Why should you model time when you use Markov models for heart sound analysis

Area of Science:

• Cardiology
• Biomedical Engineering
• Signal Processing

Background:

• Auscultation is crucial for diagnosing heart diseases but faces interpretation challenges due to rapid temporal events and frequencies outside the human audible range.
• Existing methods struggle to accurately segment heart sounds, impacting diagnostic precision.

Purpose of the Study:

• To propose and evaluate a novel semi-hidden Markov model (SHMM) for enhanced heart sound segmentation.
• To address the limitations of traditional hidden Markov models (HMMs) by incorporating temporal constraints of cardiac cycles.

Main Methods:

• Development of a semi-hidden Markov model (SHMM) tailored for heart sound analysis.
• Comparison of the SHMM's performance against state-of-the-art hidden Markov models (HMMs).
• Experimental validation focusing on the accuracy of reconstructing continuous state sequences.

Main Results:

• The proposed SHMM demonstrated superior accuracy in recreating the true continuous state sequence of heart sounds compared to HMMs.
• A mean error rate per sample of 0.23 was achieved, indicating significant improvement in segmentation accuracy.
• The SHMM effectively accounts for the temporal constraints inherent in heart cycles.

Conclusions:

• Semi-hidden Markov models offer a more accurate approach to segmenting heart sounds than traditional HMMs.
• This advancement has the potential to improve the interpretation of heart sounds and aid in the diagnosis of cardiac conditions.
• The model's ability to handle temporal constraints enhances its clinical applicability in cardiology.