Similarity-Based Signal Embedding for Heart Sound Abnormality Detection

Öz

Heart sound (HS) analysis is important for early detection of cardiovascular abnormalities. Although deep learning methods have shown strong performance, they often require large datasets and high-dimensional inputs. In this study, a similarity-based framework is proposed for heart sound abnormality detection using low-dimensional statistical features. Experiments were conducted on the publicly available HLS-CMDS dataset, and only heart sound recordings were used. Each signal was divided into five segments, and time- and frequency-domain statistical features were extracted. Instead of direct classification, a pairwise similarity learning approach was adopted. Feature-difference vectors were constructed between segment pairs, and an RBF-kernel SVM was trained to learn similarity relationships. The learned similarity function was then used to generate a prototype-based embedding for each signal. The proposed method provides a computationally efficient and interpretable alternative to deep learning models, especially for small-scale biomedical datasets.

Anahtar Kelimeler

• Anomaly detection
• Embedding learning
• Heart sound analysis
• Low-dimensional features
• Similarity learning
• Support vector machines (SVM)

Kalp Sesi Anomali Tespiti için Benzerlik Tabanlı Temsil Yaklaşımı

Öz

Kalp sesi (KS) analizi, kardiyovasküler anormalliklerin erken tespitinde önemli bir rol oynamaktadır. Derin öğrenme yöntemleri yüksek performans göstermesine rağmen, genellikle büyük veri kümeleri ve yüksek boyutlu girdiler gerektirmektedir. Bu çalışmada, düşük boyutlu istatistiksel özellikler kullanılarak kalp sesi anormalliklerinin tespiti için benzerlik tabanlı bir çerçeve önerilmektedir. Deneyler, herkese açık HLS-CMDS veri seti üzerinde gerçekleştirilmiş ve yalnızca kalp sesi kayıtları kullanılmıştır. Her bir sinyal beş segmente ayrılmış; zaman ve frekans alanına ait istatistiksel özellikler çıkarılmıştır. Doğrudan sınıflandırma yerine, ikili benzerlik öğrenme yaklaşımı benimsenmiştir. Segment çiftleri arasında özellik farkı vektörleri oluşturulmuş ve benzerlik ilişkilerini öğrenmek üzere RBF çekirdekli bir Destek Vektör Makinesi (SVM) eğitilmiştir. Öğrenilen benzerlik fonksiyonu, her sinyal için prototip tabanlı bir temsil uzayı oluşturmak amacıyla kullanılmıştır. Önerilen yöntem, özellikle küçük ölçekli biyomedikal veri kümeleri için derin öğrenme modellerine kıyasla hesaplama açısından verimli ve yorumlanabilir bir alternatif sunmaktadır.

Anahtar Kelimeler

• Anomali tespiti
• Temsil öğrenme
• Kalp sesi analizi
• Düşük boyutlu özellikler
• Benzerlik öğrenme
• Destek vektör makineleri (SVM)

Kaynakça

1. [1] “Cardiovascular diseases (CVDs).” Accessed: Feb. 26, 2026. [Online]. Available: https://www.who.int/news-room/fact-sheets/detail/cardiovascular-diseases-(cvds)
2. [2] W. Chen, Q. Sun, X. Chen, G. Xie, H. Wu, and C. Xu, “Deep Learning Methods for Heart Sounds Classification: A Systematic Review,” Entropy, vol. 23, no. 6, p. 667, May 2021, doi: 10.3390/e23060667.
3. [3] M. Hosseinzadeh et al., “Enhanced heart sound classification using Mel frequency cepstral coefficients and comparative analysis of single vs. ensemble classifier strategies,” PLoS ONE, vol. 19, no. 12, p. e0316645, Dec. 2024, doi: 10.1371/journal.pone.0316645.
4. [4] T. N. Alotaiby, N. A. Alsahle, and G. N. Alotibi, “Abnormal Heart Sound Detection Using Common Spatial Patterns and Random Forests,” Electronics, vol. 14, no. 8, p. 1512, Apr. 2025, doi: 10.3390/electronics14081512.
5. [5] A. Doreen, W. Mwangi, and P. Muriithi, “Few-shot pneumonia detection using Siamese networks and transfer learning on chest X-ray images,” Discov Artif Intell, vol. 5, no. 1, p. 269, Oct. 2025, doi: 10.1007/s44163-025-00468-6.
6. [6] G. Koch, R. Zemel, and R. Salakhutdinov, “Siamese Neural Networks for One-shot Image Recognition,” 2015.
7. [7] R. Y. Mashhoor and A. Ayatollahi, “HeartSiam: A Domain Invariant Model for Heart Sound Classification,” in 2022 8th Iranian Conference on Signal Processing and Intelligent Systems (ICSPIS), Behshahr, Iran, Islamic Republic of: IEEE, Dec. 2022, pp. 1–5. doi: 10.1109/ICSPIS56952.2022.10044047.
8. [8] Y. Torabi, S. Shirani, and J. P. Reilly, “Descriptor: Heart and Lung Sounds Dataset Recorded From a Clinical Manikin Using Digital Stethoscope (HLS-CMDS),” May 2025.

9. [9] A. A. Sánchez-Ruiz, “Analyzing the Similarity Learned by a Siamese Network with Contrastive Loss on the MNIST Dataset,” 2025.
10. [10] K. Cortacero et al., “Evolutionary design of explainable algorithms for biomedical image segmentation,” Nat Commun, vol. 14, no. 1, p. 7112, Nov. 2023, doi: 10.1038/s41467-023-42664-x.
11. [11] A. McDonald, M. J. F. Gales, and A. Agarwal, “A recurrent neural network and parallel hidden Markov model algorithm to segment and detect heart murmurs in phonocardiograms,” PLOS Digit Health, vol. 3, no. 11, p. e0000436, Nov. 2024, doi: 10.1371/journal.pdig.0000436.
12. [12] C. Boz and Y. Kocyigit, “Segmented vs. Non-Segmented Heart Sound Classification: Impact of Feature Extraction and Machine Learning Models,” Applied Sciences, vol. 15, no. 20, p. 11047, Oct. 2025, doi: 10.3390/app152011047.
13. [13] O. Deperlioglu, “Classification of segmented heart sounds with Artificial Neural Networks,” ijamec, vol. 6, no. 4, pp. 39–44, Dec. 2018, doi: 10.18100/ijamec.2018447313.
14. [14] M. Hassanuzzaman et al., “Classification of Short Segment Pediatric Heart Sounds Based on a Transformer-Based Convolutional Neural Network,” IEEE Access, vol. 13, pp. 93852–93868, 2025, doi: 10.1109/ACCESS.2025.3573870.
15. [15] Y. He, W. Li, W. Zhang, S. Zhang, X. Pi, and H. Liu, “Research on Segmentation and Classification of Heart Sound Signals Based on Deep Learning,” Applied Sciences, vol. 11, no. 2, p. 651, Jan. 2021, doi: 10.3390/app11020651.
16. [16] C. Liu et al., “An open access database for the evaluation of heart sound algorithms,” Physiol. Meas., vol. 37, no. 12, pp. 2181–2213, Dec. 2016, doi: 10.1088/0967-3334/37/12/2181.
17. [17] H. Tang, Z. Dai, Y. Jiang, T. Li, and C. Liu, “PCG Classification Using Multidomain Features and SVM Classifier,” BioMed Research International, vol. 2018, pp. 1–14, Jul. 2018, doi: 10.1155/2018/4205027.
18. [18] A. Allen et al., “A Digital Twins Machine Learning Model for Forecasting Disease Progression in Stroke Patients,” Applied Sciences, vol. 11, no. 12, p. 5576, Jun. 2021, doi: 10.3390/app11125576.
19. [19] S. Benabderrahmane, P. Valtchev, J. Cheney, and T. Rahwan, “Refining Decision Boundaries In Anomaly Detection Using Similarity Search Within the Feature Space,” 2026, arXiv. doi: 10.48550/ARXIV.2602.02925.
20. [20] A. Yaqub, M. S. Orakzai, M. F. Qureshi, Z. Mushtaq, I. Siddique, and T. Radwan, “Spectrotemporal Deep Learning for Heart Sound Classification under Clinical Noise Conditions,” CMES, vol. 145, no. 2, pp. 2503–2533, 2025, doi: 10.32604/cmes.2025.071571.