Artificial Intelligence and Machine Learning Techniques in the Early Detection of Cardiovascular Diseases: Predicting Heart Conditions before Clinical Onset”

Abstract

Background:

The first cause of death worldwide is cardiovascular diseases (CVDs). Detection is imperative to better outcomes, although conventional methods of diagnosis can only occur after the disease symptoms appear. AI and ML also present new opportunities, as they are used to analyze big, complicated data to identify early, sometimes hardly detectable signs of heart disease before clinical manifestation.

Objectives: To assess the effectiveness of an AI and machine learning model in predicting cardiovascular events before the onset of clinical symptoms, using data from electronic health records, medical imaging, and wearable devices.

Study design: A Cross-sectional study.

Place and duration of study:  Department of Cardiology, MTI LRH, Peshawar, From January 2019 to December 2020.

Methods: AI model was trained on electronic health records, echocardiogram data, and wearable sensor results to analyze a cross-sectional study of 150 patients. The algorithm used is the gradient boosting model verified through 10-fold cross-validation. Precipitating factors were heart rate, cholesterol, ECG patterns, and blood pressure. The sensitivity, specificity, and the area under the ROC curve (AUC) were used to determine predictive performance. Data were analyzed with SPSS v26 at a p-value<0.05.

Results:

Co-demographic characteristics of 150 patients (52 percent male) took part in the study. The average age was 57.6 years (SD +11.2). The AI model showed a sensitivity of 91 percent, specificity of 88 percent, and an area under the curve of 0.93 in major adverse cardiovascular event prediction in one year. The AI method outperformed the conventional risk scoring (AUC 0.74) with a significant improvement (p < 0.001). The risk of the event was 4.2 times higher among patients identified as high risk by the model compared with low-risk subjects. The model also identified subclinical patterns that were related to myocardial stress and early fibrillation of the atria that were not covered by common diagnostics.

Conclusion: Analysis based on AI meaningfully promotes the identification of cardiovascular diseases before it could be driven by identifying a patient at a higher risk, before symptoms. This predictive ability also enables clinicians to take proactive measures, which might lower the number and severity of cardiac events. The routine use of AI tools would change cardiovascular care into preventive care. Future research studies are encouraged to confirm the clinical benefit of AI across different patient groups.