BCI enable users to control external devices or interact with computer systems through brain activity alone.MI-BCI leverage the imagination of movements—eliciting neural activity akin to actual execution—to provide intuitive, non-muscular control. Inspite significant advances, MI–BCIs remain largely confined to laboratory environments due to challenges in signal reliability, algorithmic robustness, and user engagement. This PhD thesis investigates how systematic enhancements across the key MI–BCI subsystems—signal acquisition, user training, algorithmic processing, and interface design—can enable reliable, motivating, and practical use outside controlled settings, ultimately supporting Activities of Daily Living (ADL) for both healthy individuals and persons with Spinal Cord Injury (SCI). Four interconnected research questions are addressed: (1) the feasibility of MI in SCI populations, (2) the influence of instructional medium on MI skill acquisition and engagement, (3) the potential of channel reduction and advanced algorithms to enhance classification and real-time control, and (4) the design of immersive and safe interfaces for real-world ADL tasks. Empirical investigations demonstrate that immersive VR training enhances MI skill acquisition and motivation; single-channel and Gaussian Process–based classifiers enable robust and portable control; and compliant robotic interfaces facilitate continuous, safe interaction with functional tasks. Collectively, these findings show that MI–BCIs can transition from controlled laboratory paradigms to functional, user-centered systems capable of supporting ADL. By integrating advances across all subsystems, this research provides a methodological framework for translating MI–BCIs into real-world applications—offering tangible benefits for rehabilitation, independence, and quality of life.