Application of gene expression programming and sensitivity analyses in analyzing effective parameters in gastric cancer tumor size and location

• Published in: Soft computing (Berlin, Germany) Vol. 24; no. 13; pp. 9943 - 9964
• Main Authors: Dorosti, Shadi, Jafarzadeh Ghoushchi, Saeid, Sobhrakhshankhah, Elham, Ahmadi, Mohsen, Sharifi, Abbas
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.07.2020; Springer Nature B.V
• Subjects: Age; Artificial Intelligence; Cancer; Computational Intelligence; Control; Documents; Engineering; Gastric cancer; Gene expression; Health care facilities; Health services; Infections; Mathematical Logic and Foundations; Mathematical models; Mechatronics; Medical prognosis; Medical screening; Methodologies and Application; Methods; Mortality; Parameter identification; Parameter sensitivity; Patients; Risk factors; Robotics; Sensitivity analysis; Sex; Smoking; Terminal diseases; Tumors; Morris; Sensitivity analysis; Sobol’–Jansen; Distributed Evaluation of Local Sensitivity Analysis; Gastric cancer; Gene expression programming
• ISSN: 1432-7643; 1433-7479
• DOI: 10.1007/s00500-019-04507-0

Gastric cancer (GC) is the third reason for cancer-related deaths in the world. The late referral of patients to medical centers in an advanced stage can make the treatment procedure more difficult. Accurate diagnosis of risk factors in GC tumor size and tumor location can lead to taking preventive measures or determining a suitable treatment strategy. This study aims to present a general model to identify the correlation of different parameters in a GC tumor place and tumor size. The medical documents of GC patients consist of the dataset of this study. The effect of seven main parameters, namely age, smoking, Helicobacter pylori ( H. pylori ) infection, job, surgical background, sex, and nodal stage is investigated in GC tumor location and tumor size. By considering all the medical documents, data modeling is conducted using gene expression programming because of the high precision of model output. In the following, three different sensitivity analysis methods (Morris, Distributed Evaluation of Local Sensitivity Analysis (DELSA), and Sobol’–Jansen) are applied to determine the influential factors in the tumor size and location. Results show that in sequence, sex, age, and H. pylori records mostly affect tumor location; the nodal stage, smoking, and surgery record mostly affect tumor size. This method can help in identifying effective parameters and prevention of patients’ death in all types of diseases, even for terminal illnesses.