犬の胸腔内リンパ節腫大の画像診断：その有用性は？

The intrathoracic lymph nodes (LN) of clinical interest in dogs are the sternal (STLN), cranial mediastinal (CrMLN), and tracheobronchial (TBLN) groups. Although computed tomography (CT) depicts and measures these nodes well, thoracic screening commonly relies on three-view radiographs. We hypothesized that enlargement of these nodes would not be consistently identified radiographically, even by experienced observers. In this retrospective, multicenter study (2012-2023), three board-certified radiologists independently graded three-view thoracic radiographs from 74 dogs. Each LN group was scored on a 5-point scale (1, cannot assess; 2, normal; 3, mild; 4, moderate; 5, marked). Corresponding per-group CT volumes (STLN, CrMLN, and combined TBLN) were calculated using the ellipsoid formula, and analyses were performed at the LN-group level. Inter-rater agreement was estimated with Gwet's AC1, and associations between radiographic grades and CT volumes were evaluated with Spearman's rank correlation. Observer agreement on radiographs was almost perfect (AC1: STLN 0.88, CrMLN 0.93, and TBLN 0.95). Correlations between radiographic grades and CT volumes were weak or nonsignificant; only one observer for TBLN showed a weak positive correlation (ρ = 0.27, p = 0.02). No CT-derived volume threshold yielded a unanimous radiographic classification of enlargement (grade ≥ 3) across observers. Radiographs frequently failed to detect enlarged nodes and occasionally overcalled normal-sized nodes. These findings indicate poor alignment between radiographic grading and CT-measured volume for intrathoracic lymphadenomegaly, despite high interobserver agreement. When accurate intrathoracic LN assessment is expected to influence clinical decision-making, CT should be considered.