Recognizing subtle cartilage changes in the preclinical stage of osteoarthritis (OA) is essential for early diagnosis. To this end, the ability of the ultrasound signal intensity to detect macroscopically undetectable cartilage change was investigated. In this study, cartilage of rabbit OA model and human OA samples was examined by macroscopic evaluation, ultrasound signal intensity, histology with Mankin scores, and Fourier transform infrared imaging (FTIRI) analysis. Rabbit OA was induced by anterior cruciate ligament transection and evaluated at 1, 2, 4 and 12 weeks. Twenty human samples were harvested during total knee arthroplasty from OA patients who had macroscopically normal human cartilage (ICRS grade 0) on the lateral femoral condyle. In the animal study, there was no macroscopic OA change at 2 weeks, but histology detected degenerative changes at this time point. Ultrasound signal intensity also detected degeneration at 2 weeks. In human samples, all samples were obtained from macroscopically intact site, however nearly normal (0 ≤ Mankin score <2), early OA (2 ≤ Mankin score <6), and moderate OA (6 ≤ Mankin score <10) samples were actually intermixed. Ultrasound signal intensity was significantly different among these 3 stages and was well correlated with Mankin scores (R = -0.80) and FTIR parameters related to collagen and proteoglycan content in superficial zone. In conclusion, ultrasound can detect microscopic cartilage deterioration when such changes do not exist macroscopically, reflecting superficial histological and biochemical changes.