Premium
Stepwise algorithm using computed tomography and magnetic resonance imaging for diagnosis of fat‐poor angiomyolipoma in small renal masses: Development and external validation
Author(s) -
Tanaka Hajime,
Fujii Yasuhisa,
Tanaka Hiroshi,
Ishioka Junichiro,
Matsuoka Yoh,
Saito Kazutaka,
Uehara Sho,
Numao Noboru,
Yuasa Takeshi,
Yamamoto Shinya,
Masuda Hitoshi,
Yonese Junji,
Kihara Kazunori
Publication year - 2017
Publication title -
international journal of urology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.172
H-Index - 67
eISSN - 1442-2042
pISSN - 0919-8172
DOI - 10.1111/iju.13354
Subject(s) - magnetic resonance imaging , medicine , angiomyolipoma , tomography , computed tomography , radiology , nuclear medicine , kidney
Objectives To develop a stepwise diagnostic algorithm for fat‐poor angiomyolipoma in small renal masses. Methods Two cohorts of small renal masses <4 cm without an apparent fat component that was pathologically diagnosed were included: 153 cases (18 fat‐poor angiomyolipomas/135 renal cell carcinomas) for model development and 71 cases (seven fat‐poor angiomyolipomas/59 renal cell carcinomas/5 oncocytomas) for validation. Dynamic contrast‐enhanced computed tomography, magnetic resonance imaging and clinical findings were analyzed. Based on multivariate analysis, we developed two prediction models for fat‐poor angiomyolipoma, the computed tomography model and the computed tomography + magnetic resonance imaging model, and a stepwise algorithm that proposes the sequential use of computed tomography and magnetic resonance imaging. Results The computed tomography model, which was composed of female aged <50 years, high attenuation on unenhanced computed tomography, less enhancement than the normal renal cortex and homogeneity in the corticomedullary phase, differentiated tumors with none of the factors as the low angiomyolipoma‐probability group, and the others were candidates for the computed tomography + magnetic resonance imaging model. The computed tomography + magnetic resonance imaging model, consisting of the first three factors of the computed tomography model, low signal intensity and absence of pseudocapsule on T2‐weighted magnetic resonance imaging, re‐stratified the tumors into low, intermediate and high angiomyolipoma‐probability groups. The incidence of fat‐poor angiomyolipoma in each group was 0%, 26% and 93%, respectively (area under the curve 0.981). External validation by two readers showed a high area under the curve (0.912 and 0.924) for each. The interobserver agreement was good (kappa score 0.77). Conclusions The present algorithm differentiates fat‐poor angiomyolipoma in small renal masses with high accuracy by adding magnetic resonance imaging to computed tomography in selected patients.