A novel adapted MRI-based scheme for Dejour classification of trochlear dysplasia

To elaborate an optimized scheme for the Dejour classification of trochlear dysplasia based on axial and sagittal MR images and to evaluate its intra- and inter-reader reliability.

Over a period of 20 months patients with a knee MRI and the diagnosis of trochlear dysplasia were retrospectively included. Exclusion criteria were incomplete examination, qualitatively non-diagnostic examination, post trochlear surgery, missing informed consent for research purposes. Three independent evaluations were performed by two radiologists: first using an established description of the Dejour classification (types A–D) and then two evaluations using a new adapted scheme (types A–D). The adapted scheme includes a shallow trochlea, in type A no spur/no cliff, in type B with spur/no cliff, in type C no spur/with cliff, and in type D with spur/with cliff.

One hundred seventy-one knee MRIs (female:65.5%; left side:52.6%) were included with a median age of 34.3 years (range:11.3–79.2). Inter-reader reliability using the established description was fair for the four-type-classification (kappa(k) = 0.23; 95%CI:0.11–0.34), fair for differentiation low-grade versus high-grade dysplasia (k = 0.28;0.13–0.43), slight for differentiation spur versus no-spur types (k = 0.20;0.05–0.34). Inter-reader reliability using the adapted scheme was substantial (k = 0.79;0.75–0.83) for the four-type-classification, substantial for differentiation low-grade versus high-grade dysplasia (k = 0.80;0.75–0.85), substantial for differentiation spur versus no-spur presence (k = 0.76;0.71–0.81). Intra-reader reliability was almost perfect for the adapted scheme (k-values: 0.88–0.95; 95%CIs: 0.84–0.98).

The novel adapted scheme for Dejour classification shows an almost perfect intra-reader reliability and a substantially higher inter-reader reliability. It may become a helpful tool in the daily diagnostic work of radiologists.