ORIGINAL PAPER Automatic Detection of the Existence of Subarachnoid Hemorrhage from Clinical CT Images Yonghong Li & Jianhuang Wu & Hongwei Li & Degang Li & Xiaohua Du & Zhijun Chen & Fucang Jia & Qingmao Hu Received: 24 June 2010 / Accepted: 26 August 2010 / Published online: 9 September 2010 # Springer Science+Business Media, LLC 2010 Abstract Subarachnoid hemorrhage (SAH) is a medical emergency which can lead to death or severe disability. Misinterpretation of computed tomography (CT) in patients with SAH is a common problem. How to improve the accuracy of diagnosis is a great challenge to both the clinical physicians and medical researchers. In this paper we proposed a method for the automatic detection of SAH on clinical non-contrast head CT scans. The novelty includes approximation of the subarachnoid space in head CT using an atlas based registration, and exploration of support vector machine to the detection of SAH. The study included 60 patients with SAH and 69 normal controls from clinical hospitals. Thirty patients with SAH and 30 normal controls were used for training, while the rest were used for testing to achieve a testing sensitivity of 100% and specificity of 89.7%. The proposed algorithm might be a potential tool to screen the existence of SAH. Keywords Subarachnoid hemorrhage . Support vector machine . Non-contrast CT . Atlas based registration Introduction Stroke is the second major killer worldwide [1]. Subarachnoid hemorrhage (SAH) is the bleeding into the subarachnoid space (SAS)—the area between the arachnoid membrane and the pia mater surrounding the brain. Though SAH accounts for only about 5% of all strokes, it occurs at a fairly young age [2]. The loss of productive life years is similar to that for cerebral infarction. Over the past several decades, the incidence of other types of strokes has decreased but not that of SAH. Because early treatment of SAH is associated with improved outcomes, timely diagnosis is critical [3]. Computed tomography (CT) is still the first choice for diagnosis of SAH in clinical practice. But numerous studies document that misdiagnosis of SAH occurs approximately 25% of the time (12–50%), even in the era of ready access to cranial CT scanning [4]. As shown in [5], misinterpretation of CT in patients with SAH is a common problem. Various hemorrhage detection systems have been reported [6–10], but none of them are Y. Li : J. Wu : X. Du : Z. Chen : F. Jia : Q. Hu Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, China Y. Li : J. Wu : X. Du : Z. Chen : F. Jia : Q. Hu The Chinese University of Hong Kong, New Territories, Hong Kong Y. Li Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China Y. Li Graduate University of Chinese Academy of Sciences, Beijing, China H. Li Ningxia Medical University, Yinchuan, China D. Li The Third Affiliated Hospital of Inner Mongolia Medical College, Baotou, China Q. Hu (*) Research Center for Human-Computer Interaction, Shenzhen Institutes of Advanced Technology, University Town of Shenzhen, 1068 Xueyuan Boulevard, Shenzhen 518055, China e-mail: qm.hu@siat.ac.cn J Med Syst (2012) 36:1259–1270 DOI 10.1007/s10916-010-9587-8