Figure 2.

A suggested scheme of the closed loop machine learning system for improving the response to diurectis in patients with HF. Inputs from various sensors are being incorporated in a dynamic manner for determining the timings of administration and dosages in a personalized approach. In a dynamic changing system with multiple alternating parameters, a continuously changing dosing regimen is generated based on the individualized-variability pattern that comprises data from the ANS (e.g., HRV), heart, and kidney. Cardiac, renal, and ANS, along with numerous other physiological and biological processes, exhibit intrinsic randomness. These variations in their activity are also contributed from the generalized circadian rhythm and from intrinsic biological rhythms. Randomized treatment administration within the preset limits augmented with personalized signature from the clinical status, echocardiographic, HRV, pulmonary artery pressure monitoring, biomarkers, and more can improve outcomes of the diuretic treatment. In addition, randomization might prevent renal adaptivity involved in the development of diuretic resistance. ANS, autonomic nervous system; HRV, heart rate variability.