Objective <p>To improve tissue contrast visualization in water-only images from two-point turbo spin-echo (TSE) Dixon MRI using dark-fat image processing and evaluate in retrospectively acquired clinical knee images.</p> Materials and methods <p>Clinical knee MRI datasets from 36 patients were retrospectively compiled under IRB approval. The dark-fat water-only images were generated and compared with the conventional water-only images from two-point TSE-Dixon MRI. The water-only images from 20 patients were analyzed: (i) qualitatively by two experienced musculoskeletal radiologists, for overall image quality, fat suppression, anatomical visualization, and the absence of artifacts, using Wilcoxon rank tests, and (ii) quantitatively using apparent signal-to-noise ratio (aSNR) and apparent contrast-to-noise ratio (aCNR) from regions of interest (ROIs) with paired <i>t</i>-test. In the remaining 16 patients, the number of cruciate ligament tears identified on MRI was tabulated and compared with arthroscopy results.</p> Results <p>Compared to conventional water-only images, dark-fat water-only images showed improved fat suppression (reader 1, <i>P</i> &lt; 0.0001; reader 2, <i>P</i> &lt; 0.05), improved overall image quality by reader 1 (<i>P</i> &lt; 0.001), and improved visualization of the sciatic nerve for reader 1 (<i>P</i> &lt; 0.01), while there were no significant differences for the other anatomical structures. aSNR was significantly reduced for bone marrow fat (<i>P</i> &lt; 0.0001), and aCNR between meniscus and articular cartilage was significantly improved (<i>P</i> &lt; 0.0001). There were a higher number of posterior cruciate ligament tears identified in MRI compared to arthroscopy.</p> Conclusion <p>Dark-fat processed water-only images improved fat suppression and tissue contrast visualization, particularly sciatic nerves, compared to conventional water-only images from two-point TSE-Dixon MR imaging.</p>