Embedding digital chronotherapy into bioelectronic medicines

Abstract

Biological rhythms pervade physiology and pathophysiology across multiple timescales. Because of the limited sensing and algorithm capabilities of neuromodulation device technology to-date, insight into the influence of these rhythms on the efficacy of bioelectronic medicine has been infeasible. As the development of new devices begins to mitigate previous technology limitations, we propose that future devices should integrate chronobiological considerations in their control structures to maximize the benefits of neuromodulation therapy. We motivate this proposition with preliminary longitudinal data recorded from patients with Parkinson's disease and epilepsy during deep brain stimulation therapy, where periodic symptom biomarkers are synchronized to sub-daily, daily, and longer timescale rhythms. We suggest a physiological control structure for future bioelectronic devices that incorporates time-based adaptation of stimulation control, locked to patient-specific biological rhythms, as an adjunct to classical control methods and illustrate the concept with initial results from three of our recent case studies using chronotherapy-enabled prototypes.

Keywords: Bioelectronics; Biological sciences; Biotechnology; Neuroscience.

Graphical abstract

Figure 1

Control Structure for a physiological system Homeostasis in physiological systems is maintained by a combination of adaptive, feedforward, and feedback control strategies. Feedforward and feedback controllers (represented in black) provide time-localized regulation of physiological processes. To maintain homeostasis for time-varying physiological systems, the adaptive controller (represented in red) adjusts elements of the feedforward and feedback controllers in a time-dependent manner in such a way that it is synchronized to patient-specific biological rhythms or in response to slow or intermittent feedbacks. Feedforward and feedback controllers may be implemented using approaches from either classical control theory, such as on-off and proportional-integral-derivative control, or more modern control algorithms, such as fuzzy and model predictive controls. To enable long-term enhancement of future neuromodulation therapies, machine learning and optimization techniques may additionally be incorporated in the adaptive controller to enhance therapy optimization overtime. This figure has been adapted with permission from (Houk, 1988).

Figure 2

Chronic subcortical recordings from the subthalamic nucleus in a patient with Parkinson's disease Bilateral subcortical recordings from a PD patient implanted with a sensing enabled DBS device, the Medtronic Percept PC™, tracking subcortical beta band power. Note the strong rhythmic circadian fluctuations in beta amplitude over the 24-h cycle, in addition to the influence of stimulation intervention which suppresses beta and compresses the circadian beta fluctuation cycle.

Figure 3

Example seizure periodicity in canine epilepsy (A) Raw iEEG tracings. iEEG tracings are displayed at multiple timescales to illustrate a single seizure and a pair of seizure clusters separated by several days. Red triangles indicate seizure onset. (B) Circadian, circaseptan, and monthly seizure periodicity. Daily, weekly, and monthly circular histograms of seizure occurrence. Concentric rings demarcate the number of seizures (five seizures per concentric ring in the daily histogram, two seizures per ring in the weekly and monthly histograms. The red bar is the resultant vector or R value. Dog four and Dog five showed statistically significant daily and monthly periodicity, respectively, as indicated by the red font and asterisk. Dog five also indicates a trend toward significant weekly periodicity;’ however, this did not survive statistical false discovery rate correction. This figure by (Gregg et al., 2020) is licensed under CC BY-NC 4.0 and adapts Figures 1B and 3 from their original publication.

Figure 4

Fully-embedded sleep adaptive closed loop DBS control in a patient with PD (A) 24 h performance of dual motor and sleep state classifiers for closed loop amplitude modulation. When the sleep classifier detects sleep, the motor state classifier is disabled and fixed amplitude ope loop stimulation is applied. Otherwise, when sleep is not detected, the motor state classifier increases or reduces the stimulation amplitude when the monitored cortical gamma activity is low or high, respectively. (B) Sleep detector performance over 47 days. The heatmap summarizes the sleep and motor state classifier performance over 47 days measured across four patients. Blue boxes indicate periods classified as sleep, whereas yellow and green boxes indicate the motor state as summarized in the state table. This figure by Gilron et al., 2021 is licensed under CC BY 4.0 and combines Figures 2A and 4A from their original publication.

Figure 5

The Mayo EPAD distributed brain coprocessor system for daytime/nighttime algorithm scheduling The EPAD system provides bidirectional communication between implantable neuromodulation devices and commercially available electronics. The system has been investigated in both human and canine patients with epilepsy. The system is capable of sleep stage detection based on recorded neurophysiological data and enables targeted adaptation of stimulation during sleep.

Figure 6

Example of DBS chronotherapy for canine epilepsy (A) Infradian stimulation scheduling. The Picostim DyNeuMo-1 system enables time-based scheduling of stimulation adjustments synchronized to the patient's particular infradian seizure rhythm (e.g., a 2 weekperiod). (B) Rose plot illustration of embedded circadian stimulation algorithm adjustments. The inner circle represents the seizure count from the patient diary; the orange tiling is the timing of first seizure onset, whereas the blue accounts for all seizures in a cluster. The algorithm is composed of three states (represented as rings) to facilitate day/nighttime-based stimulation scheduling (inner green ring), a motion-triggered sleep mode to prevent seizure occurrence during daytime napping (middle pale orange ring), and a tap-activated boost mode for preventing breakthrough seizures (outer dark orange ring). (C) Summary of canine seizure frequency and anti-seizure drug dosage preimplantation and postimplantation. Postimplantation, the canine experiences no status epilepticus events and no significant seizure clusters thus resulting in a reduction in rescue medication. This figure by (Zamora et al., 2021) is licensed under CC-BY 4.0 and combines Figures 3 and 4A from their original publication.