QURESHI, SCHETININ: SEGMENTATION AND ESTIMATION OF BRAININDICES 1 Computer-Aided Segmentation and Estimation of Indices in Brain CT Scans Adnan Qureshi adnan.qureshi@beds.ac.uk Vitaly Schetinin vitaly.schetinin@beds.ac.uk Institute for Research in Applicable Computing University of Bedfordshire Luton, UK. Abstract The importance of neuro-imaging as one of the biomarkers for diagnosis and prog- nosis of pathologies and traumatic cases is well established. Doctors routinely perform linear measurements on neuro-images to ascertain severity and extent of the pathology or trauma from significant anatomical changes. However, it is a tedious and time consum- ing process and manually assessing and reporting on large volume of data is fraught with errors and variation. In this paper we present a novel technique for segmentation of sig- nificant anatomical landmarks using artificial neural networks and estimation of various ratios and indices performed on brain CT scans. The proposed method is efficient and robust in detecting and measuring sizes of anatomical structures on noncontrast CT scans and has been evaluated on images from subjects with ages between 5 to 85 years. Results show that our method has average ICC of ≥ 0.97 and, hence, can be used in processing data for further use in research and clinical environment. 1 Overview and problem statement Linear measurements on axial CT scans provide clinicians and surgeons opportunity to ascer- tain differential diagnoses of neuropsychiatric disorders, outcomes of clinical and surgical interventions, geriatric changes and deleterious effects of drugs. Quantitative assessment of neuro-images is an effective approach to reveal structural changes in conditions such as Alzheimer’s disease (AD), Schizophrenia, Huntington’s disease, hydrocephalus and many other neurological and psychiatric disorders [1, 2, 11, 12]. The typical measurements per- formed on the axial CT scans (fig. 1a, details in table 1) are used to estimate indices and ratios such as Evan’s Ratio (ER), Bifrontal Index (BFI), Bicaudate Index (BCI), Cella Media Index (CMI), Frontal Horn Index (FHI), Ventricular index (VI), Huckman number (HN) and 3rd ventricle width (V3) [3]. Manual measurements of regions of interest (ROI) are still con- sidered ’gold standard’, however, these are time consuming and a robust and efficient method is required to assist the researchers and to our knowledge, there is no automated system for performing measurements and calculation of these from CT scans. Computer aided quantitative radiology and volumetry studies of the human anatomy and pathology have been undertaken by many researchers and a surfeit of image segmentation methods have been proposed [10]. However, in medical images an ROI can have dissimi- larity of pixel intensity, inhomogeniety of background contribution and noise, and spatially c  2014. The copyright of this document resides with its authors. It may be distributed unchanged freely in print or electronic forms.