Açıklanabilir Yapay Zeka Teknikleri ile Fotopletismografi (PPG) Sinyalleri Kullanarak Kan Glikoz Seviyesi Tahmini

Year 2024, Volume: 9 Issue: 1, 45 - 62, 29.06.2024

Gökhan Adigüzel , Ümit Şentürk , Kemal Polat

https://doi.org/10.56171/ojn.1473276

Abstract

Kan şekeri seviyelerinin tahmini, diyabetin etkili yönetiminde kritik bir görevdir. Bu çalışma, Fotopletismografi (PPG) sinyallerini kullanarak kan şekeri seviyelerini tahmin etmek için CatBoost, XGBoost ve ekstra ağaç regresör gibi makine öğrenimi modellerinin gücünden, SHAP değerleri ve karışıklık matrisi gibi açıklanabilir yapay zeka teknikleriyle birlikte yararlanmaya odaklanıyor. Bu araştırmada kullanılan veri seti, PPG sinyallerinden glikoz tahmini için dikkatlice seçilmiştir ve 217 kişiden alınan verilerden oluşmaktadır. Her bireyin bilgileri, laboratuvar glikoz ölçümlerini ve yaklaşık bir dakikalık kaydedilen parmak PPG sinyallerini içerir. Test edilen çeşitli makine öğrenimi modelleri arasında CatBoost, kan şekeri seviyelerini tahmin etmede en iyi performansı gösteren model olarak ortaya çıktı. CatBoost modeli, 0.7191'lik etkileyici bir determinasyon katsayısı (R2) metriğine ve 25.21'lik ortalama mutlak hataya (MAE) ulaşarak glikoz seviyesi tahminlerindeki verimliliğini ve doğruluğunu ortaya koydu. Özellik önemi analizi, CatBoost ile oluşturulan tahmin modelinde medyan fark ve basıklık gibi belirli özelliklerin önemini vurgulayarak bunların kan şekeri seviyelerinin belirlenmesindeki önemli rolünün altını çizdi. Açıklanabilir yapay zeka tekniklerinin dahil edilmesi, tahmine dayalı modellerin yorumlanabilirliğini ve şeffaflığını arttırdı. Sonuç olarak bu araştırma, PPG sinyallerinden kan şekeri seviyelerinin tahmin edilmesinde makine öğrenimine dayalı yaklaşımların potansiyelini vurgulamaktadır. CatBoost gibi gelişmiş modellerden yararlanan ve açıklanabilir yapay zeka yöntemlerini kullanan bu çalışma, doğru, invaziv olmayan ve veriye dayalı tahmine dayalı metodolojiler yoluyla gelişmiş diyabet yönetiminin yolunu açıyor.

Keywords

Kan Şekeri Tahmini, Fotopletismografi, Makine Öğrenmesi, SHAP, Açıklanabilir Yapay Zeka

Blood Glucose Level Estimation Using Photoplethysmography (PPG) Signals with Explainable Artificial Intelligence Techniques

Abstract

Estimating blood sugar levels is a critical task in effective diabetes management. This study focuses on leveraging the power of machine learning models such as CatBoost, XGBoost, and Extra Trees Regressor, along with explainable AI techniques like SHAP values and confusion matrices, to predict blood sugar levels using Photoplethysmography (PPG) signals. The dataset used in this research is carefully selected for glucose prediction from PPG signals and consists of data from 217 individuals. Information for each individual includes laboratory glucose measurements and approximately one minute of recorded finger PPG signals. Among the various machine learning models tested, CatBoost emerged as the best-performing model in predicting blood sugar levels. The CatBoost model demonstrated its efficiency and accuracy in glucose level predictions by achieving an impressive coefficient of determination (R2) of 0.7191 and a mean absolute error (MAE) of 25.21. Feature importance analysis highlighted the significance of specific features like median deviation and kurtosis in the predictive model built with CatBoost, emphasizing their critical role in determining blood sugar levels. The inclusion of explainable AI techniques enhanced the interpretability and transparency of predictive models. In conclusion, this research underscores the potential of machine learning-based approaches in predicting blood sugar levels from PPG signals. By leveraging advanced models like CatBoost and utilizing explainable AI methods, this study paves the way for improved diabetes management through accurate, non-invasive, and data-driven predictive methodologies.

Keywords

Blood Sugar Prediction, Photoplethysmography, Machine Learning, SHAP, XAI

References

• International Diabetes Federation. (IDF 2009c). Diabetes Prevalence. Online Access http://www.idf.org/home/index.cfm?node=264 (Acces:12.05.2009)
• Diyabet Çığ Gibi Büyüyor. (2008). Diyabete Bakış, 7, 6-7.
• K. Wikblad, L. Wibell, , K. Montin (1990). “The Patient's Experience Of Diabetes And Its Treatment: Construction Of An Attitude Scale By A Semantic Differential Technique.” Journal Of Advanced Nursing, 15(9), 1083-1091.
• American Diabetes Association. Diagnosis and classification of diabetes mellitus, Diabetes Care, 32 (1), 62–67, 2009.
• O. Sevli. Diyabet hastalığının farklı sınıflandırıcılar kullanılarak teşhisi, Journal of the Faculty of Engineering and Architecture of Gazi University, 38:2 (2023) 989-1001
• American Diabetes Association. (2023). Standards of Medical Care in Diabetes—2023. Diabetes Care, 46(Suppl. 1): S1-S291.
• Lam, B. Q., Srivastava, R., Morvant, J., Shankar, S., Srivastava, R. K. 2021. Association of DM and Alcohol Abuse with Cancer: Molecular Mechanisms and Clinical Significance. Cells 2021, Vol. 10, 10(11):, 3077. https://doi.org/10.3390/CELLS10113077
• Quinones, S., Robert Roberts, C. C., David Cistola, C.-C., Narayan, M., Crites, S. L. 2021. Non-invasive in-vitro glucose monitoring using an optical sensor and machine searning techniques for diabetes applications, Texas.
• Beck, R. W., Riddlesworth, T., Ruedy, K., & Kollman, C. (2018). Continuous glucose monitoring versus usual care in patients with type 2 diabetes receiving multiple daily insulin injections: a randomized trial. Annals of Internal Medicine, 169(6), 379-387.
• Karon, B. S. (2016). Why is there still no international standard for glucose in blood? Clinical Chemistry and Laboratory Medicine (CCLM), 54(6), 975-977.
• Newman, J. D., & Turner, A. P. (2005). Home blood glucose biosensors: a commercial perspective. Biosensors and Bioelectronics, 20(12), 2435-2453.
• Sieg, A., Guy, R. H., & Delgado-Charro, M. B. (2012). Non-invasive and minimally invasive methods for measuring glucose. In Advances in Noninvasive Electrocardiographic Monitoring Techniques (pp. 259-277). Springer, Berlin, Heidelberg.
• Lee, Y. H., Wong, D. T., & Saliva: An emerging biofluid for early detection of diseases. (2009). American Journal of Dentistry, 22(4), 241-248.
• Rohleder, D., von Ribbeck, H. G., Brenner, T., Giessen, H., & Lemmer, U. (2012). Noninvasive continuous glucose monitoring by photoacoustic spectroscopy. Analytical Chemistry, 84(16), 6558-6565.
• Lukaski, H. C. (1987). Methods for the assessment of human body composition: traditional and new. American Journal of Clinical Nutrition, 46(4), 537-556.
• Gusev, M., Poposka, L., Spasevski, G., Kostoska, M., Koteska, B., Simjanoska, M., Ackovska, N., Stojmenski, A., Tasic, J., Trontelj, J. 2020. Noninvasive Glucose Measurement Using Machine Learning and Neural Network Methods and Correlation with Heart Rate Variability. Journal of Sensors. https://doi.org/10.1155/2020/9628281
• Kossowski, T., Stasinski, R. 2016. Robust IR attenuation measurement for non-invasive glucose level analysis. International Conference on Systems, Signals, and Image 75 Processing : International Conference on Systems, Signals, and Image Processing (Vol. 2016-June). https://doi.org/10.1109/IWSSIP.2016.7502770
• Sharma, N. K., Singh, S. 2012. Designing a non invasive blood glucose measurement sensor. 2012 IEEE 7th International Conference on Industrial and Information Systems, ICIIS 2012 : 2012 IEEE 7th International Conference on Industrial and Information Systems, ICIIS 2012. https://doi.org/10.1109/ICIInfS.2012.6304818
• Zhao, X., Zheng, Q., Yang, Z. M. 2016. Two types of photonic crystals applied to glucose sensor. . https://doi.org/10.1109/inec.2016.7589369
• Tanaka, Y., Purtill, C., Tajima, T., Seyama, M., Koizumi, H. 2017. Sensitivity improvement on CW dual-wavelength photoacoustic spectroscopy using acoustic resonant mode for 79 noninvasive glucose monitor. Proceedings of IEEE Sensors : Proceedings of IEEE Sensors. https://doi.org/10.1109/ICSENS.2016.7808685
• Arikan , K., Burhan, H., Sahin , E., Sen, F., A sensitive, fast, selective, and reusable enzyme-free glucose sensor based on monodisperse AuNi alloy nanoparticles on activated carbon support, Chemosphere, 291, 3 2022 , https://doi.org/10.1016/j.chemosphere.2021.132718
• Karimi, F., Zare, N., Bekmezci, M., Akin, M., Bayat, R., Seyitoğlu, B., Arikan, K., Isik, I., Sen, F., Enzyme-free glucose detection via scalable and economical fabrication of nickel-polyvinylpyrrolidone-modified multi-walled carbon nanotubes, Elektrochimica Acta , 496 , 8 2024 https://doi.org/10.1016/j.electacta.2024.144341
• Allen, J. (2007). Photoplethysmography and its application in clinical physiological measurement. Physiological Measurement, 28(3), R1-R39
• C. Salamea, E. Narvaez and M. Montalvo, “Database Proposal for Correlation of Glucose and Photoplethysmography Signals”, Advances in Intelligent Systems and Computing, pp. 44-53, 2019. , Available: 10.1007/978-3-030-32033-1_5.
• Online: “Utilizing the PPG/BVP signal”, support.empatica.com, 2021, Available: https://support.empatica.com/hc/en-us/articles/204954639-Utilizing-the-PPGBVP-signal
• D. Makowski, T. Pham, ZJ. Lau, JC. Brammer JC, F. Lespinasse, H . Pham, C. Scholzel, SHA. Chen. NeuroKit2 A Python toolbox for neurophysiological signal processing. Behavior 2021
• M. Elgendi, 2012. On the Analysis of Fingertip Photoplethysmogram Signals, Current Cardiology Reviews, 8, 14-25, Available: 10.2174/157340312801215782
• M. Barandas, D. Folgado, L. Fernandes, S. Santos, M. Abreu, P. Bota, H. Liu, T. Schultz, H. Gamboa. TSFEL: Time Series Feature Extraction Library, SoftwareX 11 (2020). https://doi.org/10.1016/j.softx.2020.100456.
• Fraunhofer Portugal AICOS. (n.d.). TSFEL: Time Series Feature Extraction Library. GitHub Repository. Retrieved from https://github.com/fraunhoferportugal/tsfel
• DeCarlo, L. T. (1997). On the meaning and use of kurtosis and skewness. Psychological Methods, 2(3), 292-307.
• R. M. Kimmel, S. I. Rubin. (1997). Crest Factor. CRC Press.
• Ali, M. (2020). PyCaret: An open-source, low-code machine learning library in Python. Available at: https://pycaret.org.
• T. Nguyen, N. Tran, B. M.Nguyen, G. Nguyen, 2018. A resource usage prediction system using functional-link and genetic algorithm neural network for multivariate cloud metrics. In 2018 IEEE 11th conference on service-oriented computing and applications (SOCA), 49–56. IEEE,
• James, G., Witten, D., Hastie, T., & Tibshirani, R. (2013). An Introduction to Statistical Learning: with Applications in R. Springer Science & Business Media.
• Root Mean Square Value A Dictionary of Physics (6 ed.). Oxford University Press. 2009. ISBN 9780199233991
• Kutner, M. H., Nachtsheim, C. J., Neter, J., & Li, W. (2004). Applied Linear Statistical Models. McGraw-Hill/Irwin.
• Chen, T., & Guestrin, C. (2016). XGBoost: A Scalable Tree Boosting System. Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 785–794.
• Hyndman, R. J., & Koehler, A. B. (2006). Another look at measures of forecast accuracy. International Journal of Forecasting, 22(4), 679-688.
• Anna Veronika Dorogush, Andrey Gulin, Gleb Gusev, Nikita Kazeev, Liudmila Ostroumova Prokhorenkova, Aleksandr Vorobev "Fighting biases with dynamic boosting". arXiv:1706.09516, 2017.
• Qi Meng, Guolin Ke, Taifeng Wang, Wei Chen, Qiwei Ye, Zhi-Ming Ma, Tie-Yan Liu. "A Communication-Efficient Parallel Algorithm for Decision Tree". Advances in Neural Information Processing Systems 29 (NIPS 2016), pp. 1279-1287.
• Moore, D. S., McCabe, G. P., & Craig, B. A. (2019). Introduction to the Practice of Statistics. W. H. Freeman.
• G. H Huzooree, K.K. Khedo, N. Joons, 2017. Glucose prediction data analytics for diabetic patients monitoring, 2017 1st International Conference on Next Generation Computing Applications (NextComp) 10.1109/NEXTCOMP.2017.8016197
• G. Edinson, G. Go´mez , L. E. B. Agredo, V. Martı´nez, O. F. B. Leiva. Application of Artificial Intelligence Techniques for the Estimation of Basal Insulin in Patients with Type I Diabetes, Research Article Open Access https://doi.org/10.1155/2020/7326073
• N.D Nanayakkara, S.C Munasingha, G.P. Ruwanpathirana, 2018. Non-Invasive Blood Glucose Monitoring using a Hybrid Technique, In Proceedings of the 2018 Moratuwa Engineering Research Conference (MERCon), Moratuwa, Sri Lanka, pp. 7–12. , 10.1109/MERCon.2018.8421885
• R. Sun, Y.L. Duan, Y.M. Zhang, L.G. Feng, B. Ding, R.N. Yan, J.H. Ma, X.F Su. Time in Range Estimation in Patients with Type 2 Diabetes is Improved by Incorporating Fasting and Postprandial Glucose Levels, DIABETES THERAPY, Volume14Issue8Page1373-1386, DOI10.1007/s13300-023-01432-2
• S. S. Gupta, S. Hossain, C. A. Haque and K. -D. Kim, 2020. In-Vivo Estimation of Glucose Level Using PPG Signal, 2020 International Conference on Information and Communication Technology Convergence (ICTC), Jeju, Korea (South), 2020, pp. 733-736, doi: 10.1109/ICTC49870.2020.9289629
• Pappada, S. M., Cameron, B. D., Rosman, P. M., Bourey, R. E., Papadimos, T. J., Olorunto, W., & Borst, M. J. (2011). Neural network-based real-time prediction of glucose in patients with insulin-dependent diabetes. Diabetes technology & therapeutics, 13(2), 135-141.
• Aishah, A. F. Q. A., Zainuriah, M. R., & Norhilda, A. K. (2019). Multiple linear regression model analysis in predicting fasting blood glucose level in healthy subjects. In IOP Conference Series: Materials Science and Engineering (Vol. 469, No. 1, p. 012050). IOP Publishing.