Purpose To assess the potential of methods to register changes over time of MR detected bone marrow lesions (BML) in knee osteoarthritis (KOA). Methods & Materials Repeated (3 month – 1 year) Gd-enhanced sagittal T1 fat-suppressed MRI of 13 knees with medial KOA (ACR-criteria). Female/male=12/1; median age 60.5 years (41.1-72.3). Volumes of BMLs in the medial femoral and tibial condyles (52 observations) were measured/calculated on anonymized images by two observers using manual segmentation (MS) and computer assisted tresholding (CAT) relative to signal intensity of normal marrow. Results were compared with MOAKS grading. Results Inter-observer kappa for MS/CAT=0.86/0.84. MS and CAT values were significantly correlated (ơ=0.94) and the median values were similar. MOAKS grade/percentage BML of total femoral or tibial condyle volume were: 1/<33%: n=19; 2/33-66%: n=4; 3/>66%: n=3. Detected BML-change in volumes was: <4%: n=15; <33%: n=9; 33-66%: n=0; >66%: n=2. Conclusion MS and CAT can estimate BML-changes over time not detectable by MOAKS.