Improving hepatocellular carcinoma diagnosis using an ensemble classification approach based on Harris Hawks...

Area of Science:

• Medical Informatics
• Machine Learning
• Oncology

Background:

• Hepato-Cellular Carcinoma (HCC) is a prevalent and often fatal liver cancer, frequently linked to chronic liver diseases like cirrhosis.
• Early detection of HCC is crucial for effective treatment and improved patient survival rates.
• Machine learning systems show promise for enhancing HCC diagnostic accuracy.

Purpose of the Study:

• To propose a hybrid machine learning approach for accurate Hepato-Cellular Carcinoma (HCC) detection.
• To leverage the Harris Hawks Optimization (HHO) algorithm for effective feature selection in HCC classification.
• To develop an ensemble classifier using bagging and decision trees for robust HCC diagnosis.

Main Methods:

• A hybrid machine learning model integrating feature selection and ensemble classification.
• Utilized the Harris Hawks Optimization (HHO) algorithm for selecting optimal HCC-related features.
• Implemented an ensemble classifier based on the bagging technique with decision trees, incorporating missing value imputation and data normalization strategies.
• Validated the approach using the HCC dataset from Coimbra Hospital and University Center (CHUC).

Main Results:

• The proposed hybrid method demonstrated high performance in HCC detection.
• Achieved an accuracy of 97.13%, indicating superior diagnostic capability.
• Outperformed established methods such as LASSO and DTPSO in experimental comparisons.

Conclusions:

• The developed hybrid machine learning approach, utilizing HHO for feature selection and an ensemble classifier, is effective for Hepato-Cellular Carcinoma detection.
• The method offers a promising tool for improving early diagnosis and potentially patient outcomes in HCC.
• The high accuracy achieved suggests the clinical utility of this advanced computational approach in liver cancer management.