[Volumetry of the lateral ventricles in dogs: Examination of intra- and inter-rater reliability based on manual image segmentation].

Volumetric methods offer new approaches for the diagnosis and treatment of neurologic diseases, as well as for an improved assessment of prognosis. The aim of this study was to determine the intra- and inter-rater reliability of volumetric measurements in manual image segmentation of the lateral ventricles of dogs, taking different sequences, body mass and lateral ventricle size into account.MRI datasets of 9 dogs, representing 3 different body mass groups and 3 different ventricular size groups, were obtained retrospectively from the image database of our clinic. For each patient, 2D sequences were available in T2, T1 and FLAIR weighting in the transverse plane. The lateral ventricles were segmented and measured volumetrically 10 times each by 5 raters. A further 9 dogs were selected for evaluation of a 3D T1 sequence. One of the raters (JD) performed segmentation and measurement 10 times. Statistical analysis included descriptive data and calculation of intraclass correlation coefficient (ICC) for intra- and inter-rater reliability.Intra-rater reliability: The smallest variation in values was measured in the T1 sequence for dogs with small body mass and small ventricular size. For larger dogs and larger lateral ventricles, the 3D T1 sequence showed less variation. Only 2/36 calculated ICCs indicated moderate reliability (1 each for T1 and 3D T1 sequences). The remaining ICC (34/36) indicated poor reliability. Inter-rater reliability: The sequence with the least variation was observer dependent. All calculated ICCs (27/27) indicated poor reliability regardless of sequence, body mass or subjective ventricular size.Volumetric measurement of canine lateral ventricles by manual segmentation shows major weaknesses in terms of reliability, both for single and multiple examiner measurements.Manual segmentation as performed in this study is not suitable for routine clinical use due to poor reliability and time-consuming application.