Background: Knee ligament injuries are a common consequence of sporting and traffic-related trauma, and accurate, timely diagnosis is essential for guiding management and preventing long-term joint instability. Magnetic resonance imaging (MRI) has become the principal non-invasive modality for evaluating the cruciate and collateral ligaments, but its diagnostic performance across individual structures requires continued prospective validation against arthroscopic findings. Objective: To prospectively evaluate the diagnostic accuracy of MRI in detecting anterior cruciate ligament (ACL), posterior cruciate ligament (PCL), medial collateral ligament (MCL), and lateral collateral ligament (LCL) injuries, using arthroscopy as the reference standard, and to describe the pattern and severity of injury in a cohort of patients with acute knee trauma. Methods: This prospective observational study enrolled 136 patients with clinically suspected knee ligament injury who underwent 1.5-Tesla MRI followed by diagnostic or therapeutic arthroscopy within four weeks. MRI findings were independently graded by two musculoskeletal radiologists and compared against arthroscopic findings to calculate sensitivity, specificity, positive and negative predictive values, accuracy, and inter-observer kappa statistics. Results: ACL injury was the most frequent finding (52.2%), followed by MCL (33.8%), PCL (13.2%), and LCL (8.8%) injuries; 28.7% of patients had multiligament involvement. MRI demonstrated high sensitivity and specificity for ACL (93.0%, 89.5%) and PCL (88.9%, 91.8%) tears, with comparatively lower but acceptable performance for MCL and LCL injuries. Inter-observer agreement was almost perfect for ACL (κ = 0.84) and substantial for the remaining structures (κ = 0.62–0.79). Conclusion: MRI is a highly reliable, reproducible, and non-invasive tool for diagnosing knee ligament injuries, particularly cruciate ligament tears, and should remain the imaging modality of choice prior to arthroscopic planning.