Cardiovascular disease detection from high utility rare rule mining

Area of Science:

• Cardiology
• Data Mining
• Medical Informatics

Background:

• Cardiovascular diseases (CVDs) pose a significant global health challenge.
• Early detection of rare symptoms is crucial for timely intervention and improved patient outcomes.
• Existing methods may overlook subtle or rare indicators of cardiovascular conditions.

Purpose of the Study:

• To develop a data-driven method for discovering rare cardiovascular disease symptom rules.
• To enhance early detection of cardiovascular disease by identifying high-utility rare patterns.
• To provide decision support for medical experts and early alerts for patients.

Main Methods:

• Utilizing fuzzy sets to handle uncertainty in continuous health data like age.
• Defining hazard ratio utility for itemsets to guide the mining process.
• Employing High Utility Rare Itemset Mining (HURIM) to discover significant symptom patterns.
• Integrating a prediction step for diagnosing cardiovascular disease from new health records.

Main Results:

• Successfully identified rare cardiovascular disease symptom patterns from historical health examination records.
• Demonstrated the effectiveness of hazard ratio utility in pinpointing clinically relevant rare symptoms.
• The proposed method accurately detected cardiovascular disease in new datasets based on identified rare symptoms.

Conclusions:

• The developed method effectively extracts high-utility rare symptoms for cardiovascular disease detection.
• Rare symptoms, when identified through utility risk, offer valuable insights for medical decision-making.
• This approach supports early diagnosis and treatment strategies for cardiovascular conditions.