Helping stroke survivors regain function has long been a challenge in neurorehabilitation, and the field is changing from traditional, experience-based clinical judgments towards smarter and more interactive human-machine models. In this review, we follow a simple thread: start with clinical needs, then look at technology integration, then decision support, and finally system reshaping. We survey recent advances in intelligent sensing and digital therapeutics for stroke rehabilitation, and three stand out. First, flexible smart sensors plus multimodal data fusion now enable us to measure dysfunction precisely and quantitatively. Second, brain-computer interfaces with closed-loop feedback systems can drive active and adaptive neural remodeling. Third, a "Meta-consultation" model helps build a coordinated rehab network which connects hospital, community, and home-based training. Together, these changes push stroke rehabilitation toward more precise, more personalized, and more affordable care. We also discuss the major challenges that remain before these technologies can be used widely in clinics, and suggest directions for future interdisciplinary research. Our goal is to provide both theoretical frameworks and practical pathways for building a human-centered smart rehab ecosystem.