Predicting mixed venous oxygen saturation (SvO 2 ) impairment in COPD patients using clinical-CT radiomics data: A preliminary study

Abstract

BACKGROUND:

Chronic obstructive pulmonary disease (COPD) is one of the most common chronic airway diseases in the world.

OBJECTIVE:

To predict the degree of mixed venous oxygen saturation (SvO₂) impairment in patients with COPD by modeling using clinical-CT radiomics data and to provide reference for clinical decision-making.

METHODS:

A total of 236 patients with COPD diagnosed by CT and clinical data at Xiangyang No. 1 People’s Hospital ( $n=$ 157) and Xiangyang Central Hospital ( $n=$ 79) from June 2018 to September 2021 were retrospectively analyzed. The patients were divided into group A (SvO ${}_{2}\geqslant$ 62%, $N=$ 107) and group B (SvO ${}_{2}<$ 62%, $N=$ 129). We set up training set and test set at a ratio of 7/3 and time cutoff spot; In training set, Logistic regression was conducted to analyze the differences in general data (e.g. height, weight, systolic blood pressure), laboratory indicators (e.g. arterial oxygen saturation and pulmonary artery systolic pressure), and CT radiomics (radscore generated using chest CT texture parameters from 3D slicer software and LASSO regression) between these two groups. Further the risk factors screened by the above method were used to establish models for predicting the degree of hypoxia in COPD, conduct verification in test set and create a nomogram.

RESULTS:

Univariate analysis demonstrated that age, smoking history, drinking history, systemic systolic pressure, digestive symptoms, right ventricular diameter (RV), mean systolic pulmonary artery pressure (sPAP), cardiac index (CI), pulmonary vascular resistance (PVR), 6-min walking distance (6MWD), WHO functional classification of pulmonary hypertension (WHOPHFC), the ratio of forced expiratory volume in the first second to the forced vital capacity (FEV1%), and radscore in group B were all significantly different from those in group A ( $P<$ 0.05). Multivariate regression demonstrated that age, smoking history, digestive symptoms, 6MWD, and radscore were independent risk factors for SvO₂ impairment. The combined model established based on the abovementioned indicators exhibited a good prediction effect [AUC: 0.903; 95%CI (0.858–0.937)], higher than the general clinical model [AUC: 0.760; 95%CI (0.701–0.813), $P<$ 0.05] and laboratory examination-radiomics model [AUC: 0.868; 95%CI (0.818–0.908), $P=$ 0.012]. The newly created nomogram may be helpful for clinical decision-making and benefit COPD patients.

CONCLUSION:

SvO₂ is an important indicator of hypoxia in COPD, and it is highly related to age, 6MWD, and radscore. The combined model is helpful for early identification of SvO₂ impairment and adjustment of COPD treatment strategies.

Keywords

Cardiac function cardiac index chronic obstructive pulmonary disease (COPD)digestive symptoms hemodynamics mixed venous oxygen saturation (SvO2)pulmonary systolic pressure WHO functional classification of pulmonary hypertension

1. Introduction

Chronic obstructive pulmonary disease (COPD) is one of the most common chronic airway diseases in the world, characterized by continuous progression and repeated acute exacerbations, it has a high risk of death and places a high economic burden on society and families. The incidence rate of people over 40 years old in China is 13.6–15.4%. Nowadays, it is generally believed that the goal of COPD treatment is to slow the decline in lung function, reduce the recurrence of exacerbations, lower the risk of hypoxemia, decrease the risk of death, and prolong life in COPD patients [1, 2]. In China, affected by the accelerated industrialization and heavy air pollution, there is a big proportion of COPD patients with impaired mixed venous oxygen saturation (SvO₂). While the general family’s financial situation and community medical environment make it very difficult to monitor and treat this disease. SvO₂ impairment is a common complication of COPD, and is associated with increased risk of hospitalization, reduced exercise capacity, and survival of COPD patients. Its pathogenesis remains elusive, and the etiology is complex. It is often accompanied by progressive hyperplasia, obstruction, and pulmonary circulatory resistance in pulmonary capillaries. It may cause changes in hemodynamics and cardiac structure, and further lead to right ventricular failure and central hypoxia, and even result in death in severe cases [2, 3, 4]. At present, the treatment of SvO₂ impairment includes rehabilitation therapy, interventional therapy, and drug therapy. It is necessary to create a combined treatment plan based on the hemodynamic indicators of COPD to improve the therapeutic effect. In recent years, N-terminal brain natriuretic peptide precursor (NT-proBNP) and peripheral venous oxygen saturation have also been used to quantitatively monitor the disease, however, there were disadvantages such as complicated operation, invasiveness, infection, and poor repeatability. In the past three years, Li et al. reported “Non-Invasive Physiological Data”, Abbas et al. reported “Index of Tissue Oxygen Delivery IDO2” and Patel et al. reported “Near-Infrared Spectroscopy”. These methods may have a certain effect on SvO₂ injury prediction [5, 6, 7], but it is still in laboratory development stage or only applied to children and is rather not suitable for COPD-SvO₂ injury. Therefore, it is extremely important to search for potential alternatives [6, 8, 9, 10]. In this study, we established a combined model to predict SvO₂ impairment using clinical and CT radiomics data and achieved good results. The newly proposed model employs the advantages of convenience and low price, shed light on the clinical management of COPD, the report is as follows.

2. Materials and methods

2.1 Cases

The clinical data (age, gender, heart rate, height, weight, the systemic systolic blood pressure, diastolic blood pressure, smoking/drinking history, digestive symptoms (e.g. anorexia, nausea, indigestion, diarrhea, vomit)) of 271 patients with COPD diagnosed by pulmonary function, CT and laboratory examinations in the respiratory department of Xiangyang No. 1 People’s Hospital and Xiangyang Central Hospital from June 2018 to September 2021 were retrospectively analyzed. The study included 162 males and 109 females, aged 36–88 years old, with an average age of 63.4 $\pm$ 12.3.

Inclusion criteria: patients with COPD diagnosed by pulmonary function, CT and laboratory examinations in the respiratory department, and with complete follow-up data Exclusion criteria: COPD patients with lung or other tumors; incomplete medical record. Finally, 236 patients were enrolled and included in Xiangyang No. 1 People’s Hospital ( $n=$ 157) and Xiangyang Central Hospital ( $n=$ 79). The training set and test set was established at a ratio of 7/3 and time cutoff spot. The COPD further prognostic criteria of the World Health Organization (WHO) and Chinese Medical Association were adopted. In detail, good prognosis means stable condition, $<$ 3 acute attacks per year, no need for continuous oxygen therapy, and normal cardiac function; while poor prognosis means hospitalization for $\geqslant$ 3 attacks per year, continuous home oxygen therapy, and exercise limitation. Based on the above prognosis guidelines, the receiver operating characteristic curve (ROC) analysis was performed and obtained a largest area under the curve [AUC: 0.911; OR: 0.883; 95%CI (0.859–0907), $P<$ 0.05] for predicting the prognosis of COPD at the SvO₂ of 62% (as cutoff value). Furthermore, according to the inclusion and exclusion criteria of this study, we took SvO₂ $=$ 62% as the cut-off to divide the COPD patients into two groups, the SvO₂ $\geqslant$ 62% group (group A, good prognosis) and the SvO₂ $<$ 62% group (group B, poor prognosis). There were 107 cases in group A, 59 males and 48 females, aged 36 $\sim$ 82 years, and 129 cases in group B, 69 males and 60 females, aged 50 $\sim$ 80 years old [3, 9] (Fig. 1).

Figure 1.

Flowchart showing inclusion and exclusion of subjects in this study.

2.2 Methods

2.2.1 Ultrasound, laboratory and pulmonary function examination

Echocardiography was used to measure the right atrioventricular end-diastolic diameter of the COPD patients. The 6-min walking distance (6MWD) of COPD patients were recorded. Right heart catheterization was used to record the hemodynamic data of COPD patients, including mean systolic pulmonary artery pressure (sPAP), mean right atrial pressure, pulmonary vascular resistance (PVR), pulmonary capillary wedge pressure (PCWP), cardiac output (CO), SvO₂ and cardiac index (CI). According to the WHO functional classification of pulmonary hypertension (WHOPHFC), the cases classified as WHO functional class I and II were considered good exercise tolerance, and III and IV were considered poor exercise tolerance [8, 9]. Regarding pulmonary function measurement, forced vital capacity (FVC), forced expiratory volume in the first second (FEV1), residual volume and total lung capacity (TLC) were measured using a spirometer (Jeager, Germany). Lung volume was measured by plethysmography. Then the patients took two puffs (200 $\mu$ g) of salbutamol aerosol, FEV1 was remeasured 15 minutes later, and FEV1% was estimated [11, 12, 13].

2.2.2 CT scan

CT scans were performed using GE Light Speed 64-row 128-slice spiral CT scanner, with tube voltage 120 kV, tube current 50–200 mAx, matrix size 512 $\times$ 512, and the pixel spacings of/the slice thickness of CT scans 5 mm. The scan was performed from bilateral lung apexes to the pancreas. The CT images were reconstructed at slice thickness of 1.0 mm and slice interval of 1.0 mm. The images were sent to the PACS server in the Dicom format. Two senior radiologists with over 10 years of working experience reviewed the CT images of both lung on the PACS server. An agreement was reached concerning the interpretation of the CT scan results through the negotiations between the two radiologists. The window width was set to 300–400 HU, and the window position $-$ 50 to $-$ 60 HU [14, 15]. To ensure the uniformity of radiomics texture parameters extraction, we have made unified parameter adjustments to all CT scanners, including grayscale, tube voltage, tube current and interlayer spacing.

2.2.3 Chest CT radiomic analysis

The transverse diameter, anteroposterior diameter, superoinferior diameter, and average CT value of both lungs were measured for chest texture parameters analysis. During CT value measurement, the region of interest (ROI) was delineated to cover the entire lung, avoiding artifacts and mediastinal fat. Each measurement was repeated three times, and the average was taken. The ROI was delineated in the entire both lung and the high-throughput radiographic texture parameters were extracted using 3D Slicer 5.13 (including morphological features, gray-level size zone matrix, histogram features, gray-level co-occurrence matrix, and length matrix). First step, these texture parameters were analyzed using the six machine learning algorithms with k-fold cross-validation (including XGboost, SVM, Naive Bayes, Random forest, KNN, Bagging); the largest area under the ROC curve of the mentioned algorithms were compared. Second step, CT radiomics was analyzed and radscore was generated using LASSO regression with 10-fold cross-validation after the differences were verified between the two groups of radiomics using machine learning (all AUCs $>$ 0.60) [16, 17, 18] (Fig. 2).

Figure 2.

Schematic representation and 3D reconstruction of COPD chest CT texture data extracted with 3D slicer software.

A brief description of texture features is as follows. First order features are calculated from the signal strength distribution in the region of interest (ROI), including features that describe the trend of the data center, such as mean, median, and mode, and features that describe the symmetry and heterogeneity of the distribution, such as percentile, skewness, kurtosis, and entropy. This represents the geometric attributes of the lesion, such as volume, diameter, surface area and intensity distribution within ROI, but ignores the spatial location of each voxel. Texture or second-order features represent joint statistical data of two or more voxels, so in coarse texture examples, adjacent pixel pairs may have similar grayscale levels, while in fine texture examples, adjacent pixel values are independent. In radiological images, the statistical dependencies between adjacent voxels may be more complex than these simple examples, so features derived from grayscale co-occurrence matrix (GLCM), grayscale run length matrix (GLRLM), and other indicators can effectively quantify image texture differences. Other complex texture parameter descriptions are no longer discussed in this article [17, 18].

2.3 Statistical analysis

Data processing was performed using SPSS Version 22.0 (IBM Corp., Armonk, NY, USA). Measurement data were expressed as mean $\pm$ standard deviation. The between-group comparisons were conducted using the independent samples $t$ test (for normal distribution data) and rank sum test (non-normal distribution data). The ratio comparison between groups were conducted using the chi-square test. The k/10-fold cross-validation was selected to adjust the elastic network parameters to select the best combination of feature parameters to avoid overfitting in 6 machine learning and least absolute shrinkage and selection operator (LASSO). The logistic regression was used to establish prediction models, and the ROC and AUC were used to evaluate the model efficiency. R software was used to make nomograms, decision curve, calibration curve. $P<$ 0.05 was considered statistically significant [19, 20, 21].

3. Results

According to the WHO functional classification of pulmonary hypertension, there were 85 cases of grade I, 39 cases of grade II, 41 cases of grade III, and 71 cases of grade IV among the 236 patients in this study.

Our team extract the high-throughput texture information from the conventional chest CT scan images using used 3D Slicer software version 5.1. We finally obtained 85 groups of texture parameters after deleting the confounding factors and lost packets; including Histogram features, GLCM, GLRLM, GLSZM, NGTDM and GLDM.

The above available texture parameters were analyzed using two groups of methods: (1) six machine learning algorithms AUC curve value (bagging-0.739, SVM-0.650, Naive Bayes-0.719, Random forest-0.831, XGboost-0.856, KNN-0.604, all AUCs $>$ 0.60) finally, the XGboost was confirmed obtained the maximum AUC ( $P<$ 0.05); That is to say, chest CT radiomics is feasible to differentiate COPD-SVO₂ impairment confirmed by machine learning. (2) Lasso regression method: 17 useful radiomics texture data were selected by lasso regression, and then the final weighted radiomics score (radscore) was calculated and generated ( $P<$ 0.05) (Fig. 3).

Table 1
Regression analysis results of establishing the general clinical model based on clinical characteristics to predict SvO₂impairment in COPD

General clinical model	Univariate analysis		Multivariate analysis
	$P$	Hazard ratio	$P$	Hazard ratio
Height (cm)	0.12	1.02(0.99–1.05)
Weight (kg)	0.09	0.98(0.95–1.01)
Age (years)	$<$ 0.05^*	1.09(1.05–1.14)	$<$ 0.05^*	1.08(1.03–1.12)
Gender (male/female)	0.81	0.94(0.56–1.57)
Smoking history	0.04^*	1.03(1.01–1.06)	0.02^*	1.04(1.01–1.08)
Drinking history	0.04^*	1.04(1.01–1.08)	$<$ 0.05^*	1.07(1.02–1.13)
Systolic blood pressure (mmHg)	$<$ 0.05^*	0.97(0.95–0.98)	$<$ 0.05^*	0.96(0.94–0.98)
Diastolic blood pressure (mmHg)	0.27	0.98(0.97–1.01)
Heart rate (times/min)	0.09	0.98(0.96–1.01)
Digestive symptoms	$<$ 0.05^*	2.04(1.21–3.46)	0.02^*	2.03(1.13–3.64)

Table 2

Regression analysis results of establishing the laboratory examination model based on laboratory, pulmonary function, echocardiography characteristics, and chest CT radiomics to predict SvO₂ impairment in COPD

Laboratory examination-radiomics model	Univariate analysis		Multivariate analysis
	$P$	Hazard ratio	$P$	Hazard ratio
PaCO₂ (mmHg)	0.62	1.01(0.98–1.03)
PaO₂ (mmHg)	0.06	0.97(0.95–1.01)
Arterial oxygen saturation (%)	0.23	0.94(0.84–1.04)
RV (mm)	0.04^*	1.07(1.01–1.14)
RA (mm)	0.08	0.98(0.95–1.01)
sPAP (mmHg)	0.01^*	1.02(1.01–1.04)
CI (Lmin^{- 1}m^{- 2})	$<$ 0.05*	0.59(0.39–0.88)
PVR (Wood U)	0.01^*	1.11(1.03–1.19)	0.04^*	1.11(1.00–1.24)
6min walking distance-6MWD (m)	$<$ 0.05^*	0.98(0.97–0.99)	0.01^*	0.99(0.97–0.99)
WHO PHFC (III, IV)	0.04^*	1.24(1.01–1.53)
FEV₁ (L)	0.12	0.49(0.19–1.20)
FEV₁%	0.04*	0.98(0.96–1.01)
Radscore	$<$ 0.05^*	2.16(1.76–2.64)	$<$ 0.05^*	2.17(1.75–2.69)

Notes: PaCO₂: partial pressure of carbon dioxide in artery; PaO₂: arterial oxygen partial pressure; RV: right ventricle diameter; RA: right atrium diameter; CI: cardiac index; PVR: pulmonary vascular resistance; sPAP: systolic pulmonary artery pressure; WHO PHFC: WHO pulmonary hypertension functional class.

Table 3

Regression analysis results of establishing the combined model based on the above risk factors to predict SvO₂impairment in COPD

Combined model	Univariate analysis		Multivariate analysis
	$P$	Hazard ratio	$P$	Hazard ratio
Age (years)	$<$ 0.05*	1.09(1.05–1.14)	0.01*	1.08(1.02–1.15)
Smoking history	0.04*	1.03(1.00–1.06)	$<$ 0.05*	1.07(1.02–1.11)
Drinking history	0.04*	1.04(1.00–1.08)
Systolic blood pressure (mmHg)	$<$ 0.05*	0.97(0.95–0.98)
Digestive symptoms	$<$ 0.05*	2.04(1.21–3.46)	0.03*	2.37(1.11–5.07)
RV (mm)	0.04*	1.07(1.01–1.14)
sPAP (mmHg)	0.01*	1.02(1.00–1.04)
CI (Lmin^{- 1}m^{- 2})	$<$ 0.05*	0.59(0.39–0.88)
PVR (Wood U)	0.01*	1.11(1.02–1.19)
6min walking distance-6MWD (m)	$<$ 0.05*	0.98(0.97–0.99)	0.01*	0.98(0.97–0.99)
WHO PHFC (III, IV)	0.04*	1.24(1.01–1.53)
FEV₁%	0.04*	0.98(0.96–0.99)
Radscore	$<$ 0.05*	2.16(1.76–2.64)	$<$ 0.05*	2.24(1.74–2.87)

Figure 3.

The comparison results of 6 machine learning ROC curves confirm that XGboost has the highest diagnostic efficiency; and all AUCs $>$ 0.60, that is to say, chest CT radiomics is feasible to differentiate COPD-SVO2 impairment confirmed by machine learning.

Figure 4.

Delong nonparametric test showed that the AUC of combined model is higher than another two models between training set and test set (C: combined model; B: laboratory examination-radiomics model; A: general clinical model).

Univariate analysis demonstrated that the systemic systolic blood pressure, CI, FEV1%, and 6MWD in group B (SvO₂ $<$ 62%) were all lower than those in group A (SvO₂ $\geqslant$ 62%), with statistically significant differences ( $P<$ 0.05). The age, smoking history drinking history digestive symptoms, RV, sPAP PVR, WHOPHFC, radscore were all higher than those in group A, and the differences were statistically significant ( $P<$ 0.05) (Tables 1 and 2). The multi-factor logistic regression indicated that age, smoking history, digestive symptoms, 6MWD and radscore were independent risk factors associated with SvO₂ impairment in COPD patients. In the training set, the combined model based on the above risk factors performed nicely on predicting SvO₂ impairment [AUC:0.903; 95%CI (0.858–0.937)], accuracy-0.818, specificity-0.785, sensitivity-0.845, better than the general clinical model [AUC: 0.760; 95%CI (0.701–0.813), $P<$ 0.05] and laboratory examination-radiomics model [AUC: 0.868; 95%CI (0.818–0.908), $P=$ 0.012]. The external validation on the testing set yielded the same result: The combined model [AUC:0.922; 95%CI (0.832–0.972)], accuracy-0.871, specificity-0844, sensitivity-0.895, outperformed the general clinical model [AUC:0.759; 95%CI (0.642–0853), $P<$ 0.05], and laboratory examination-radiomics model [AUC:0.809; 95%CI (0.698–0.893)], $P=$ 0.023; The decision curve analysis also confirmed the higher net benefit of the combined model in both sets (Figs 4, 5 and Table 3). The nomogram simplified the method of SvO₂ impairment prediction (Fig. 6). The final equation of their binary logistic regression analysis is as follow:

$\displaystyle P(Y=1/X)=$ $\displaystyle\quad[1+e^{-(-7.89+0.08\ast\textit{age}+0.06\ast\textit{smoking % history}+0.86\ast\textit{digestive symptoms}-0.02\ast\textit{6MWD}+0.81\ast% \textit{Radscore})}]^{-1}$

4. Discussion

COPD, as a common progressive disease in respiratory medicine, tends to reoccur and is often accompanied by changes in cardiopulmonary hemodynamics. Mixed venous blood oxygen saturation (SvO₂) impairment often accompanies COPD. SvO₂ impairment can not only reflect the blood perfusion at the tissue and cell level, but also indirectly reflect the cardiac output, as well as the oxygenation degree of tissue and cells. In severe cases, it may even damage the cardiac function. The etiology of COPD is complex and uncertain, and it may be caused by various pathogenic factors including cold air, tobacco, dust, bacteria and viruses. It often manifests as progressive shortness of breath with light exercises, dyspnea, fatigue, and decreased exercise tolerance. In severe cases, accompanied by severe persistent SvO₂ impairment, shock or even disability may occur, and the morbidity and mortality rates of COPD are high, 9% $\sim$ 10% and 5.7%, respectively. The treatments of COPD include adjusted physical activity, avoidance of high altitudes, prevention of infection, drug treatment, and surgical treatment. But common climate change can easily lead to prolonged recurrence of COPD and produce a more severe COPD-SvO₂ impairment. Therefore, the monitor and prevention of COPD is crucial. At present, COPD is often monitored using indicators such as NT-proBNP and sPAP in most hospitals. However, the effect of the above indicators on COPD-SvO₂ monitoring is very limited due to the complicated operation, invasiveness, large error, and high price [6, 12, 22]. The combined model newly developed in this study based on the clinical-chest CT radiomics indicators provides an opportunity for CPOD-SVO₂ monitoring. This non-invasive tool provides support for COPD clinical decision-making and greatly benefits COPD patients.

Figure 5.

Decision curve (DCA) of different prediction models (C: combined model; B: laboratory examination-radiomics model; A: general clinical model) of COPD was established, the net benefit of the combined model is the largest compared with another two models (all $P<$ 0.05).

Figure 6.

Nomogram (left) and calibration curve (right) of risk factors of SvO₂ impairment in COPD was applied in clinical trials.

The superior vena cava receives venous blood from the head and neck, upper limbs, chest wall and some thoracic organs, the inferior vena cava receives venous blood from the lower limbs, abdomen and pelvis, and the coronary sinus receives 90% of the venous blood of the heart. These three meet together as mixed venous blood (SvO₂). SvO₂ is a classic indicator evaluating the body’s oxygen utilization. It is often used to direct fluid resuscitation, fluid management and the use of vasoactive drugs in septic shock patients, and it can show the severity of a patient’s condition. When SvO₂ fluctuates, each influencing factor should be investigated one by one, and a comprehensive analysis should be conducted in combination with cardiac output and arterial lactate. Dynamical monitoring of the changes of SvO₂ is more meaningful for the evaluation and management of critically ill patients [23, 24]. However, it requires repeatedly blood drawing through the cardiac catheter for examination. It is difficult to take care of the central venous catheter, which may easily cause infection, therefore the application in clinical practice remains inconvenient [25, 26]. Our previous initial results demonstrated that age, smoking history, 6MWD were consistent and strongly correlated with SvO₂, suggesting that the abovementioned factors may substitute SvO₂ to a certain extent. With simple operation and low-cost features, they show great potential in patients with COPD and other diseases. This further study used modeling to predict SvO₂ impairment in COPD patients based on clinical-laboratory-CT radiomics data. In this study, univariate analysis indicated that the systemic systolic blood pressure, CI, FEV1%, 6MWD age, smoking history drinking history digestive symptoms, RV, sPAP PVR, WHOPHFC, radscore were risk factors for SvO₂ impairment ( $P<$ 0.05). The 6MWD test is the most important and low-cost method for the evaluation of exercise tolerance in COPD patients, and it is significantly correlated with prognosis [3, 9, 28]. We revealed that it was also associated with SvO₂ impairment. However, the specificity is weak. Nevertheless, the 6MWD test is simple and requires no deep vein catheterization. It is an economical and convenient predicting method. The progressive increase in PVR may lead to right ventricular failure and eventually death, and it is an important indicator of the prognosis of COPD patients with impaired SvO₂ and pulmonary hypertension [27, 28, 29]. The WHO pulmonary hypertension functional class and FEV1% are closely related to the severity of COPD. They demonstrate the patient’s comprehensive cardiopulmonary function reserve, that is, the compensatory ability with increased load. Therefore, as predictors, these two could better present the underlying hemodynamics of COPD patients. COPD patients with poor basal cardiac function and low CI have relatively insufficient cardiorespiratory reserve. Therefore, when increased exercise load and pulmonary vasoconstriction induce severe hemodynamic fluctuations, decompensation will occur soon and thus lead to severe right ventricular failure, resulting in severe impaired SvO₂. Therefore, sPAP and CI are important predictors for SvO₂ impairment. sPAP can be estimated by echocardiographic tricuspid regurgitation velocity (4V ${}^{2}+\Delta P$ ), providing strong data support and convenience for clinical management of COPD [9, 30]. In addition, the newly discovered digestive symptoms in this study have important significance as a new influencing factor of SvO₂ impairment, It is reported that this may be related to the gastrointestinal hypoxia caused by blood oxygen dysfunction in patients with COPD [30, 31, 32]. However, this study design is observational and cannot establish causation between both. It would be more crucial to discuss these associations as correlations or potential relationships that require further investigation and massive data analysis in future. The right atrioventricular diameter and systemic systolic/diastolic blood pressure are the more classic hemodynamic indicators of COPD, so they are also effective in predicting SvO₂ impairment. At the same time, these indicators demonstrated that the basic condition and exercise tolerance of the group with SvO ${}_{2}<$ 62% were poor. The concept of radiomics was first proposed by Dutch scholar Lambin in 2012. It refers to the high-throughput extraction of a large amount of image information from images (e.g. CT, MRI and PET), achieving tumor segmentation, feature extraction, and model building. By deeper mining, prediction, and analysis of massive data, it assists physicians in making the most accurate diagnosis. In recent years, with the addition of machine learning algorithms and artificial intelligence technology, radiomics research has played a greater role in the diagnosis and treatment of tumors and inflammatory diseases. The texture score of chest CT image is an important index for predicting SVO2 impairment newly discovered in this study. Its AUC and sensitivity are as high as 0.825/0.814. It not only reflects the morphological changes of the lungs of COPD patients, but also includes important information such as lung air content, lung tension, interlobular thickening, pulmonary edema, lung tissue fibroelasticity, which indicates its potential prediction performance. In addition to showing the respiratory function and oxygenation status of COPD patients, these indicators may also tell the severity of the COPD disease and can be used to evaluate the prognosis preliminarily, so that necessary measure and treatment can be taken in advance In other words, we may imply that SvO₂ may reflect also the cardiac function elimination and severity to some extent [33, 34, 35, 36, 37]. In addition, the newly developed nomogram tool in this study has high clinical value in COPD-SvO₂ impairment prediction. The various variables included are readily more available and easily measurable in clinical settings, such as age, smoking history, and gastrointestinal symptoms, which can be obtained from clinical data. 6MWD can be obtained through simple physical fitness test, and radscore can be obtained through texture extraction software included in CT scanners or “3D slicer” free software. The newly created nomogram has received better clinical praise from our hospital, without any additional expense, and is convenient and fast, worthy of promotion by basic health institutions.

4.1 Limitations

The cases scale of this study was insufficient. In the future, multi-center studies involving multiple hospitals will be considered. This study also lacked radiomics extraction based on echocardiography, therefore, the mining of ultrasound-CT data was not deep enough. In addition, we plan to apply deep learning and artificial intelligence (AI) to build more perfect models to verify our results in the future. Deep learning can not only improve the efficiency and accuracy of lung texture delineation, develop semi-automatic delineation tools, but more importantly, discover more valuable clinical imaging parameters, not limited to $P<$ 0.05 influencing factors in this study [38, 39, 40]. Based on AI Big data training model, SVO₂ impairment in COPD population may be automatically identified, just as AI identifies benign and malignant pulmonary nodules [41, 42, 43]. In addition, this study also has the following limitations, such as the reliance on CT radiomics data and the challenges associated with the interpretation and reproducibility of radiomics features, furthermore, suggestions and explorations for future research directions, including potential areas of improvement or investigation, would enhance the study’s contribution to the field [44, 45].

5. Conclusions

Age, smoking history, digestive symptoms, 6MWD and radscore may partially substitute SvO₂ to monitor tissue hypoxia in COPD patients. The nomogram created based on the combined model can be used to preliminarily predict the SvO₂ impairment preliminarily in COPD patients and provides a fast and effective method for clinical assessment of COPD patients.

Ethics statement

The experimental protocol was established according to the ethical guidelines of the Declaration of Helsinki and was approved by the Human Ethics Committee of Xiangyang No.1 People’s Hospital, Hubei University of Medicine, Affiliated Hospital of Nanjing University of Chinese Medicine (Issue No. 7836-1 [2018]).

Informed consent

Written informed consent was obtained from participants or their guardians.

Availability of data and materials

All data generated or analysed during this study are included in this published article.

Author contributions

PA, JL and ZW conceived and drafted the manuscript. AP and JW contributed to the literature review MY and JW are responsible for the quality control of the statistics. ZW critically revised the manuscript for important intellectual content ZW and JW approved the final version to be published and agreed to act as guarantors of the work.

Funding

The study was supported by the “323” Public Health Project of the Hubei health Commission and the Xiangyang No. 1 People’s Hospital (No. XYY2022-323) and the Natural Science Foundation of Hubei Province (Grant no. 2022BCE021 and 2022CFD010).

Footnotes

Acknowledgments

The authors gratefully acknowledge Peng Duan, Weiping Gu and Kai Lian for assistance with translating references and improving statistical technology.

Conflict of interest

The authors declare no conflicts of interest.

References

Zhou

Zhao

Chen

. The recent advances of phenotypes in acute exacerbations of COPD. Int J Chron Obstruct Pulmon Dis. 2017; 12: 1009-1018. Published 2017 Mar 27. doi: 10.2147/COPD.S128604.

Ahrens

. Continuous mixed venous (SvO) monitoring. Too expensive or indispensible? Crit Care Nurs Clin North Am. 1999; 11(1): 33-48.

Christenson

Smith

Bafadhel

Putcha

. Chronic obstructive pulmonary disease. Lancet. 2022 Jun 11; 399(10342): 2227-2242. doi: 10.1016/S0140-6736(22)00470-6. Epub 2022 May 6. PMID: 35533707.

López-Campos

Tan

Soriano

. Global burden of COPD. Respirology. 2016; 21(1): 14-23. doi: 10.1111/resp.12660.

Pinsky

Dubrawski

. Automated Assessment of Cardiovascular Sufficiency Using Non-Invasive Physiological Data. Sensors (Basel). 2022; 22(3): 1024. Published 2022 Jan 28. doi: 10.3390/s22031024.

Abbas

Hussain

MZH

Shahbaz

Siddiqui

NUR

Hasan

. Performance of a Risk Analytic Tool (Index of Tissue Oxygen Delivery “IDO2”) in Pediatric Cardiac Intensive Care Unit of a Developing Country. Front Pediatr. 2022; 10: 846074. Published 2022 Jun 3. doi: 10.3389/fped.2022.846074.

Patel

NTP

Templeton

Lane

Williams

Neff

Goenaga-Diaz

. Near-Infrared Spectroscopy for Determination of Cardiac Output Augmentation in a Swine Model of Ischemia-Reperfusion Injury. Crit Care Explor. 2022; 4(8): e0749. Published 2022 Aug 15. doi: 10.1097/CCE.0000000000000749.

Soriano

Polverino

Cosio

. What is early COPD and why is it important? Eur Respir J. 2018; 52(6): 1801448. Published 2018 Dec 6. doi: 10.1183/13993003.01448-2018.

Galiè

Humbert

Vachiery

, et al. 2015 ESC/ERS Guidelines for the diagnosis and treatment of pulmonary hypertension: The Joint Task Force for the Diagnosis and Treatment of Pulmonary Hypertension of the European Society of Cardiology (ESC) and the European Respiratory Society (ERS): Endorsed by: Association for European Paediatric and Congenital Cardiology (AEPC), International Society for Heart and Lung Transplantation (ISHLT). Eur Heart J. 2016; 37(1): 67-119. doi: 10.1093/eurheartj/ehv317.

10.

Refaee

Ibrahim

, et al. The Emerging Role of Radiomics in COPD and Lung Cancer. Respiration. 2020; 99(2): 99-107. doi: 10.1159/000505429.

11.

Jung

Park

Hwang

You

JSH

Choi

. Effects of core-postural stabilisation on fluoroscopy diaphragmatic measurement and dyspnea in chronic obstructive pulmonary disease: A randomized single-blinded clinical trial. Technol Health Care. 2021; 29(S1): 359-366. doi: 10.3233/THC-218034.

12.

Wang

Ying

Liu

Zhang

. Findings of ventilator-measured P01 in assessing respiratory drive in patients with severe ARDS [published online ahead of print, 2023 Jun 20]. Technol Health Care. 2023. doi: 10.3233/THC-230096.

13.

Peng

Lin

, et al. Exploration of quality criteria for the detection of peak inspiratory flow under different resistance conditions [published online ahead of print, 2023 Jun 20]. Technol Health Care. 2023. doi: 10.3233/THC-220905.

14.

Wang

Zhao

Zhu

Dou

. False positive reduction in pulmonary nodule classification using 3D texture and edge feature in CT images. Technol Health Care. 2021; 29(6): 1071-1088. doi: 10.3233/THC-181565.

15.

Wang

Shi

Cui

, et al. A robust and efficient framework for tubular structure segmentation in chest CT images Technol. Health Care. 2021; 29(4): 655-665. doi: 10.3233/THC-202431.

16.

Yang

Guo

, et al. Lung radiomics features for characterizing and classifying COPD stage based on feature combination strategy and multi-layer perceptron classifier. Math Biosci Eng. 2022; 19(8): 7826-7855. doi: 10.3934/mbe.2022366.

17.

Lambin

Rios-Velazquez

Leijenaar

, et al. Radiomics: extracting more information from medical images using advanced feature analysis. Eur J Cancer. 2012; 48(4): 441-446. doi: 10.1016/j.ejca.2011.11.036.

18.

Gillies

Kinahan

Hricak

. Radiomics: Images Are More than Pictures, They Are Data. Radiology. 2016; 278(2): 563-577. doi: 10.1148/radiol.2015151169.

19.

Nambi

Abdelbasset

Alrawaili

Alsubaie

Abodonya

Saleh

. Virtual reality or isokinetic training; its effect on pain, kinesiophobia and serum stress hormones in chronic low back pain: A randomized controlled trial. Technol Health Care. 2021; 29(1): 155-166. doi: 10.3233/THC-202301.

20.

Begić

Aziri

Omeragić

, et al. Heat-not-burn tobacco products and cardiovascular risk reduction: A systematic review of randomized controlled trials. Technol Health Care. 2023; 31(4): 1457-1491. doi: 10.3233/THC-220677.

21.

Liu

, et al. Risk assessment of type 2 diabetes in northern China based on the logistic regression model. Technol Health Care. 2021; 29(S1): 351-358. doi: 10.3233/THC-218033.

22.

Gloeckl

Marinov

Pitta

. Practical recommendations for exercise training in patients with COPD. Eur Respir Rev. 2013; 22(128): 178-186. doi: 10.1183/09059180.00000513.

23.

Smith

Wrobel

. Epidemiology and clinical impact of major comorbidities in patients with COPD. Int J Chron Obstruct Pulmon Dis. 2014; 9: 871-888. Published 2014 Aug 27. doi: 10.2147/COPD.S49621.

24.

Wedzicha

Miravitlles

Hurst

, et al. Management of COPD exacerbations: a European Respiratory Society/American Thoracic Society guideline. Eur Respir J. 2017; 49(3): 1600791. Published 2017 Mar 15. doi: 10.1183/13993003.00791-2016.

25.

Negewo

Gibson

McDonald

. COPD and its comorbidities: Impact, measurement and mechanisms. Respirology. 2015; 20(8): 1160-1171. doi: 10.1111/resp.12642.

26.

Pooler

Beech

. Examining the relationship between anxiety and depression and exacerbations of COPD which result in hospital admission: a systematic review. Int J Chron Obstruct Pulmon Dis. 2014; 9: 315-330. Published 2014 Mar 29. doi: 10.2147/COPD.S53260.

27.

Chan

SMH

Selemidis

Bozinovski

Vlahos

. Pathobiological mechanisms underlying metabolic syndrome (MetS) in chronic obstructive pulmonary disease (COPD): clinical significance and therapeutic strategies. Pharmacol Ther. 2019; 198: 160-188. doi: 10.1016/j.pharmthera.2019.02.013.

28.

André

Conde

Fragoso

, et al. COPD and Cardiovascular Disease. Pulmonology. 2019; 25(3): 168-176. doi: 10.1016/j.pulmoe.2018.09.006.

29.

Onishi

. Total management of chronic obstructive pulmonary disease (COPD) as an independent risk factor for cardiovascular disease. J Cardiol. 2017; 70(2): 128-134. doi: 10.1016/j.jjcc.2017.03.001.

30.

Samant

Tran

Uddo

deBoisblanc

Saketkoo

Saito

Helmcke

Lammi

. Use of Handheld Ultrasound to Estimate Right Atrial Pressure in a Pulmonary Hypertension Clinic. Ann Am Thorac Soc. 2022 Feb; 19(2): 179-185. doi: 10.1513/AnnalsATS.202101-092OC. PMID: 34214012.

31.

van Buul

Kasteleyn

Chavannes

Taube

. Association between morning symptoms and physical activity in COPD: a systematic review. Eur Respir Rev. 2017; 26(143): 160033. Published 2017 Jan 3. doi: 10.1183/16000617.0033-2016.

32.

Raftery

Tsantikos

Harris

Hibbs

. Links Between Inflammatory Bowel Disease and Chronic Obstructive Pulmonary Disease. Front Immunol. 2020; 11: 2144. Published 2020 Sep 11. doi: 10.3389/fimmu.2020.02144.

33.

Lewandowski

Koller

. Transforming medical sciences with high-performance computing, high-performance data analytics and AI. Technol Health Care. 2023; 31(4): 1505-1507. doi: 10.3233/THC-237000.

34.

Jamshidnezhad

Hosseini

Mahmudi

Mohammadi-Asl

. A machine learning technology to improve the risk of non-invasive prenatal tests. Technol Health Care. 2022; 30(4): 951-965. doi: 10.3233/THC-213628.

35.

Yang

Liu

Sui

Wang

Gai

. Construction of U-Net++ pulmonary nodule intelligent analysis model based on feature weighted aggregation. Technol Health Care. 2023; 31(S1): 477-486. doi: 10.3233/THC-236041.

36.

Xiao

Wei

Huang

Chen

. SVM classifier of cervical histopathology images based on texture and morphological features. Technol Health Care. 2023; 31(1): 69-80. doi: 10.3233/THC-220031.

37.

Zhu

Zhang

, et al. Severity detection of COVID-19 infection with machine learning of clinical records and CT images. Technol Health Care. 2022; 30(6): 1299-1314. doi: 10.3233/THC-220321.

38.

Fojnica

Osmanovi

Badnjevic

. Dynamical Model of Tuberculosis-Multiple Strain Prediction based on Artificial Neural Network[C]//5th Mediterranean Conference on Embedded Computing (MECO). IEEE, 2016. doi: 10.1109/MECO.2016.7525763.

39.

Badnjevic

Cifrek

Koruga

. Classification of Chronic Obstructive Pulmonary Disease (COPD) Using Integrated Software Suite[J]Springer. International Publishing, 2014. doi: 10.1007/978-3-319-00846-2_225.

40.

Badnjevic

Cifrek

Koruga

. Integrated software suite for diagnosis of respiratory diseases[C]//EuroconIEEE. 2013; 564-568. doi: 10.1109/EUROCON.2013.6625037.

41.

Tiwari

Tripathi

Pandey

Sharma

. Detection of COVID-19 Infection in CT and X-ray images using transfer learning approach. Technol Health Care. 2022; 30(6): 1273-1286. doi: 10.3233/THC-220114.

42.

Liu

Yin

Wei

. The human-AI scoring system: A new method for CT-based assessment of COVID-19 severity. Technol Health Care. 2022; 30(1): 1-10. doi: 10.3233/THC-213199.

43.

Zhi

Qing

. Intelligent medical image feature extraction method based on improved deep learning. Technol Health Care. 2021; 29(2): 363-379. doi: 10.3233/THC-202638.

44.

Kha

Minh

TNT

Nguyen

NQK

. Development and Validation of CT-Based Radiomics Signature for Overall Survival Prediction in Multi-organ Cancer. J Digit Imaging. 2023 Jun; 36(3): 911-922. doi: 10.1007/s10278-023-00778-0. Epub 2023 Jan 30. PMID: 36717518; PMCID: PMC10287593.

45.

Lam

LHT

Diep

DTN

Nguyet

DLN

Truong

Tri

Thanh

NQK

. Molecular subtype classification of low-grade gliomas using magnetic resonance imaging-based radiomics and machine learning. NMR Biomed. 2022 Nov; 35(11): e4792. doi: 10.1002/nbm.4792. Epub 2022 Jul 13. PMID: 35767281.

Predicting mixed venous oxygen saturation (SvO 2 ) impairment in COPD patients using clinical-CT radiomics data: A preliminary study

Abstract

BACKGROUND:

OBJECTIVE:

METHODS:

RESULTS:

CONCLUSION:

Keywords

1. Introduction

2. Materials and methods

2.1 Cases

2.2.1 Ultrasound, laboratory and pulmonary function examination

2.2.2 CT scan

2.2.3 Chest CT radiomic analysis

3. Results

Table 1 Regression analysis results of establishing the general clinical model based on clinical characteristics to predict SvO2 impairment in COPD

5. Conclusions

Ethics statement

Informed consent

Availability of data and materials

Author contributions

Funding

Footnotes

Acknowledgments

Conflict of interest

References

Table 1
Regression analysis results of establishing the general clinical model based on clinical characteristics to predict SvO₂impairment in COPD