Table I. Definition of DyNeuMo Mk-2 algorithm system requirements; full requirements can be found in [14].

User Needs
Predicate Therapy Support	The research system must support existing stimulation parameters for therapy delivery (amplitude, frequency, pulse width)
Slow-Adaptive Stimulation Scheme	Stimulation based on assignment of discrete stimulation parameters to specific time intervals in throughout the circadian cycle. Temporal mapping facilitated through the clinician programmer
Fast-Adaptive Sensing Scheme	Inertial accelerometer (three axis) – with DC accuracy for posture detection and AC capability for activity, tremor, gait, shocks and free-fall – flexibility for configuration to specific therapy needs; fully configurable through telemetry update
	Biopotential amplifier – local field potentials measured from implanted leads, including spectral power analysis or evoked potentials; fully configurable through telemetry update
Algorithm Methods and Priority	Slow-adaptive and Fast-adaptive algorithms classify a defined state and map this state to a specific stimulation parameter set pre-defined by the clinician. Priority is currently defined to the latest algorithm interrupt; upon termination of the fast-adaptive state, the signal will return to the slow-adaptive setting
Algorithm Power Allowance	Desired: the adaptive algorithm must draw no more than 25% of the nominal therapy power (e.g. 100$\mathrm{\mu }$W for deep brain stimulation). Mandatory: the power consumption will not require in excess of a daily recharge
Algorithm Slow-Adaptive Granularity	Stimulation epochs will be provided in a 30 minute (max) intervals through a 24-hour calendar
Algorithm Fast-Adaptive Latency	< 20 ms from event detection to stimulation adjustment
Algorithm Risk Mitigations	Please reference [17] [14] for an overview of therapy limits, ramp rates, and fallback modes used for algorithms