A practical approach for colorectal cancer diagnosis based on machine learning

Area of Science:

• Medical Informatics
• Computational Biology
• Oncology

Background:

• Colorectal cancer diagnosis relies on timely and accurate identification.
• Electronic Medical Records (EMRs) contain valuable data for diagnostic tools.
• Machine learning offers potential for improving diagnostic accuracy and efficiency.

Purpose of the Study:

• To develop and evaluate machine learning models for colorectal cancer diagnosis.
• To identify key features supporting early colorectal cancer detection.
• To create a system assisting clinicians in patient management.

Main Methods:

• Data collection from Electronic Medical Records (EMRs).
• Feature engineering and selection, identifying 21 relevant attributes.
• Training and evaluation of machine learning models: CART, Random Forest, and XGBOOST.

Main Results:

• XGBOOST demonstrated superior performance among the evaluated models for colorectal cancer diagnosis.
• Identification of 21 significant feature attributes for early diagnosis.
• The developed system aids clinicians in selecting appropriate tests and procedures.

Conclusions:

• Machine learning, particularly XGBOOST, is effective for colorectal cancer diagnosis.
• The system has the potential to reduce patient waiting times and healthcare costs.
• This approach provides a framework for applying machine learning in medical diagnostics.