Mining Prognosis Index of Brain Metastases Using Artificial Intelligence

Area of Science:

• Oncology
• Biostatistics
• Machine Learning

Background:

• Brain metastases represent a significant challenge in cancer care, necessitating accurate prognostic tools.
• Predicting patient survival is crucial for effective treatment planning and clinical decision-making.

Purpose of the Study:

• To identify the optimal machine learning-based prognosis index for patients with brain metastases.
• To compare the performance of various machine learning methods against traditional statistical approaches.

Main Methods:

• Utilized a dataset of 700 cancer patients with brain metastases, split into training (446) and testing (254) cohorts.
• Evaluated seven prediction methods and seven features using the mutual information and rough set with particle swarm optimization (MIRSPSO) approach.
• Assessed performance using metrics including area under the curve (AUC), accuracy, sensitivity, and specificity.

Main Results:

• The MIRSPSO method achieved the highest accuracy with an AUC of 0.978 ± 0.06.
• MIRSPSO demonstrated superior performance over traditional statistical methods and other machine learning variations (SFS, MIPSO, MISFS).
• The developed prognosis index showed improved accuracy, sensitivity, and specificity in clinical performance compared to conventional methods.

Conclusions:

• Identifying optimal machine learning methods is essential for predicting overall survival in brain metastases.
• Machine learning approaches offer significantly higher accuracy for prognosis prediction compared to conventional statistical methods.
• The MIRSPSO method presents a promising tool for enhancing clinical applications in managing brain metastases.