FROM HEMORRHAGE TO MIDLINE SHIFT: A NEW METHOD OF TRACING THE DEFORMED MIDLINE IN TRAUMATIC BRAIN INJURY CT IMAGES Ruizhe Liu*, Shimiao Li*, Chew Lim Tan* Boon Chuan Pang**, C.C. Tchoyoson Lim **, Cheng Kiang Lee** Qi Tian***, Zhuo Zhang *** * School of Computing, National University of Singapore ** National Neuroscience Institute *** Institute for Infocomm Research ABSTRACT In intracranial pathological examinations using CT scan, brain midline shift (MLS) is an important diagnostic feature indicating the pathological severity and patient’s survival possibility. In this paper, we develop a new method of tracing the brain midline shift in traumatic brain injury (TBI) CT images using its original cause - the hemorrhage. Firstly, we model the relationship between the hemorrhage and the midline deformation caused by it using a linear regression model (H-MLS model). Secondly, using the H-MLS model, the deformed midline is predicted from the hemorrhage de- tected in CT images. Finally, the predicted deformed midline is adjusted according to the visual symmetry information. Preliminary experiments show that the proposed method is effective and time-efficient. Index Terms— Medical image analysis, midline shift, computed tomography, traumatic brain injury, hemorrhage 1. INTRODUCTION CT scans are widely used in today’s neurology diagnosis. In intracranial pathological examinations, brain midline shift (MLS) is an important diagnostic feature. Despite the functional difference of the brain hemispheres, the normal anatomical structure of the brain is symmetric to the so called the midsagittal plane, which is shown in a single CT slice as the brain midline. Intracranial pathology changes, such as hemorrhage or tumour may cause the brain midline shift (MLS). Patients presenting MLS may suffer from contin- ual disequilibrium. Moreover, MLS is often associated with high intracranial pressure (ICP), which can be deadly [1] [2] [3]. Furthermore, studies show that MLS is significant in indicating the survival probability of patients[4]. Therefore, MLS is used as a measurement of the change of the brain symmetry and an important indicator of pathological severity. Figure 1 shows brain CT slices of the normal case and a case presenting MLS. THIS WORK IS UNDER THE SUPPORT OF THE MOE RESEARCH GRANT R-252-000-349-112. Given the significance of MLS in diagnosis, automated detection and computation of MLS using image processing techniques is an important task. A robust and efficient al- gorithm to compute MLS is an essential component of a computer-aided neurology diagnosis system. However, the task has not been given enough attention. Few works can be found in the literature that focus on MLS automated de- tection and computation. To the best of our knowledge, the only work to date is [5], which models the deformed midline as a quadratic Bezier curve. Genetic algorithm was used to minimize the summed score of each point of the deformed midline on the symmetry map. The method was proved to be effective, however, it is mainly based on the symmetry of the brain structure along the image vertical direction that may be lost if large hemorrhage or tumour exists. This shortcoming often causes its failure in the intracerebral hemorrhage (ICH) case, though ICH is one of the major causes of MLS. More- over, the use of generic algorithm makes the method time inefficient. In this paper, we develop a new method of tracing the brain midline shift in traumatic brain injury (TBI) CT images. Instead of being based on the symmetry information in the im- age, we resort to the cause of the midline shift. In TBI, hem- orrhage is the main cause of brain midline shift. Therefore, firstly, we model the relationship between the hemorrhage and the midline deformation caused by it using a linear re- gression model (H-MLS model). Secondly, using the H-MLS model, the deformed midline is predicted from the hemor- rhage detected in CT images. Finally, the predicted deformed midline is adjusted according to the visual symmetry infor- mation. Preliminary experimental results demonstrate the ef- fectiveness of the method. 2. TBI HEMORRHAGE AND MIDLINE SHIFT: A LINEAR REGRESSION MODEL (H-MLS MODEL) We define the ideal midline (iML) as the intersection of the CT scan slice and the midsagittal plane [6]. As large mass of pathology such as hemorrhage emerges, iML deforms and