Exploring the value of pleural fluid biomarkers for complementary pleural effusion disease examination

https://doi.org/10.1016/j.compbiolchem.2021.107559Get rights and content

Highlights

  • The high information weight of white blood cells and cell morphological classification was explored for the first time.

  • The model based on deep learning was built to distinguish pleural effusion disease.

  • The idea could have an efficient in terms of time-saving, easy-to-use, point-of-care and low-cost distribution. •The method can be used not only in clinics but also door-to-door screening in remote areas.

Abstract

Objective

Pleural fluid biomarkers are beneficial for the complementary diagnosis of pleural effusion etiologies. This study focuses on the multidimensional evaluation of deep learning to investigate the pleural effusion biomarkers value and the diagnostic utility of combining these markers, in distinguishing pleural effusion etiologies.

Methods

Pleural effusion were divided into three groups according to the diagnosis and treatment guidelines: malignant pleural effusion (MPE), parapneumonic effusion (PPE), and congestive heart failure (CHF). First, the value of the biomarker was analyzed by a receiver operating characteristic (ROC) curve. Then by utilizing deep learning and entropy weight method (EWM), the clinical value of biomarkers was computed multidimensionally for complementary diagnosis of pleural effusion diseases.

Results

There were significant differences in the six biomarkers, TP, ADA, CEA, CYFRA211, NSE, MNC% (p < 0.05) and no significant differences in three physical characteristics including color, transparency, specific gravity and six other biomarkers such as WBC, PNC%, MTC%, pH level, GLU, LDH (p > 0.05) among the three pleural effusion groups. The comprehensive test of pleural fluid biomarkers based on deep learning is of high accuracy. The clinical value of cytomorphology biomarkers WBC, MNC %, PNC %, MTC % was higher among pleural fluid biomarkers.

Conclusion

The clinical value of multi-dimensional analysis of biomarkers by deep learning and entropy weight method is different from the ROC curve analysis. It is suggested that during the clinical examination process, more attention should be paid to the cell morphology biomarkers, but the physical properties of the pleural fluid are less clinical significance.

Introduction

Lung cancer, the most frequent cause of cancer death, is also a major cause of the pleural effusion. Approximately 10% of lung cancer patients have pleural effusion at the time of initial diagnosis, while 30–40% develop it later in the course of their disease (Khan et al., 2013). One of the main issues in the differential diagnosis of pleural effusion is distinguishing exudates from transudates. Determining the nature of pleural effusion (exudate or transudate) allows reducing the list of potential pleural causes and indicates the direction for further diagnosis. Another important clinical issue is the etiology of effusion – malignant or benign – being crucial for pleural effusion management and prognosis. In clinical practice, the preoperative diagnosis etiology of pleural effusion still depends primarily on magnetic resonance imaging (MRI), computed tomography (CT), bronchoscopy, sputum analysis (Pongnikorn et al., 2018; Saraya et al., 2019). A biopsy followed by the histopathological examinations with high specificity is usually performed to confirm the diagnosis. In contrast, the conventional test biomarkers are less invasive and faster than histopathological examination, which has been widely used in supporting clinical tests (Porcel et al., 2004; Chen et al., 2006; Nicolini et al., 2008). Pleural fluid tests are a useful guideline when accessing the etiology of pleural effusions. The differences in pleural fluid biochemical profiles may in part mirror the relative contributions of diverse etiologies. The routine pleural fluid evaluation usually includes the character of the fluid (color, odor), specific gravity, pH level, lactate dehydrogenase (LDH), adenosine deaminase levels (ADA), glucose (GLU), total protein (TP) and cell count for differential and cytological examination.

To help differentiate the etiologies of pleural effusion, many studies have examined tumor markers, such as cytokeratin 19 fragments (CYFRA21–1), carcinoembryonic antigen (CEA), neuron-specific enolase (NSE), as a possible alternative to invasive procedures (Porcel et al., 2004, Volarić et al., Feb. 2018, Duffy et al., 2007). Tumor markers have been extensively studied, but none are specific for detecting malignant pleural effusion (MPE) (Volarić et al., Feb. 2018, Lai et al., Sep. 1999). The value of individual tumor markers depends on the dominating mechanism involved in the development of the pleural effusion. The level of serum CEA was found to be significantly increased in tumors such as colorectal cancer (Duffy et al., 2007, Mroczko et al., 2007), breast cancer (Nicolini et al., 2008, Sölétormos et al., 2004) and endometrial cancer (Yurkovetsky et al., 2007). Also, serum CYFRA 21–1 concentrations were used primarily for the detection of colorectal cancer (Sawant et al., 2008), bladder cancer (ANDREADIS et al., 2005, Nisman et al., Jun. 2002), esophageal cancer (Shimada et al., 2005), and gynecological cancer (Pras et al., 2002, Yamamoto et al., 2005). On another hand, the median fluid concentrations for the investigated tumor markers were higher in malignant effusions than benign effusions (Korczynski et al., 2009). For example, the serum and pleural fluid levels of CYFRA21–1 are considered as a useful means in differentiating malignant from benign pleural effusion (Pras et al., 2002, Yamamoto et al., 2005). Besides, when CYFRA21–1 and CEA are combined, the diagnostic sensitivity for malignant pleural effusion would be superior, unlikely to CEA or CYFRA21–1 used alone (Saraya et al., 2019). These tumor markers exhibit great specificity, and the low sensitivity of each marker limits their diagnostic value. Some studies suggested that the analysis of tumor markers must be performed in combination with other pleural fluid tests for greater diagnostic accuracy (Chen et al., 2006, Trapé et al., 2017). The increasing number of tumor markers can improve detection efficiency, as well as reducing the specificity accordingly. Thus, monitoring of the type and number of tumor markers in pleural effusions can help to distinguish benign from malignant pleural effusions (Korczynski et al., 2009).

It is well known that there are many reports on the clinical application value of single or multiple combined biomarkers (Otoshi et al., 2017). However, whether it is based on single or multiple joint detections, it cannot simultaneously improve the specificity and sensitivity of the tested items at the same time. Recently, the combination of multiple biomarkers in this research group has improved the clinical diagnostic value of markers and basic theory for multidimensional analysis of biomarkers was provided (Zhang et al., 2018 Oct, Wu et al., 2019).

Deep learning is a multidisciplinary field combining computer science and mathematics and focused on implementing computer algorithms capable of maximizing predictive accuracy from statistic or dynamic data sources using analytical or probabilistic models. In the present study, the entropy weight method (EMW) which is a branch of information theory, that can capture the implied interactions among markers and determine the information weight of each marker, was used for deep learning. Combining clinical data using deep learning can enable the development of a novel model for distinguishing the etiology of pleural effusion. Furthermore, we investigate the contribution and direction of each variable within the model. The large number of biomarkers and their intertwining relations necessitates advanced deep learning models, rather than simple statistical and correlation analysis. The multidimensional combination of many biomarkers based on deep learning may assist effectively in the pleural effusion diagnosis. Moreover, the combination of biomarkers and deep learning led to enhancement in accuracy over single marker (Saleh et al., 2016).

This paper aims to investigate the discriminative properties of pleural fluid and serum biomarkers in distinguishing malignant and non-malignant pleural effusion. We also asked for the optimal combination of these markers to improve the diagnostic certainty in clinical use. The multidimensional combination of tumor markers, biochemical markers, and cytomorphological markers was explored by deep learning. It helps support the diagnosis etiology of pleural effusion, including malignant pleural effusion (MPE), parapneumonic effusion (PPE) and pleural effusion caused by congestive heart failure (CHF).

Section snippets

Patients and samples

A total of 276 pleural fluid samples were collected prospectively from 276 consecutive patients diagnosed with pleural effusion and admitted to the Clinical Laboratory at Tianjin Chest Hospital, China, from Jun 2018 until Jun 2019. Measurements were made at the time of admission to the Tianjin Chest Hospital. 193 patients had malignant pleural effusion (MPE) related lung cancer diagnosed by histologic and immunohistological analyses, 47 patients had parapneumonic effusion (PPE) resulting from a

Biomarkers for the determination of pleural effusion in three groups of diseases

According to the current guideline for medical examination of pleural effusion in China, the routine pleural fluid evaluation usually includes determination of physical character, biochemical markers, immune markers, pleural effusion cell counting, and classified counting according to cytomorphology after staining. The principle examinations and the approximate time consumption are listed (Table 1).

Comparison of biomarkers in three groups of pleural effusion

For each group, 11 indicators including ADA, TP, GLU, LDH, CEA, CYFRA21–1, NSE, WBC, MNC %, PNC

Discussion

According to the guideline of China Medical Examination of Pleural Effusion, the routine pleural fluid evaluation usually includes determination of the physical features of pleural fluid such as specific gravity, color, transparency; biochemical markers including pH level, TP, LDH, GLU, ADA, tumor markers (CEA, CYFRA21–1, NSE); WBC and the percentage of cells of various types (MNC%, PNC%, MTC%). Determining the diagnostic utility of pleural effusions markers has many advantages regarding both

Conclusions

Our study applied deep learning and entropy weight method to appraise the clinical value of pleural fluid biomarkers for the first time. As a result, a comprehensive examination of pleural fluid biomarkers is of high diagnosis ability and also suggests that we should attach special importance to cytomorphology tests in the pleural effusion routine examinations. In addition, the idea of combining pleural fluid markers and deep learning techniques could lead to an efficient future differential

Funding

This work was supported by the National Natural Science Foundation of China (Grant Nos. 61703304, 31600761).

CRediT authorship contribution statement

Pham Thi Huyen: Designed the study, Interpreted data, Wrote the manuscript. Lei Li: Made the clinical diagnosis. Meiyu Li, Zhexiang Wang: Collected samples and data. Sike Ma, Yan Zhao, Jing Yan: Performed data analyses. Meng Zhao, Xuguo Sun: Corrected the manuscript and contributed to the study design. All authors agreed with the results and conclusions.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

We are grateful to the patients, physicians, and pathologists at the Clinical Laboratory, Tianjin Chest Hospital, who contributed to patient material.

References (37)

  • W. Wang et al.

    Evaluation of information transfer and data transfer models of rain-gauge network design based on information entropy

    Environ. Res.

    (2019)
  • Y. Yamamoto et al.

    Cleavage of carcinoembryonic antigen induces metastatic potential in colorectal carcinoma

    Biochem. Biophys. Res. Commun.

    (2005)
  • Z. Yurkovetsky et al.

    Development of multimarker panel for early detection of endometrial cancer. High diagnostic power of prolactin

    Gynecol. Oncol.

    (2007)
  • M. Bradshaw et al.

    The role of vascular endothelial growth factor in the pathogenesis, diagnosis and treatment of malignant pleural effusion

    Curr. Oncol. Rep.

    (2013)
  • R. Braunschweig et al.

    Detection of malignant effusions: comparison of a telomerase assay and cytologic examination

    Diagn. Cytopathol.

    (2001)
  • T. Eason et al.

    An information theory-based approach to assessing spatial patterns in complex systems

    Entropy

    (2019)
  • E. Fafliora et al.

    Systematic review and meta-analysis of vascular endothelial growth factor as a biomarker for malignant pleural effusions

    Physiol. Rep.

    (2016)
  • D. Jimenez-Carretero et al.

    Tox_(R)CNN: deep learning-based nuclei profiling tool for drug toxicity screening

    PLoS Comput. Biol.

    (2018)
  • Cited by (2)

    1

    Represents the first author.

    View full text