Computed tomography-derived pathological phenotypes in COPD

CT-based COPD identification using multiple instance learning with two-stage attention

BACKGROUND AND OBJECTIVE

Chronic obstructive pulmonary disease (COPD) is one of the leading causes of morbidity and mortality worldwide. However, COPD remains underdiagnosed globally. Spirometry is currently the primary tool for diagnosing COPD, but it has unneglected difficulties in detecting mild COPD. Chest computed tomography (CT) has been validated for COPD diagnosis and quantification. Whereas many CT-based deep learning approaches have been developed to identify COPD, it remains challenging to characterize CT-based pathological alternations of COPD which are multidimensional and highly spatially heterogeneous, and the diagnosis performance still needs to be improved.

METHODS

A multiple instance learning (MIL) with two-stage attention (TSA-MIL) is proposed to identify COPD using CT images. Based on transfer learning, a Resnet-50 model pre-trained on natural images is used to extract multicomponent and multidimensional features of COPD abnormalities, in which a pseudo-color method is designed to transfer single-channel CT slices to RGB-like three channels and meanwhile increase the richness of feature representations. To generate more robust attention score for each instance, a two-stage attention module is utilized with the first stage aiming at discovering the key instance while the second stage correcting the attention score for each instance by calculating its average relative distance to the key instances; besides, an instance-level clustering over feature domain is exploited to further improve feature separability and therefore facilitate the subsequent attention module. CT scans, spirometry and demographic data of a total of 800 participants were collected from a large public hospital, with 720 and 80 participants used for model development and evaluation, respectively. In addition, data of 260 participants from another large hospital were also collected for external validation.

RESULTS AND CONCLUSIONS

The proposed TSA-MIL approach outperforms not only most of the advanced MIL models, but also other up-to-date COPD identification methods, with an accuracy of 0.9200 and an area under curve (AUC) of 0.9544 on the test set, and with an accuracy of 0.8115 and an AUC of 0.8737 on the external validation set without multicenter effect reduction, which is clinically acceptable. Therefore, this approach is promising to be a powerful tool for COPD diagnosis in clinical practice.

Image Quality Assessment of Low-Dose CT using Hybrid Iterative Reconstruction

Abstract Background: X-ray computed tomography is the first imaging technology that supports accurate nondestructive interior image reconstruction of an object from sufficient projection data. Low-dose computed tomography (LDCT) has been considered to relieve the harm to patients caused by X-ray radiation. However, LDCT images can be degraded by quantum noise and streak artifacts.Methods: The objective of the authors’ study is to evaluate the optimal level of the hybrid iterative reconstruction (HIR) that generates images with the best diagnostic quality on different dose and noise levels. HIR with optimizations is proposed to reduce image noise and provide better performance at a low dose. The Catphan®504 phantom is employed to assess various image qualities (IQ).Results: For any given scanning protocols, there is linear noise reduction and linear increase of contrast-to-noise ratio (CNR) using optimal HIR. The evidence from various module tests demonstrates that the shape of the noise power spectrum is continuously shifted to low frequency with increasing HIR levels compared with that of filtered-back-projection (FBP). This may describe the difference between the human observer performance and features of the ideal low-contrast objects.Conclusion: Optimal HIR is clearly demonstrated to be a superior method for reducing image noise and improving CNR compared to FBP. Optimal HIR also inhibits texture change or spectrum shift compared with the pure IR method. Even though there are continuous noise reduction and CNR increase with HIR at increasing levels, the human observer performance does not seem to improve simultaneously due to coarser noise (low-frequency noise). HIR level 3 to 5 is optimal for their study. It is possible for the optimal HIR to offer equivalent diagnostic IQ at a lower dose compared with FBP at a routine dose.

DCT-MIL: Deep CNN transferred multiple instance learning for COPD identification using CT images

While many pre-defined computed tomographic (CT) measures have been utilized to characterize chronic obstructive pulmonary disease (COPD), it is still challenging to represent pathological alternations of multiple dimensions and highly spatial heterogeneity. Deep CNN transferred multiple instance learning (DCT-MIL) is proposed to identify COPD via CT images. After the lung is divided into eight sections along the axial direction, one random axial CT image is taken out from each section as one instance. With one instance as the input, the activations of neural layers of AlexNet trained by natural images are extracted as features. After dimension reduction through principle component analysis, features of all instances are input into three MIL methods: Citation k-Nearest-Neighbor (Citation-KNN), multiple instance support vector machine, and expectation-maximization diverse density. Moreover, the performance dependence of the resulted models on the depth of the neural layer where activations are extracted and the number of features is investigated. The proposed DCT-MIL achieves an exceptional performance with an accuracy of 99.29% and area under curve of 0.9826 while using 100 principle components of features extracted from the fourth convolutional layer and Citation-KNN. It outperforms not only DCT-MIL models using other settings and the pre-trained AlexNet with fine-tuning by montages of eight lung CT images, but also other state-of-art methods. Deep CNN transferred multiple instance learning is suited for identification of COPD using CT images. It can help finding subgroups with high risk of COPD from large populations through CT scans ordered doing lung cancer screening.

Impact of COPD on prognosis of lung cancer: from a perspective on disease heterogeneity

Background

COPD is an important comorbidity of lung cancer, but the impact of COPD on the outcomes of lung cancer remains uncertain. Because both COPD and lung cancer are heterogeneous diseases, we evaluated the link between COPD phenotypes and the prognosis of different histological subtypes of lung cancer.

Methods

In this retrospective study, subjects with a newly and pathologically confirmed diagnosis of lung cancer were enrolled from patients preparing for lung cancer surgery. All participants underwent pulmonary function test (PFT). The diagnosis of COPD was based on GOLD criteria. Lung cancer subtypes and COPD phenotypes were categorized by WHO classification of lung tumors and computer quantitative analysis of PFT. The HRs were estimated by Cox regression analysis.

Results

Among 2,222 lung cancer patients, 32.6% coexisted with COPD. After adjustment for age, sex, body mass index (BMI), smoking status, and therapy method, COPD was significantly associated with the decreased overall survival (OS) of lung cancer (HR 1.28, 95% CI 1.05-1.57). With the increased severity of COPD, the OS of lung cancer was gradually worsened (HR 1.23, 95% CI 1.08-1.39). But surgical treatment and high BMI were independent prognostic protective factors (HR 0.46, 95% CI 0.37-0.56; HR 0.96, 95% CI 0.94-0.99). Moreover, in terms of disease heterogeneity, emphysema-predominant phenotype of COPD was an independent prognostic risk factor for squamous carcinoma (HR 2.53, 95% CI 1.49-4.30). No significant relationship between COPD phenotype and lung cancer prognosis was observed among adenocarcinoma, small cell lung cancer, large cell lung cancer, and other subtype patients.

Conclusion

These findings suggest that COPD, especially emphysema-predominant phenotype, is an independent prognostic risk factor for squamous carcinoma only.

The link between chronic obstructive pulmonary disease phenotypes and histological subtypes of lung cancer: a case-control study

Background

COPD is considered an independent risk factor for lung cancer. COPD and lung cancer are both very heterogeneous diseases, and the study herein investigates the link between COPD phenotypes and specific histological subtypes of lung cancer.

Methods

This case–control study comprised 2,283 patients with newly diagnosed pathological lung cancer and 2,323 non-lung cancer controls. All participants underwent pulmonary function tests. The diagnosis of COPD was based on Global Initiative for Chronic Obstructive Lung Disease criteria. Subtypes of the two diseases were categorized according to 2015 World Health Organization classification of lung cancer and computer quantification of airway collapse on maximum expiratory flow volume. ORs were estimated using logistic regression analysis.

Results

The prevalence of COPD was higher (32.8%) in lung cancer patients compared to controls (16.0%). After adjustment for age, sex, body-mass index, and smoking status, the presence of COPD significantly increased the risk of lung cancer (OR 2.88, 95% CI 2.48–3.34) and all common histological subtypes (ORs 2.04–5.26). Both emphysema-predominant and non-emphysema-predominant phenotypes of COPD significantly increased the risk of lung cancer (OR 4.43, 95% CI 2.85–6.88; OR 2.82, 95% CI 2.40–3.31). Higher risk of squamous-cell carcinoma and small-cell lung cancer was observed in patients with the emphysema-predominant than the non-emphysema-predominant phenotype (OR 1.73, 95% CI 1.03–2.89; OR 3.74, 95% CI 1.64–8.53).

Conclusion

COPD was an independent risk factor for lung cancer and all common histological subtypes. Both emphysema-predominant and non-emphysema-predominant phenotypes of COPD significantly increased the risk of lung cancer. Relative to non-emphysema-predominant phenotype of COPD, emphysema-predominant phenotype had a higher risk of squamous-cell carcinoma and small-cell lung cancer.