Automatic pulmonary fissure detection and lobe segmentation in CT chest images

Lightweight Techniques to Improve Generalization and Robustness of U-Net Based Networks for Pulmonary Lobe Segmentation

Lung lobe segmentation in chest CT is relevant to a wide range of clinical applications. However, existing segmentation pipelines often exhibit vulnerabilities and performance degradations when applied to external datasets. This is usually attributed to the size of the available dataset or model. We show that it is possible to enhance generalizability without huge resources by carefully curating the dataset and combining machine learning with medical expertise. Multiple machine learning techniques (self-supervision (SSL), attention (A), and data augmentation (DA)) are used to train a fast and fully-automated lung lobe segmentation model based on 2D U-Net. Our study involved evaluating these techniques on a diverse dataset collected under the RACOON project, encompassing 100 CT chest scans from patients with bacterial, viral, or SARS-CoV2 infections. We compare our model to a baseline U-Net trained on the same dataset. Our approach significantly improved segmentation accuracy (Dice score of 92.8% vs. 82.3%, p < 0.001). Moreover, our model achieved state-of-the-art performance (Dice score of 92.8% vs. 90.8% for the literature's state-of-the-art, p = 0.102) with reduced training examples (69 vs. 231 CT Scans). Among the techniques, data augmentation with expert knowledge displayed the most significant impact, enhancing the Dice score by +0.056. Notably, these enhancements are not limited to lobe segmentation but can be seamlessly integrated into various medical imaging segmentation tasks, demonstrating their versatility and potential for broader applications.

A Systematic Review of Automated Segmentation Methods and Public Datasets for the Lung and its Lobes and Findings on Computed Tomography Images

OBJECTIVES

Automated computational segmentation of the lung and its lobes and findings in X-Ray based computed tomography (CT) images is a challenging problem with important applications, including medical research, surgical planning, and diagnostic decision support. With the increase in large imaging cohorts and the need for fast and robust evaluation of normal and abnormal lungs and their lobes, several authors have proposed automated methods for lung assessment on CT images. In this paper we intend to provide a comprehensive summarization of these methods.

METHODS

We used a systematic approach to perform an extensive review of automated lung segmentation methods. We chose Scopus, PubMed, and Scopus to conduct our review and included methods that perform segmentation of the lung parenchyma, lobes or internal disease related findings. The review was not limited by date, but rather by only including methods providing quantitative evaluation.

RESULTS

We organized and classified all 234 included articles into various categories according to methodological similarities among them. We provide summarizations of quantitative evaluations, public datasets, evaluation metrics, and overall statistics indicating recent research directions of the field.

CONCLUSIONS

We noted the rise of data-driven models in the last decade, especially due to the deep learning trend, increasing the demand for high-quality data annotation. This has instigated an increase of semi-supervised and uncertainty guided works that try to be less dependent on human annotation. In addition, the question of how to evaluate the robustness of data-driven methods remains open, given that evaluations derived from specific datasets are not general.

The Value of Effective Lung Ventilation Area Ratio Based on CT Image Analysis Is a New Index to Predict the Shorter Outcome of Anti-melanoma Differentiation-Associated Protein 5 Positive Dermatomyositis Associated Interstitial Lung Disease: A Single-Center Retrospective Study

Background: Anti-melanoma differentiation-associated protein 5 (MDA5) positive dermatomyositis (MDA5⁺DM) patients have poor outcomes due to rapidly progressive interstitial lung disease (ILD). The accurate assessment of lung involvement is an urgent focus of research. Methods: A computer-aided lung interstitial image analysis technology has been developed, and a quantitative indicator named effective lung ventilation area ratio (ELVAR) that calculates the proportion of the area outside the lung interstitium in lung tissue has been established. 55 newly diagnosed MDA5⁺DM patients and 46 healthy individuals, matched for age and gender, were enrolled in this study. MDA5⁺DM patients were classified into early death group or early survival group according to their survival state within 3 months after diagnosis. Clinical characteristics, laboratory and immunological test results, lung involvement (including ELVAR value) and treatment were compared between early death group and early survival group to determine an index that can predict prognoses of patients with MDA5⁺DM. Results: There were significant differences between early death MDA5⁺DM patients and early survival MDA5⁺DM patients about 12 indices including age of onset, CRP, ferritin, albumin, and pulmonary involvement including severity of type I respiratory failure at diagnosis, P/F ratio, oxygen supplementation, values of ELVAR, FVC, and DLCO. The results of ROC analysis and correlation analysis showed the value of ELVAR had good diagnostic value and widely correlation with many clinical characteristics. Univariate analysis and Multivariate analysis showed four factors including age of onset, ferritin, value of ELVAR, and oxygen supplementation >4 L/min significantly value for poor prognosis in MDA5⁺DM patients. A cutoff value of 0.835 about ELVAR had good predictive power for mortality within 3 months in 54.2% of MDA5⁺DM patients. Conclusion: The value of ELVAR derived from computed tomography image analysis is a new index that can predict poor outcomes in MDA5⁺DM patients with rapidly progressive interstitial lung disease.

Ultrasound mapping of lung changes in idiopathic pulmonary fibrosis

INTRODUCTION

Idiopathic pulmonary fibrosis (IPF) is the most common and severe form of idiopathic interstitial pneumonia, accounted for 20% of cases of interstitial lung disease (ILD). In this study, we sought to compare the lung changes of IPF using a lung ultrasound (LUS) protocol of 12 zones with "current standard" high resolution computed tomography (HRCT) diagnostic method and overlap it with the functional pulmonary test as a complete clinical and imaging evaluation.

METHODS

Thirty-one patients were included in the study and performed HRCT and pulmonary functional tests (PFT). A 12-lung zones' LUS protocol was used and compared with HRCT and PFT.

RESULTS

The HRCT total fibrotic score had a correlation coefficient of 0.454 (P < 0.005) with predicted FVC and 0.713 with predicted DLCO (P < 0.001). Both the median of the number of B-lines and the average of the thickness of the pleural line obtained in the LUS assessment had a positively and statistically significant correlation with the HRCT fibrotic score P < 0.001. The pleural thickness of 2.4 mm is the cut-off value of the mild form of fibrosis with a sensitivity of 0.958 and a specificity of 0.994.

CONCLUSION

B-lines and the average thickness of the pleural line as LUS markers of the fibrotic interstitial syndrome are highly and positively correlated with HRCT score, FVC and DLCO. LUS as a complementary method in the clinical management of IPF could be used more often by skilled clinicians to assess patients in terms of possible diagnosis and monitoring of IPF.

An Efficient Method for the Detection of Oblique Fissures from Computed Tomography images of Lungs

Detection of a pulmonary fissure in lungs is difficult due to its anatomical changeability among humans and it is essential in the clinical environment for accurate localizing and treating the lung abnormalities on a lobe level in human lungs. In this work, an algorithmic approach is proposed to detect the lung oblique fissures from lung computed tomography (CT) images. In the preprocessing module of our approach, the lung structures are enhanced using morphological operation and lung images are de-noised using Wiener filter. In the second module, lung regions are segmented using techniques, namely, thresholding and background subtraction. In the third module of our algorithm, initially, fissure regions are segmented using the active contour model, then by applying the rule based approach on the fissure regions, the oblique fissures are segmented. The proposed algorithm has been tested on 50 images collected from Lung Image Database Consortium (LIDC) and 30 images obtained from Early Lung Cancer Action Program (ELCAP).

Ultrasound Fissure Observation: Assessment of Lung by Pleural-Hub Affiliates

BACKGROUND

Lung ultrasound examination is becoming an even more important part of pulmonologists' clinical routine. As indicated in the literature, the coordinates of any findings on lung parenchyma are based on surface landmarks or conventional quadrants. In our experience we have noticed that lung fissures are clearly detectable as interruptions of the pleural line, but this has never been investigated previously. The aim of this study was to evaluate whether lung fissures are detectable under normal conditions in routine clinical practice.

METHODS

Ten operators were enrolled from Pleural-Hub, a scientific discussion group. After compiling a prestudy survey to investigate whether they had observed fissures previously, they were asked to note if the following fissures were detectable: posterior right (PR) and left (PL), lateral right (LR) and left (LL), anterior right (AR), and anterolateral left (AL). Enrollment was competitive, aiming to reach 100 case subjects.

RESULTS

We found that general fissure detection was 61.3%, in particular: PR, 59%; LR, 75%; AR, 69%; PL, 45%; LL, 64%; AL, 56%. Single operators yielded different detection rates ranging from 90% to 25%, showing strong operator dependency. Before being made aware of general results operators indicated operator's skill and rib shadow artifacts as the principal factors affecting fissure detection.

CONCLUSIONS

Lung fissures may be detected with ultrasound once adequate training is provided. This may allow the clinician a more precise anatomical delineation of pathology identified by lung ultrasound.

Transient Dynamics Simulation of Airflow in a CT-Scanned Human Airway Tree: More or Fewer Terminal Bronchi?

Using computational fluid dynamics (CFD) method, the feasibility of simulating transient airflow in a CT-based airway tree with more than 100 outlets for a whole respiratory period is studied, and the influence of truncations of terminal bronchi on CFD characteristics is investigated. After an airway model with 122 outlets is extracted from CT images, the transient airflow is simulated. Spatial and temporal variations of flow velocity, wall pressure, and wall shear stress are presented; the flow pattern and lobar distribution of air are gotten as well. All results are compared with those of a truncated model with 22 outlets. It is found that the flow pattern shows lobar heterogeneity that the near-wall air in the trachea is inhaled into the upper lobe while the center flow enters the other lobes, and the lobar distribution of air is significantly correlated with the outlet area ratio. The truncation decreases airflow to right and left upper lobes and increases the deviation of airflow distributions between inspiration and expiration. Simulating the transient airflow in an airway tree model with 122 bronchi using CFD is feasible. The model with more terminal bronchi decreases the difference between the lobar distributions at inspiration and at expiration.

Influence of fissure integrity on quantitative CT and emphysema distribution in emphysema-type COPD using a dedicated COPD software

OBJECTIVES

Fissure integrity (FI) plays a key role in selecting patients for interventional emphysema therapy. We investigated its interference with automated lobar segmentation in quantitative computed tomography (CT) and emphysema distribution.

METHODS

CT was available for 50 patients with chronic obstructive pulmonary disease (COPD). Lobe segmentation was performed fully automated by software and corrected manually. FI was evaluated visually using a %-scale. The influence of FI on emphysema ratio (ER=percentage of lung volume with density values<-950 HU), mean lung density (MLD), emphysema and total volume of adjacent lobes was analyzed. Lobe-based results were compared with respect to FI.

RESULTS

Differences in ER in adjacent lobes for complete vs. incomplete fissures were 12.4% for the right horizontal, 0.2% and 3% for the right oblique and 4.4% for the left oblique fissure (all p>0.05). Results for emphysema comparing automated vs. manually corrected segmentation exceeded clinically acceptable values, but were not significantly affected by FI (p>0.05). The widest limits of agreement for ER and MLD were noted in the right middle lobe ([-14, 17.4%], [-22.4, 32.4 Hounsfield Units]).

CONCLUSIONS

Automated lobe segmentation and emphysema distribution are not significantly affected by FI. Manual correction of automated lobar segmentation is still recommended in severe emphysema.

Severity of pulmonary emphysema and lung cancer: analysis using quantitative lobar emphysema scoring

The aim of this study was to determine the relationship between lobar severity of emphysema and lung cancer using automated lobe segmentation and emphysema quantification methods.This study included 78 patients (74 males and 4 females; mean age of 72 years) with the following conditions: pathologically proven lung cancer, available chest computed tomographic (CT) scans for lobe segmentation, and quantitative scoring of emphysema. The relationship between emphysema and lung cancer was analyzed using quantitative emphysema scoring of each pulmonary lobe.The most common location of cancer was the left upper lobe (LUL) (n = 28), followed by the right upper lobe (RUL) (n = 27), left lower lobe (LLL) (n = 13), right lower lobe (RLL) (n = 9), and right middle lobe (RML) (n = 1). Emphysema ratio was the highest in LUL, followed by that in RUL, LLL, RML, and RLL. Multivariate logistic regression analysis revealed that upper lobes (odds ratio: 1.77; 95% confidence interval: 1.01-3.11, P = 0.048) and lobes with emphysema ratio ranked the 1st or the 2nd (odds ratio: 2.48; 95% confidence interval: 1.48-4.15, P < 0.001) were significantly and independently associated with lung cancer development.In emphysema patients, lung cancer has a tendency to develop in lobes with more severe emphysema.

Lung segmentation on standard and mobile chest radiographs using oriented Gaussian derivatives filter

Background

Unsupervised lung segmentation method is one of the mandatory processes in order to develop a Content Based Medical Image Retrieval System (CBMIRS) of CXR. The purpose of the study is to present a robust solution for lung segmentation of standard and mobile chest radiographs using fully automated unsupervised method.

Methods

The novel method is based on oriented Gaussian derivatives filter with seven orientations, combined with Fuzzy C-Means (FCM) clustering and thresholding to refine the lung region. In addition, a new algorithm to automatically generate a threshold value for each Gaussian response is also proposed. The algorithms are applied to both PA and AP chest radiographs from both public JSRT dataset and our private datasets from collaborative hospital. Two pre-processing blocks are introduced to standardize the images from different machines. Comparisons with the previous works found in the literature on JSRT dataset shows that our method gives a reasonably good result. We also compare our algorithm with other unsupervised methods to provide fairly comparative measures on the performances for all datasets.

Results

Performance measures (accuracy, F-score, precision, sensitivity and specificity) for the segmentation of lung in public JSRT dataset are above 0.90 except for the overlap measure is 0.87. The standard deviations for all measures are very low, from 0.01 to 0.06. The overlap measure for the private image database is 0.81 (images from standard machine) and 0.69 (images from two mobile machines). The algorithm is fully automated and fast, with the average execution time of 12.5 s for 512 by 512 pixels resolution.

Conclusions

Our proposed method is fully automated, unsupervised, with no training or learning stage is necessary to segment the lungs taken using both a standard machine and two different mobile machines. The proposed pre-processing blocks are significantly useful to standardize the radiographs from mobile machines. The algorithm gives good performance measures, robust, and fast for the application of the CBMIRS.