Magnetic Resonance Imaging (MRI) is an important paraclinical tool for diagnosing and following-up of Multiple Sclerosis (MS). The detection of MS lesions in MRI may require complementary information to filter false detections. Given that MS lesions cannot be located within cerebrospinal fluid (CSF), detection of this region is very helpful for our purpose. Although T1-weighted images are usually chosen to detect CSF regions, the gray level similarity between some MS lesions and CSF regions difficult this task. With the aim of discriminating CSF region within intracranial region, but considering aforementioned drawback, we propose a fuzzy-based algorithm that involves the regional analysis of the fuzzy information obtained from a previous local analysis. The proposed algorithm introduces location, shape and size constraints in CSF detection, and provides confidence degrees associated with the possibility of including MS lesion pixels.