Our recent study on patient-specific feature selection revealed the importance of computational methods in guid-ing deep brain stimulation (DBS), a neuromodulation treatment for Parkinson’s disease (PD) that is fast expand-ing to a wide range of brain disorders. Contrary to the prevailing clinical practice, a parallel study of ours showed evidence that neurophysiological, not anatomical, information extracted from the stimulated subthalamic nucleus (STN) should guide the DBS surgery in PD. The neural information is acquired from the STN and nearby nuclei during a subjective, cumbersome and error-prone surgical procedure of finding the optimal stimulation location via microelectrode recordings, with sub-millimeter precision. Surprisingly, our and others’ data-driven algorithms that inform the intranuclear placement of the stimulator do not incorporate neurophysiologically derived features. Here we propose expanding by seven times our set of available features to include potential neuromarkers that can guide, secure, and even predict the surgical outcome. The most informative combination of the new features, identified by a random forest classifier, will become the input to a conditional random field (CRF) model intro-ducing contextual interference among features and nuclei. The CRFs will make structured predictions of the STN boundaries by taking into account for the first time the sequential order of the neurophysiological signals as microelectrodes penetrate through the STN. The blind identification of the recorded nuclei will enable intranuclear neuromarkers associated with the STN sub-territories to predict the DBS outcome. We anticipate that our neuro-computational approach can expand to intraoperatively guide clinical decisions for alleviating the symptoms of neurological diseases.