Cyfra 21.1 CEA for both serum and pleural fluid of 168 patients

Citation Author(s):
Gianni
D'Angelo
Department of Computer Science, University of Salerno, Italy
Submitted by:
Gianni D'Angelo
Last updated:
Fri, 04/19/2019 - 13:59
DOI:
10.21227/wk3e-5x64
Data Format:
License:
0
0 ratings - Please login to submit your rating.

Abstract 

Malignant pleural effusions (MPEs) are a challenging public health problem, causing significant morbidity and often being the first presenting sign of cancer. Pleural fluid cytology is the most common method used to differentiate malignant from non-malignant effusions. However, its sensitivity reaches 50-70% and depends on the experience of the cytologist, the tumor load, and the amount of fluid tested. Therefore, diagnostic inaccuracy and a high incidence of false negatives may endanger patients with clinical mistreatment and mismanagement. Analysis of tumor markers in pleural effusions has been considered so far as a valid and less invasive alternative for identifying MPEs, and a reliable diagnostic aid in advanced stages when cytology examination is inconclusive. Although thoracoscopic and video assisted pleural biopsy have high specificity as minimally invasive procedures, they are not well tolerated by some patients and may not be available in all hospitals. Therefore, novel non-invasive methods are required that could be widely available, and would be able to differentiate MPEs from benign effusions. This work describes the use of Genetic Programming to determine the optimal panel of markers able to fix a highly sensitive differentiation between benign and MPEs. A formula is also provided mining the complex relation existing between the expressed markers and the presence of lung cancer. The proposed formula indicates CEA and CYFRA21-1 as high efficient tumor markers in the diagnosis of cytologically negative lung cancer associated MPE, and emphasizes the complex relationships between these markers and the lung cancer. The experimental results put into evidence the effectiveness of the proposed approach in predict the typology of the lung cancer (benign or malignant) given biomarkers.

Instructions: 

The file DataSet.pdf contains the data samples used in this study. It reports the values of Cyfra 21.1 and CEA for both serum and pleural fluid of 168 patients. The dataset includes 124 patients pathologically diagnosed with lung cancer (LC), and 44 patients suffering from non-malignant diseases (INF).