Related Experiment Video
Updated: May 11, 2025
Author Spotlight: Advancing Alzheimer's Research – Exploring Early Detection and Multi-Omics Approaches
Published on: December 15, 2023
Optimized Feature Selection and Deep Neural Networks to Improve Heart Disease Prediction
Changming Tan1, Zhaoshun Yuan2, Feng Xu3
1Department of Cardiovascular Surgery, The Second Xiangya Hospital of Central South University, No139 Renmin Road, Changsha, Hunan Province, 410011, People's Republic of China. tanchangming79@csu.edu.cn.
Insights
This study introduces a novel deep learning model for accurate heart disease prediction using routine physical markers. The system achieves high accuracy, enabling faster and more reliable early diagnosis of cardiac conditions.
Area of Science:
- Cardiology
- Artificial Intelligence
- Biomedical Informatics
Background:
- Heart disease poses a significant global health challenge, characterized by high mortality and increasing prevalence.
- Early detection through routine physical markers is vital for effective intervention, but manual data analysis is inefficient and prone to errors.
Purpose of the Study:
- To develop a rapid and reliable system for predicting cardiac disease using a combination of deep learning and feature selection.
- To overcome the limitations of manual analysis in large-scale health datasets.
Main Methods:
- A hybrid model integrating a deep convolutional neural network (CNN) with Linear Support Vector Classification (LinearSVC) for feature selection.
- Hyperparameter tuning of the CNN using a random search algorithm to optimize performance and prevent training issues.
- Validation on the UCI and MIT public datasets.
Main Results:
- The proposed model achieved high predictive performance, with accuracy rates reaching up to 98.2% on the UCI dataset.
- Excellent performance metrics including precision, recall, and F1 score were reported.
- An average Matthews Correlation Coefficient (MCC) score of 90% was obtained, indicating robust predictive power.
Conclusions:
- The developed deep learning model demonstrates significant efficacy and reliability for predicting heart disease.
- This approach offers a promising automated solution for early cardiac disease detection, improving diagnostic speed and accuracy.
- The integration of advanced machine learning techniques can enhance the analysis of clinical data for better patient outcomes.
Abstract:
Heart disease remains a significant health threat due to its high mortality rate and increasing prevalence. Early prediction using basic physical markers from routine exams is crucial for timely diagnosis and intervention. However, manual analysis of large datasets can be labor-intensive and error-prone. Our goal is to rapidly and reliably anticipate cardiac disease using a variety of body signs. This research presents a unique model for heart disease prediction. We provide a system for predicting cardiac disease that blends the deep convolutional neural network with a feature selection technique based on the LinearSVC. This integrated feature selection method selects a subset of characteristics that are strongly linked with heart disease. We feed these features into the deep conventual neural network that we constructed. Also to improve the speed of the predictor and avoid gradient varnishing or explosion, the network's hyperparameters were tuned using the random search algorithm. The proposed method was evaluated using the UCI and MIT datasets. The predictor is evaluated using a number of indicators, such as accuracy, recall, precision, and F1 score. The results demonstrate that our model attains accuracy rates of 98.16%, 98.2%, 95.38%, and 97.84% in the UCI dataset, with an average MCC score of 90%. These results affirm the efficacy and reliability of the proposed technique to predict heart disease.
