On the development of diagnostic support algorithms based on CPET biosignals data via machine learning and wavelets

Area of Science:

• Medical Informatics
• Biomedical Engineering
• Machine Learning in Healthcare

Background:

• Early disease identification is crucial for preventing complications and reducing healthcare system burden.
• Artificial intelligence (AI) offers significant support for medical diagnosis.
• Cardiopulmonary exercise testing (CPET) generates biosignals valuable for health assessments.

Purpose of the Study:

• To develop and evaluate an AI-driven algorithm for identifying metabolic syndrome (MS), heart failure (HF), and healthy individuals (H) using CPET data.
• To enhance disease diagnosis through advanced signal processing and machine learning techniques.

Main Methods:

• Utilized support vector machine (SVM) classification combined with wavelet transforms for feature extraction from CPET biosignals.
• Trained and tested models on data from 45 participants (15 MS, 15 HF, 15 healthy).
• Developed both binary (SVM-POL-BW5) and multi-class (SVM-LIN-MW3) classification algorithms.

Main Results:

• The SVM with a polynomial kernel and 5-level wavelet transform (SVM-POL-BW5) demonstrated superior performance in binary classification tasks compared to existing methods.
• The multi-class classification algorithm (SVM-LIN-MW3) achieved an average accuracy of 95% for diagnosing MS, HF, and healthy states.
• The study highlights the effectiveness of SVM and wavelet transforms in analyzing complex biosignals.

Conclusions:

• SVM-based algorithms coupled with wavelet transforms show significant promise for diagnosing various diseases using CPET data.
• The developed AI approach demonstrates adaptability and potential for broader applications in clinical healthcare settings.
• This methodology offers a novel pathway for early and accurate disease detection, improving patient outcomes.