Measures of mean cartilage thickness over predefined regions in the femoral plate using magnetic resonance imaging have provided important insights into the characteristics of knee osteoarthritis (OA), however, this quantification method suffers from the limited ability to detect OA-related differences between knees and loses potentially important information regarding spatial variations in cartilage thickness. The objectives of this study were to develop a new method for analyzing patterns of femoral cartilage thickness and to test the following hypotheses: (1) asymptomatic knees have similar thickness patterns, (2) thickness patterns differ with knee OA, and (3) thickness patterns are more sensitive than mean thicknesses to differences between OA conditions. Bi-orthogonal thickness patterns were extracted from thickness maps of segmented magnetic resonance images in the medial, lateral, and trochlea compartments. Fifty asymptomatic knees were used to develop the method and establish reference asymptomatic patterns. Another subgroup of 20 asymptomatic knees and three subgroups of 20 OA knees each with a Kellgren/Lawrence grade (KLG) of 1, 2, and 3, respectively, were selected for hypotheses testing. The thickness patterns were similar between asymptomatic knees (coefficient of multiple determination between 0.8 and 0.9). The thickness pattern alterations, i.e., the differences between the thickness patterns of an individual knee and reference asymptomatic thickness patterns, increased with increasing OA severity (Kendall correlation between 0.23 and 0.47) and KLG 2 and 3 knees had significantly larger thickness pattern alterations than asymptomatic knees in the three compartments. On average, the number of significant differences detected between the four subgroups was 4.5 times greater with thickness pattern alterations than mean thicknesses. The increase was particularly marked in the medial compartment, where the number of significant differences between subgroups was 10 times greater with thickness pattern alterations than mean thickness measurements. Asymptomatic knees had characteristic regional thickness patterns and these patterns were different in medial OA knees. Assessing the thickness patterns, which account for the spatial variations in cartilage thickness and capture both cartilage thinning and swelling, could enhance the capacity to detect OA-related differences between knees.