Inter-Reader and Inter-Modality Agreement for Radiographic and Magnetic Resonance Imaging Subchondral Bone Assessment in the Equine Proximal Phalanx Is Location- and Pattern-Dependent

Standing low-field magnetic resonance imaging (MRI) is widely used for subchondral bone (SCB) assessment of the equine proximal phalanx, but radiography remains the most common first-line imaging modality, and no direct comparison between the two is available for SCB sclerosis and resorption. Fifty-four fetlocks from 52 horses referred for standing MRI were retrospectively assessed at the medial glenoid, sagittal groove and lateral glenoid, using an ordinal scale for sclerosis and a four-type nominal classification for resorption. Seven readers of varied clinical backgrounds performed independent radiographic readings; two board-certified radiologists then established a consensus reading, blinded to MRI. Agreement was quantified with Gwet’s agreement coefficients (inter-reader), weighted and unweighted Cohen’s κ (reader-MRI) and symmetric metrics (consensus-MRI). Inter-reader agreement for sclerosis was location-dependent (Gwet’s agreement coefficient = 0.27–0.73, lowest medially) and substantial-to-almost-perfect for resorption (Gwet’s agreement coefficient ≥ 0.72). The radiographic consensus aligned closely with MRI at the sagittal groove but underestimated marked sclerosis medially and over-called it laterally; fracture-type resorption was reliably detected at the sagittal groove but misclassified medially. Radiography can be relied upon for sagittal groove lesions, but complementary MRI is warranted medially when resorption or clinical suspicion of fracture is present, and to confirm radiographic positive calls laterally.