Sub-optimal clinical outcomes of conventional deep brain stimulation (cDBS) in treating Parkinson's Disease (PD) have boosted the development of new solutions to improve DBS therapy. Adaptive DBS (aDBS), consisting of closed-loop, real-time changing of stimulation parameters according to the patient's clinical state, promises to achieve this goal and is attracting increasing interest in overcoming all of the challenges posed by its development and adoption. In the design, implementation, and application of aDBS, the choice of the control variable and of the control algorithm represents the core challenge. The proposed approaches, in fact, differ in the choice of the control variable and control policy, in the system design and its technological limits, in the patient's target symptom, and in the surgical procedure needed. Here, we review the current proposals for aDBS systems, focusing on the choice of the control variable and its advantages and drawbacks, thus providing a general overview of the possible pathways for the clinical translation of aDBS with its benefits, limitations and unsolved issues.
