease and the effect of different LVR treatments.

Comparison between multivolume CT-based surrogates of regional ventilation in healthy subjects.

RATIONALE AND OBJECTIVES: The assessment of regional ventilation is of critical importance when investigating lung function during disease progression and planning of pulmonary interventions. Recently, different computed tomography (CT)-based parameters have been proposed as surrogates of lung ventilation. The aim of the present study was to compare these parameters, namely variations of density (ΔHU), specific volume (sVol), and specific gas volume (ΔSVg) between different lung volumes, in relation to their topographic distribution within the lung. MATERIALS AND METHODS: Ten healthy volunteers were scanned via high-resolution CT at residual volume (RV) and total lung capacity (TLC); ΔHU, sVol, and ΔSVg were mapped voxel by voxel after registering TLC onto RV. Variations of the three parameters along the vertical and horizontal directions were analyzed. RESULTS: Along the vertical direction (from ventral to dorsal regions), a strong dependence on gravity was found in ΔHU and sVol, with greater values in the dorsal regions of the lung (P < .001), whereas ΔSVg was more homogeneously distributed within the lung. Conversely, along the caudocranial direction (from lung bases to apexes) where no gravitational gradient is present, the three parameters behaved similarly, with lower values at the apices. CONCLUSIONS: ΔHU, sVol, and ΔSVg behave differently along the gravity direction. As the greater amount of air delivered to the dependent portion of the lung supplies a larger number of alveoli, the amount of gas delivered to alveoli compared to the mass of tissue is not gravity dependent. The minimization of gravity dependence in the distribution of ventilation when using ΔSVg suggests that this parameter is more reliable to discriminate healthy from pathologic regions.

Alveolar mechanics studied by in vivo microscopy imaging through intact pleural space.

In six male anesthetized, tracheotomized, and mechanically ventilated rabbits we derived indications on alveolar mechanics from in vivo imaging, using a "pleural window" technique (pleural space intact) that allows unrestrained movement of the same subpleural alveoli (N=60) on increasing alveolar pressure from 4 to 8 cmH2O. Absolute compliance (C(abs), ratio of change in alveolar surface area to the change in alveolar pressure) was significantly lower in smaller compared to larger alveoli. Specific compliance, C(sp), obtained by normalizing C(abs) to alveolar surface area, was essentially independent of alveolar size. Both C(abs) and C(sp) were affected by large variability likely reflecting the complex matching between elastic and surface forces. We hypothesize that the relative constancy of C(sp) might contribute to reduce interregional differences in parenchymal and surface forces in the lung tissue by contributing to assure a uniform stretching in a model of mechanically inter-dependent alveoli.

From morphological heterogeneity at alveolar level to the overall mechanical lung behavior: an in vivo microscopic imaging study.

In six male anesthetized, tracheotomized, and mechanically ventilated rabbits, we imaged subpleural alveoli under microscopic view (60×) through a "pleural window" obtained by stripping the endothoracic fascia and leaving the parietal pleura intact. Three different imaging scale levels were identified for the analysis on increasing stepwise local distending pressure (P ld) up to 16.5 cmH2O: alveoli, alveolar cluster, and whole image field. Alveolar profiles were manually traced, clusters of alveoli of similar size were identified through a contiguity-constrained hierarchical agglomerative clustering analysis and alveolar surface density (ASD) was estimated as the percentage of air on the whole image field. Alveolar area distributions were remarkably right-skewed and showed an increase in median value with a large topology-independent heterogeneity on increasing P ld. Modeling of alveolar area distributions on increasing P ld led to hypothesize that absolute alveolar compliance (change in surface area over change in P ld) increases fairly linearly with increasing initial alveolar size, the corollary of this assumption being a constant specific compliance. Clusters were reciprocally interweaved due to their highly variable complex shapes. ASD was found to increase with a small coefficient of variation (CV <25%) with increasing P ld. The CV of lung volume at each transpulmonary pressure was further decreased (about 6%). The results of the study suggest that the considerable heterogeneity of alveolar size and of the corresponding alveolar mechanical behavior are homogenously distributed, resulting in a substantially homogenous mechanical behavior of lung units and whole organ.

Experimental model to evaluate the effect of hydrothorax and lobar resection on lung compliance.

OBJECTIVES: The objective of this study was to evaluate to what extent lung compliance is affected by the individual and combined action of lung resection and hydrothorax in an animal model. METHODS: Anaesthetized and mechanically ventilated rabbits (weight range 2 ÷ 2.2 kg) were randomized in two groups: (i) experimental hydrothorax (from 2 to 8 ml) (n = 5) and (ii) right lower lobe lobectomy (n = 4) and right middle plus lower lobe resection (n = 2). To obtain lung compliance, we measured alveolar, oesophageal pressures and lung volume during slow inflation manoeuvres in control conditions and after hydrothorax or lung resection. Lung compliance was estimated as the change in lung volume divided by the change in transpulmonary pressure. Based on the changes in compliance of the whole lung, we calculated the corresponding changes in compliance of the right lung, which was directly exposed to unilateral hydrothorax and lobectomy. RESULTS: Average total lung compliance in the control was 3.3 ± 0.8 (SD) ml/cmH2O. Eight millilitres of hydrothorax significantly decreased (P < 0.001) lung compliance to 2.7 ± 0.7 ml/cmH2O and increased pleural liquid pressure at the bottom of the cavity from -1 cmH2O up to ∼ 2.5-3 cmH2O. Resection of the right lower lobe significantly decreased (P < 0.001) lung compliance to 1.75 ± 0.3 ml/cmH2O. Resection of the right middle plus lower lobes significantly decreased (P < 0.001) lung compliance to 1.52 ± 0.4 ml/cmH2O. CONCLUSIONS: Following hydrothorax, the decrease in right lung compliance (∼ 45%) was much greater than that expected based on the estimated decrease in right lung volume (20%). We attribute this difference to the fact that hydrothorax causes the lung to be exposed to positive, rather than sub-atmospheric, pressure, causing atelectasis. Following lobectomy, right lung compliance decreased by 62 and 80% for estimated decreases in lung volume of 30 and 60%. This difference could reflect inaccuracy in the estimate of lung volume reduction based on resected weight and/or surgical damage. We conclude that potential detrimental effects of hydrothorax and lobar resection decrease lung compliance and expose the lung to the risk of over-distension when a chest drain is applied.

Regional lung function and heterogeneity of specific gas volume in healthy and emphysematous subjects.

The aim of our study was to study regional lung function by standard computed tomography (CT) and characterise regional variations of density and specific gas volume (SVg) between different lung volumes. We studied 10 healthy and 10 severely emphysematous subjects. Corresponding CT images taken at high and low lung volumes were registered by optical flow to obtain two-dimensional maps of pixel-by-pixel differences of density (ΔHU) and SVg (ΔSVg) at slice levels near the aortic arch, carina and top diaphragm. In healthy subjects, ΔHU was higher at all levels (p<0.001) with higher variability expressed as interquartile range (p<0.001), largely due to its differences between dorsal and ventral regions. In patients, median ΔSVg values were 3.2 times lower than healthy volunteers (p<0.001), while heterogeneity of ΔSVg maps, expressed as quartile coefficient of variation, was 5.4 times higher (p<0.001). In all patients, there were areas with negative values of ΔSVg. In conclusion, ΔSVg is uniform in healthy lungs and minimally influenced by gravity. The significant ΔSVg heterogeneity observed in emphysema allows identification of areas of alveolar destruction and gas trapping and suggests that ΔSVg maps provide useful information for evaluation and planning of emerging treatments that target trapped gas for removal.

Influence of CT reconstruction settings on extremely low attenuation values for specific gas volume calculation in severe emphysema.

RATIONALE AND OBJECTIVES: Emphysema is characterized by lung tissue destruction and trapped gas. On computed tomographic (CT) images, this may be expressed by widespread areas with high specific gas volume (SV(g)). SV(g) is highly sensitive to very low attenuation values, which frequently occur in the CT images of patients with severe emphysema. The purpose of the present work was to study if and how different reconstruction settings and different scanners significantly influence SV(g) distribution, particularly in the very low attenuation range. MATERIALS AND METHODS: Two sets of CT images taken from two different CT scanners at two different lung volumes in 10 healthy volunteers and 18 subjects with severe emphysema were analyzed. Images were reconstructed using two different settings of reconstruction parameters: (1) thin slice thickness and sharp filter and (2) thick slice thickness and smooth filter. For each set of images, average values of SV(g) and their variation (ΔSV(g)) from total lung capacity to residual volume were calculated in the whole lung. RESULTS: Very low attenuation values are always present in CT images when reconstructed with thin slice thickness and a sharp filter and in very large numbers in patients with severe emphysema. SV(g) values were in general significantly higher in patients with emphysema than in healthy subjects, at both total lung capacity and residual volume (P < .001), and were significantly influenced by the reconstruction filter (P < .001) and CT scanner (P < .001). ΔSV(g) was lower in patients with emphysema than in healthy subjects (P < .001) and was significantly affected by the reconstruction setting but not by the CT scanner. CONCLUSIONS: The disproportionate effect of low-attenuation pixels on SV(g) likely causes overestimation of the severity of emphysema and trapped gas. This can be significantly reduced, however, by using thick slices and a smooth filter for image reconstruction. ΔSV(g) is generally robust for quantifying the functional impairment of the lung in severe emphysema.

3D airway tree reconstruction in healthy subjects and emphysema.

Several algorithms for the segmentation of the 3D human airway tree from computed tomography (CT) images have recently been proposed, but the effects of lung volume and the presence of emphysema on segmentation accuracy has not been investigated. Two different sets of CT images taken on nine healthy subjects and nine patients with severe emphysema (FEV(1) = 19 ± 4.1 SD % pred) were used to reconstruct the trachea-bronchial tree by a region-growing algorithm at two different lung volumes: total lung capacity (TLC) and residual volume (RV). The sixth generation was reached in 67% of the healthy subjects and 22% of the emphysematous patients at TLC. At RV, fifth generation was reached in 33 and 11% of healthy subjects and emphysematous patients. At TLC, 67 ± 2 and 39 ± 2% of airways belonging to the fourth generation were successfully reconstructed, respectively in healthy and emphysematous subjects. At RV, the percentage of successful reconstruction was 33 ± 2 and 16 ± 2%, respectively. Segmentation was significantly influenced by the presence of disease (P < 0.001) and lung volume (P < 0.001) at which the CT scans were acquired. Airway tree reconstruction performed by means of a region-growing algorithm depends on lung volume and presence of emphysema, both of which have significant effect, even at the level of lobar and segmental bronchi.

Quantification of trapped gas with CT and 3 He MR imaging in a porcine model of isolated airway obstruction.

PURPOSE: To quantify regional gas trapping in the lung by using computed tomographic (CT)-determined specific gas volume and hyperpolarized helium 3 ((3)He) magnetic resonance (MR) imaging in a porcine model of airway obstruction. MATERIALS AND METHODS: Four porcine lungs were removed after sacrifice for unrelated cardiac experiments, for which animal studies approval was obtained. Dynamic expiratory thin-section CT and (3)He MR imaging were performed during passive deflation from total lung capacity after obstructions were created with inverted one-way endobronchial exit valves in segmental or lobar bronchi to produce identifiable regions of trapped gas. Changes in specific gas volume were assessed from CT data for defined regions of interest within and outside of obstructed segments and for entire lobes. Helium 3 data were analyzed according to the corresponding regional signal reduction during expiration, compared with the total magnetic moment at each time point. RESULTS: In 4.5 seconds of free collapse, volume decreased by 6% +/- 2 (standard error) and 53% +/- 3, respectively, in trapped-gas lobes and in unobstructed regions (P < .0001). Specific gas volume changed by 6% +/- 2 in areas of trapped gas and decreased by 56% +/- 3 in unobstructed regions, from 3.4 mL/g +/- 0.2 to 1.5 mL/g +/- 0.1 (P < .0001). The (3)He signal intensity decreased by 25% +/- 6 and 71% +/- 3, respectively, in trapped-gas and normal regions (P = .0008). In unobstructed regions, the percentage decreases in specific gas volume and (3)He signal intensity were not statistically different from one another (P = .89). CONCLUSION: The results obtained from the model of gas trapping demonstrate that CT-determined specific gas volume and (3)He MR imaging can help identify and quantify the extent of regional trapped gas in explanted porcine lungs.