Heart failure (HF) is the end-stage of all cardiac diseases, characterized by high prevalence, high mortality, and heavy social and economic burden. Early warning of HF exacerbation is of great value for outpatient management and reducing readmission rates. Currently, remote dynamic monitoring technology, which captures changes in hemodynamic and physiological parameters of HF patients, has become the primary method for early warning and is a hot research topic in clinical studies. This paper systematically reviews the progress in this field, which was categorized into invasive monitoring based on implanted devices, non-invasive monitoring based on wearable devices, and other monitoring technologies based on audio and video. Invasive monitoring primarily involves direct hemodynamic parameters such as left atrial pressure and pulmonary artery pressure, while non-invasive monitoring covers parameters such as thoracic impedance, electrocardiogram, respiration, and activity levels. These parameters exhibit characteristic changes in the early stages of HF exacerbation. Given the clinical heterogeneity of HF patients, multi-source information fusion analysis can significantly improve the prediction accuracy of early warning models. The results of this study suggest that, compared with invasive monitoring, non-invasive monitoring technology, with its advantages of good patient compliance, ease of operation, and cost-effectiveness, combined with AI-driven multimodal data analysis methods, shows significant clinical application potential in establishing an outpatient management system for HF.
Patients with acute heart failure (AHF) often experience dyspnea, and monitoring and quantifying their breathing patterns can provide reference information for disease and prognosis assessment. In this study, 39 AHF patients and 24 healthy subjects were included. Nighttime chest-abdominal respiratory signals were collected using wearable devices, and the differences in nocturnal breathing patterns between the two groups were quantitatively analyzed. Compared with the healthy group, the AHF group showed a higher mean breathing rate (BR_mean) [(21.03 ± 3.84) beat/min vs. (15.95 ± 3.08) beat/min, P < 0.001], and larger R_RSBI_cv [70.96% (54.34%–104.28)% vs. 58.48% (45.34%–65.95)%, P = 0.005], greater AB_ratio_cv [(22.52 ± 7.14)% vs. (17.10 ± 6.83)%, P = 0.004], and smaller SampEn (0.67 ± 0.37 vs. 1.01 ± 0.29, P < 0.001). Additionally, the mean inspiratory time (TI_mean) and expiration time (TE_mean) were shorter, TI_cv and TE_cv were greater. Furthermore, the LBI_cv was greater, while SD1 and SD2 on the Poincare plot were larger in the AHF group, all of which showed statistically significant differences. Logistic regression calibration revealed that the TI_mean reduction was a risk factor for AHF. The BR_ mean demonstrated the strongest ability to distinguish between the two groups, with an area under the curve (AUC) of 0.846. Parameters such as breathing period, amplitude, coordination, and nonlinear parameters effectively quantify abnormal breathing patterns in AHF patients. Specifically, the reduction in TI_mean serves as a risk factor for AHF, while the BR_mean distinguishes between the two groups. These findings have the potential to provide new information for the assessment of AHF patients.