Optimizing early diagnosis by integrating multiple classifiers for predicting brain stroke and critical diseases

Area of Science:

• Medical Informatics
• Computational Biology
• Artificial Intelligence in Medicine

Background:

• Machine learning (ML) is increasingly vital for medical prognosis.
• Early disease prediction, such as for brain stroke, is crucial for patient outcomes.
• Brain stroke is a leading cause of death, necessitating advanced diagnostic tools.

Purpose of the Study:

• To develop robust ML models for the early prediction of brain stroke, Alzheimer's, heart attack, cancer, and Parkinson's.
• To identify a superior ML technique for early brain stroke detection, aiming to reduce severe post-complication incidence.
• To evaluate and compare the performance of various ML classifiers and ensemble models across multiple neurological and cardiovascular diseases.

Main Methods:

• Trained ML models using five distinct datasets for disease prediction.
• Employed eight individual classifiers and 56 ensemble models (soft and hard voting) for brain stroke prediction.
• Utilized eight individual classifiers for early prediction of heart attack, cancer, Alzheimer's, and Parkinson's.

Main Results:

• A hybrid model combining Random Forest and Decision Tree with hard voting achieved 99% accuracy for early brain stroke prediction.
• The proposed brain stroke model demonstrated 98% precision, 100% recall, and 99% F1 score.
• XGBoost excelled in predicting cancer, Parkinson's, and Alzheimer's, while Bernoulli Naive Bayes was optimal for heart attack prediction.

Conclusions:

• The developed ML models, particularly the Random Forest and Decision Tree ensemble, show high efficacy for early disease prediction.
• The study highlights the potential of ML in significantly improving diagnostic accuracy for critical neurological and cardiovascular conditions.
• The findings suggest that ensemble methods can outperform individual classifiers in complex medical prediction tasks.