Analyzing a Lung Cancer Patient Dataset with the Focus on Predicting Survival Rate One Year after Thoracic Surgery

Area of Science:

• Biomedical Data Mining
• Surgical Oncology
• Predictive Analytics

Background:

• Data mining offers novel insights into complex biomedical datasets.
• Early lung cancer diagnosis and treatment are crucial for patient survival.
• Predictive models can improve diagnostic accuracy in medicine.

Purpose of the Study:

• To evaluate risk factor patterns for mortality one year after thoracic surgery for lung cancer.
• To develop a predictive model for post-operative mortality risk.
• To identify key variables influencing lung cancer patient survival.

Main Methods:

• Utilized data mining algorithms (Naive Bayes, Expectation-Maximization) for variable extraction.
• Employed regression analysis to develop a risk prediction questionnaire.
• Applied clustering for outlier detection and a scorecard algorithm for risk estimation.
• Analyzed a dataset of 470 records with 17 features.

Main Results:

• Large tumor size identified as the primary risk factor for increased mortality.
• Type II diabetes and preoperative dyspnea were associated with lower survival rates.
• Forced expiratory volume in the first second (FEV1) emerged as a key classification variable for patient stratification.

Conclusions:

• A questionnaire-based risk assessment tool can enhance clinical practice by identifying knowledge gaps.
• The study provides a method for predicting one-year mortality risk after lung cancer surgery.
• Findings support the integration of data mining in clinical decision-making for lung cancer patients.