Kernel based support vector machine for the early detection of syncope during head-up tilt test

Area of Science:

• Cardiology
• Autonomic Neuroscience
• Medical Diagnostics

Background:

• Syncope is a common clinical issue often diagnosed using the head-up tilt test (HUTT).
• Understanding autonomic nervous system (ANS) responses during HUTT is crucial for accurate syncope diagnosis.
• Dynamic properties of heart rate variability (HRV) offer insights into ANS function.

Purpose of the Study:

• To analyze ANS response during HUTT by examining HRV dynamic properties.
• To identify predictive markers for HUTT outcomes in patients with and without syncope.
• To evaluate the efficacy of machine learning models in classifying syncope during HUTT.

Main Methods:

• Collected ECG data from 66 subjects (35 with syncope, 31 without) during HUTT.
• Extracted baroreflex response, linear, and non-linear HRV parameters from RR-interval time series.
• Utilized kernel support vector machines (SVM) for patient classification.

Main Results:

• In the first 15 minutes of HUTT, increased total power spectrum, standard deviation, fractal scale of RR-interval and ΔRR-interval, and decreased sample entropy were observed in the syncope group.
• These HRV indices show potential as predictors for positive HUTT responses in reflex syncope patients.
• Kernel SVM achieved 85% classification accuracy (80.6% specificity, 88.5% sensitivity) within the first 15 minutes.

Conclusions:

• Specific HRV dynamic properties during the initial phase of HUTT can predict syncope.
• Machine learning models, particularly SVM, can aid in early and accurate syncope detection during HUTT.
• This approach may reduce examination time and avoid false-negative diagnoses, improving clinical syncope management.