Research on Lung Cancer Classification Based on Multidimensional Hematological Indicators and Machine Learning...

Area of Science:

• Oncology
• Bioinformatics
• Machine Learning

Background:

• Accurate lung cancer subtyping is essential for personalized medicine and improved patient outcomes.
• Current methods like pathological biopsy can be invasive and lack dynamic monitoring capabilities.

Purpose of the Study:

• To develop and validate non-invasive machine learning models for lung cancer subtyping using hematological indicators.
• To compare the clinical applicability of these models against traditional diagnostic methods.

Main Methods:

• Utilized data from 771 lung cancer patients for model development and validation.
• Screened ten supervised learning algorithms, including XGBoost and Random Forest.
• Validated models on an independent cohort of 510 lung cancer cases from two clinical centers.

Main Results:

• The XGBoost model achieved 95% accuracy in differentiating small cell lung cancer from non-small cell lung cancer.
• The Random Forest model achieved 91% accuracy in distinguishing lung squamous cell carcinoma from lung adenocarcinoma.
• Both models demonstrated significant clinical applicability in independent validation.

Conclusions:

• Machine learning models integrating hematological data offer a non-invasive, repeatable, and dynamic approach to lung cancer subtyping.
• These models serve as a valuable complement to pathological biopsy, enhancing diagnostic accuracy and facilitating personalized treatment strategies.