Cerebellum segmentation employing texture properties and knowledge based image processing: applied to normal adult controls and patients

• Published in: Magnetic resonance imaging Vol. 20; no. 5; pp. 425 - 429
• Main Authors: Saeed, N, Puri, B.K
• Format: Journal Article
• Language: English
• Published: New York, NY Elsevier Inc 01.06.2002; Elsevier Science
• Subjects: Biological and medical sciences; Cerebellum; Cerebellum - anatomy & histology; Cerebellum - pathology; Edge-enhancement; Female; Humans; Image Processing, Computer-Assisted; Imaging, Three-Dimensional; Investigative techniques, diagnostic techniques (general aspects); Magnetic Resonance Imaging - methods; Male; Medical sciences; MRI; Nervous system; Radiodiagnosis. Nmr imagery. Nmr spectrometry; Schizophrenia - pathology; Segmentation; Texture; Cerebellum; Edge-enhancement; Segmentation; MRI; Texture; Human; Medical imagery; Exploration; Technique; Nuclear magnetic resonance imaging
• ISSN: 0730-725X; 1873-5894
• DOI: 10.1016/S0730-725X(02)00508-8

A semi-automated method is described for segmenting the cerebellum from T 1-weighted 3-dimensional magnetic resonance imaging scans of adult controls and patients. The method relies on prior knowledge involving a user-defined template as a guide to aid the segmentation of the cerebellum. As the gray and white matter intensity distribution in the cerebellum has a complex pattern, texture information that identified the “graininess” was employed to capture the intensity distribution of voxels. The textural information was used to group voxels in a small circular structuring element as belonging to the cerebellum region. The cerebella from scans of 15 of the 20 subjects were segmented both manually and using the semi-automated procedure; the results were strongly correlated ( r = 0.985, n = 15, p < 0.0001), and the volumes obtained from the two methods differed by 2.3%. The cerebellar volumes in 10 normal subjects and 10 age- and sex-matched patients with a neuropsychiatric disorder (schizophrenia) did not differ significantly ( p = 0.18). The whole cerebellum was segmented in approximately 30 min using the semi-automated procedure. The method described is robust, easy-to-use, fairly fast and gives objective measurements.