Effect of MRI-based semiautomatic size-assessment in cerebral metastases on the RANO-BM classification.

Abstract

Evaluation of asemiautomatic software algorithm for magnetic resonance imaging (MRI)-based assessment of cerebral metastases in cancer patients. Brain metastases (n = 131) in 38patients, assessed by contrast-enhanced MRI, were retrospectively evaluated at two timepoints (baseline, follow-up) by two experienced neuroradiologists in ablinded manner. The response assessment in neuro-oncology (RANO) criteria for brain metastases (RANO-BM) were applied by means of asoftware (autoRANO-BM) as well as manually (manRANO-BM) at an interval of 3weeks. The average diameter of metastases was 12.03 mm (SD ± 6.66 mm) for manRANO-BM and 13.97 mm (SD ± 7.76 mm) for autoRANO-BM. Diameter figures were higher when using semiautomatic measurements (median = 11.8 mm) as compared to the manual ones (median = 10.2 mm; p = 0.000). Correlation coefficients for intra-observer variability were 0.993 (autoRANO-BM) and 0.979 (manRANO-BM). The interobserver variability (R1/R2) was 0.936/0.965 for manRANO-BM and 0.989/0.998 for autoRANO-BM. A total of 19 lesions (15%) were classified differently when using semiautomatic measurements. In 14cases with suspected disease progression by manRANO-BM astable course was found according to autoRANO-BM. Computerized measuring techniques can aid in the assessment of cerebral metastases by reducing examiner-dependent effects and may consequently result in adifferent classification according to RANO-BM criteria.