Ingrid Agartz poster abstract

Reliability and reproducibility of MR brain tissue segmentation

I. Agartz1, G. Okugawa2, G. Sedvall1

1Karolinska Institutet, Department of Clinical Neuroscience, Section of Psychiatry, Karolinska Hospital, Stockholm, Sweden; 2UWCM, Department of Neuropsychiatry, Kansai Medical University, Osaka, Japan

Reliability and reproducibility of segmented tissue class volumes grey matter (GM), white matter (WM) and CSF were investigated using BRAINS and a 1.5 T MRI. Right and left side volume measurements were obtained using continuous and a discrete classifier.

Two different sets of MR-scanc were used: 10 subjects (inter- and intrareliability) and 11 subjects (scan-rescan reproducibility). Intracllass correlation (ICC) coefficient (was used as reproducibility index.

For the first scan set, values were 87.0% total brain volume (TBR) of intracranial volume (54.4% GM, 32.5% WM) and 13.0% CSF (continuous) and 93.5% TBV (57.1% GM, 36.5% WM) and 6.4% CSF (discrete). For the second set, values were 86.3% TBV (54.5% GM, 31.8% WM) and 13.7% CSF (continuous) and 93.0% TBV (57.5% GM, 35.5% WM) and 7.0% CSF (discrete). For both sets, TBV was 7% larger and CSF volume 7% smaller by the discrete classification than the continuous.

Inter- and intrareliability: ICCs were above 0.99 for continuous and discrete measures exept GM (discrete) (r2>0.96). SCan-rescan reproducibility: ICCs for continuous and discrete classifications were excellent (r2>0.99 and r2>0.97).

View poster