Prediabetes is an important stage of abnormal glucose metabolism in the development of diabetes, and its early diagnosis is crucial for global diabetes prevention and control. To explore non-invasive detection methods for prediabetes, heart rate variability (HRV) signals were utilized. By introducing a multi-scale analysis strategy, the global information of the signals was revealed, as well as subtle but important changes at different scales. The CatBoost algorithm was used for classification task. The results show that this method achieves an accuracy of 88.52%, a sensitivity of 83.40%, a specificity of 91.82%, a precision of 86.73%, and an F1-score of 87.40% on the dataset. This study provides a new approach for the diagnosis of prediabetes. The results are especially suitable for wearable devices, offering a potential solution for daily self-health monitoring and disease prevention.
为了解决这些问题并提供更加实用和精确的方法,本文提出使用5 min HRV信号来检测糖尿病前期,同时引入多尺度分析策略,以进一步提高检测效果.本文首先关注特定生理尺度的整体特征并进行分析,以捕捉HRV信号在特定生理尺度下的显著变化,这有助于理解信号的整体模式与糖尿病前期之间的关系.其次,为了更细致地揭示信号中的变化,本文利用小波散射网络将信号分解为不同尺度下的散射系数,以捕捉信号在精细化尺度内的微小变化,从而能够更准确地分析HRV信号的细节特征,揭示出信号中微弱但重要的模式.最后,本文构建了基于CatBoost算法的分类模型,将提取到的特定生理尺度的整体特征与精细化尺度特征结合起来,实现了分类任务.CatBoost算法能够有效地整合所提供的特征,进一步扩展特征维度,从而实现更精确的决策.该分类模型通过对多尺度特征学习到的信息进行综合考量,为糖尿病前期的检测提供了更准确的结果.
由图9可知,在基于特定生理尺度特征的单尺度检测模型中,LFnu,Pt和HFnu的贡献值较大,其他特征的贡献值较小;在基于精细化尺度特征的单尺度检测模型中,列出了前10个贡献值最大的尺度特征,J94,J85,J67和J195的贡献值较大,其他特征的贡献值表现相当;在基于多尺度特征融合的检测模型中,列出了前10个贡献值最大的尺度特征,LFnu,Pt和HFnu的贡献值依旧是最大的,其余特征的贡献较小.然而J10,J94等精细化尺度特征的贡献值仍然高于HF等其他特定生理尺度特征.多尺度模型的各种性能指标皆得到了大幅度提升,说明J10,J94等特征可以有效地辅助模型决策,小波散射网络提取的精细化尺度特征发挥了较为重要的作用.在这项工作中,所提出的多尺度特征模型在全部性能指标评估中表现最好,准确率达到88.52%,敏感度达到83.40%,特异度达到91.82%,精确度达到86.73%,F1分数达到87.40%,比单一的特定生理尺度模型和精细化尺度模型皆有大幅提升.尤其是模型中代表检测糖尿病前期患者能力的敏感度指标,这意味着该模型可以检测到绝大多数糖尿病前期患者.此外,模型完全基于5 min HRV信号检测糖尿病前期,具有无创、简便的优点.
International Diabetes Federation. IDF diabetes atlas[M]. 10th ed. Brussels: International Diabetes Federation, 2021.
[2]
WanH, WangY Y, FangS J, et al. Associations between the neutrophil-to-lymphocyte ratio and diabetic complications in adults with diabetes: a cross-sectional study[J]. Journal of Diabetes Research, 2020, 2020(1): 6219545.
[3]
ElSayedN A, AleppoG, ArodaV R, et al. 17 diabetes advocacy: standards of care in diabetes—2023[J]. Diabetes Care, 2023, 46(sup1): 279-280.
[4]
Echouffo-TcheuguiJ B, SelvinE. Prediabetes and what it means: the epidemiological evidence[J]. Annual Review of Public Health, 2021, 42: 59-77.
[5]
FaerchK, HulmánA, SolomonT P J. Heterogeneity of prediabetes and type 2 diabetes: implications for prediction, prevention and treatment responsiveness[J]. Current Diabetes Reviews, 2016, 12(1): 30-41.
[6]
LiuQ, ZhouQ, HeY F, et al. Predicting the 2-year risk of progression from prediabetes to diabetes using machine learning among Chinese elderly adults[J]. Journal of Personalized Medicine, 2022, 12(7): 1055.
[7]
TabákA G, HerderC, RathmannW, et al. Prediabetes: a high-risk state for diabetes development[J]. The Lancet, 2012, 379(9833): 2279-2290.
[8]
BrannickB, WynnA, Dagogo-JackS. Prediabetes as a toxic environment for the initiation of microvascular and macrovascular complications[J]. Experimental Biology and Medicine, 2016, 241(12): 1323-1331.
[9]
CaiX Y, ZhangY L, LiM J, et al. Association between prediabetes and risk of all cause mortality and cardiovascular disease: updated meta-analysis[J]. BMJ, 2020, 370: m2297.
[10]
MutieP M, Pomares-MillanH, Atabaki-PasdarN, et al. An investigation of causal relationships between prediabetes and vascular complications[J]. Nature Communications, 2020, 11: 4592.
[11]
CosicV, JakabJ, PravecekM K, et al. The importance of prediabetes screening in the prevention of cardiovascular disease[J]. Medical Archives, 2023, 77(2): 97-104.
[12]
ToboreI, KandwalA, LiJ Z, et al. Towards adequate prediction of prediabetes using spatiotemporal ECG and EEG feature analysis and weight-based multi-model approach[J]. Knowledge-Based Systems, 2020, 209: 106464.
[13]
WangL Y, MuY, ZhaoJ, et al. IGRNet: a deep learning model for non-invasive, real-time diagnosis of prediabetes through electrocardiograms[J]. Sensors, 2020, 20(9): 2556.
[14]
LinY C, LinC S, ChangT S, et al. Early sensory neurophysiological changes in prediabetes[J]. Journal of Diabetes Investigation, 2020, 11(2): 458-465.
[15]
OliveiraC M, GhezziA C, CambriL T. Higher blood glucose impairs cardiac autonomic modulation in fasting and after carbohydrate overload in adults[J]. Applied Physiology, Nutrition, and Metabolism, 2021, 46(3): 221-228.
[16]
CoopmansC, ZhouT L, HenryR M A, et al. Both prediabetes and type 2 diabetes are associated with lower heart rate variability: the maastricht study[J]. Diabetes Care, 2020, 43(5): 1126-1133.
[17]
RajendraA U, PaulJ K, KannathalN, et al. Heart rate variability: a review[J]. Medical and Biological Engineering and Computing, 2006, 44(12): 1031-1051.
[18]
VijayC, DarshanM, VishnuR. Cardiac autonomic dysfunction and ECG abnormalities in patients with type 2 diabetes mellitus—a comparative cross-sectional study[J]. National Journal of Physiology, Pharmacy and Pharmacology, 2016, 6(3): 178.
[19]
IgbeT, LiJ Z, KandwalA, et al. An absolute magnitude deviation of HRV for the prediction of prediabetes with combined artificial neural network and regression tree methods[J]. Artificial Intelligence Review, 2022, 55(3): 2221-2244.
Chinese Diabetes Society. Type 2 diabetes prevention and treatment comprehensive guide (2020 edition) [J]. Chinese Journal of Diabetes,2021, 13(4): 317-411.
[22]
CuiX R, TianL R, LiZ W, et al. On the variability of heart rate variability—evidence from prospective study of healthy young college students[J]. Entropy, 2020, 22(11): 1302.
[23]
SassiR, CeruttiS, LombardiF, et al. Advances in heart rate variability signal analysis: joint position statement by the E-cardiology ESC working group and the European Heart Rhythm Association co-endorsed by the Asia Pacific Heart Rhythm Society[J]. Europace, 2015, 17(9): 1341-1353.
[24]
PanJ, TompkinsW J. A real-time QRS detection algorithm[J]. IEEE Transactions on Biomedical Engineering, 1985, 32(3): 230-236.
[25]
CataiA M, PastreC M, de GodoyM F, et al. Heart rate variability: are you using it properly? standardisation checklist of procedures[J]. Brazilian Journal of Physical Therapy, 2020, 24(2): 91-102.
[26]
ForteG, FavieriF, CasagrandeM. Heart rate variability and cognitive function: a systematic review[J]. Frontiers in Neuroscience, 2019, 13: 710.
[27]
JwoD J, ChangW Y, WuI H. Windowing techniques, the welch method for improvement of power spectrum estimation[J]. Computers, Materials and Continua, 2021, 67(3): 3983-4003.
[28]
JinY, DuanY L. Wavelet scattering network-based machine learning for ground penetrating radar imaging: application in pipeline identification[J]. Remote Sensing, 2020, 12(21): 3655.
[29]
MallatS. Group invariant scattering[J]. Communications on Pure and Applied Mathematics, 2012, 65(10): 1331-1398.
[30]
OstroumoraL, GusevG, VorobevA, et al. CatBoost: unbiased boosting with categorical features[C]// Advances in Neural Information Processing Systems. Montreal, 2018: 6639-6649.
[31]
BentéjacC, CsörgőA, Martínez-MuñozG. A comparative analysis of gradient boosting algorithms[J]. Artificial Intelligence Review, 2021, 54(3): 1937-1967.
[32]
ZhangY X, ZhaoZ G, ZhengJ H. CatBoost: a new approach for estimating daily reference crop evapotranspiration in arid and semi-arid regions of Northern China[J]. Journal of Hydrology, 2020, 588: 125087.