Particle Swarm Optimized Gaussian Process Classifier for Treatment Discontinuation Prediction in Multicohort...

Area of Science:

• Oncology
• Machine Learning
• Biostatistics

Background:

• Prostate cancer is a significant global health concern, particularly in men.
• Metastasis remains a challenge in prostate cancer treatment.
• Docetaxel treatment, while common, can cause adverse reactions.

Purpose of the Study:

• To develop and evaluate a machine learning model for optimal discontinuation of prostate cancer treatment.
• To compare the performance of various classifiers in identifying appropriate treatment cessation points.
• To reduce adverse effects and mortality in metastatic castration-resistant prostate cancer patients.

Main Methods:

• Implementation and comparison of multiple classification models, including Particle Swarm Optimized Gaussian Process Classifier (PSO-GPC).
• Utilizing three independent cohorts of prostate cancer patient data.
• Evaluating model performance using Area Under the Receiver Operating Characteristics Curve (AUC) and Area Under the Precision-Recall Curve (AUPRC) due to data skewness and class imbalance.

Main Results:

• PSO-GPC demonstrated superior performance compared to state-of-the-art methods.
• Achieved AUCs ranging from 0.6717 to 0.8499 and AUPRCs from 0.1392 to 0.5423.
• Statistical analysis confirmed the unbiased and robust performance of PSO-GPC across independent datasets.

Conclusions:

• The proposed PSO-GPC model offers a promising approach for personalized treatment decisions in prostate cancer.
• Optimizing treatment discontinuation can mitigate adverse reactions and potentially lower mortality rates.
• This study highlights the potential of advanced machine learning in managing complex oncological treatments.