Abstract—This study proposes a technique for automated detection and diagnosis of stroke lesions based on diffusion- weighted imaging (DWI). The technique consists of several stages which are pre-processing, segmentation, feature extraction, and classification. The proposed analytical framework of this study is based on Fuzzy C-Means (FCM) segmentation, statistical parameters for features extraction and rule-based classification. The three-dimensional (3D) view is developed to enable observing directions of the gained 3D structure along the three axes. The segmentation results have been validated by using Jaccard and Dice indices, false positive rate (FPR), and false negative rate (FNR). The results for Jaccard, Dice, FPR and FNR of acute stroke are 0.7, 0.84, 0.049 and 0.205, respectively. The accuracy for acute stroke is 90% and chronic stroke is 70%, while the sensitivity and the specificity is 84.38% and 83.33%, respectively. Index Terms—Diffusion-Weighted Imaging (DWI), Segmentation, Fuzzy C-Means, Three-Dimensional Reconstruction I. INTRODUCTION TROKE is a major health burden in Malaysia as well as worldwide. It is also one of the top five leading causes of death in Malaysia [1]. National Stroke Association of Malaysia (NASAM) stated that one of six people worldwide will suffer from the stroke in their lifetime and it is the third leading cause of adult disability [2]. Stroke is a clinical symptom that happens when the blood vessel is blocked or burst due to a blood clot. All the oxygen and nutrients supply will be cut off causing a syndrome characterize by rapidly developing symptoms or sign of focal neurologic Manuscript received January 23, 2017. N. Mohd Saad is with the Center for Robotics & Industrial Automation (CeRIA) and Faculty of Electronics and Computer Engineering, Universiti Teknikal Malaysia Melaka, Malaysia. (email:norhashimah@utem.edu.my). N. S. M. Noor is with the Center for Robotics & Industrial Automation (CeRIA) and Faculty of Electronics and Computer Engineering, Universiti Teknikal Malaysia Melaka, Malaysia. (email:m021610012@student.utem.edu.my). A. R. Abdullah is an Associate Professor and a coordinator with the Center for Robotics & Industrial Automation (CeRIA) and Faculty of Electrical Engineering, Universiti Teknikal Malaysia Melaka, Malaysia. (email:abdulr@utem.edu.my). Sobri Muda is a Professor with the Radiology Department, Fakulti Perubatan dan Sains Kemanusiaan, Universiti Putra Malaysia, Malaysia. (email: asobri@upm.edu.my). A. F. Muda is with the Faculty for Electronics and Computer Engineering, Universiti Teknikal Malaysia Melaka, Malaysia. (email:fateehaa@yahoo.com). N. N. S. Abdul Rahman is with the Center for Robotics & Industrial Automation (CeRIA) and Faculty of Electronics and Computer Engineering, Universiti Teknikal Malaysia Melaka, Malaysia. (email:mo21610013@student.utem.edu.my). dysfunction due to a vascular cause [3]. Urgent treatment is needed to obtain less debilitating of stroke and for better recovery. Detection and diagnosis of brain stroke are the keys to implementing successful therapy and treatment planning. However, the diagnosis is a very challenging task and can only be performed by professional neuroradiologists [4]. Computer Aided Diagnosis (CAD) is needed to assist radiologists in interpreting medical images by using dedicated computer systems to provide a second opinion for clinical validation [5]. Studies on CAD systems and technology show that CAD can help to improve the diagnostic accuracy of radiologists. Today, diffusion- weighted imaging (DWI) of the brain has become common in emergency setting, such that it is considered as a key sequence for stroke imaging, and essential in the proper evaluation of most pathologic conditions [6]. This imaging technique provides high lesion contrast compared to other magnetic resonance image (MRI) sequences [7]. DWI measures diffusion of water molecules within the tissue structure on a pixel basis. Tissues in which water diffusion is reduced can, therefore, be readily detected as a hyperintense on DWI images, which has become the hallmark of detection of recent stroke lesion. Manual detection and segmentation of different tissues are very time-consuming compared to computer-aided techniques [8]. However, automatic image segmentation is still a troublesome issue with larger part of utilization due to the presence of noise and intensity inhomogeneity. Fuzzy C-Means (FCM) is a popular technique proposed by many researchers for segmentation of medical images [8]. However, because of noises and intensity inhomogeneity, FCM technique fails in producing accurate results. Although, the original FCM algorithm provides superior results for segmenting noise free images, but it’s accuracy is low when the image is corrupted with noises. Therefore, FCM algorithm is modified to improve the segmentation results [9, 10, 11]. The purpose of this research is to develop an automatic method to detect and segment the brain stroke lesions from DWI images and to visualize the brain scan in three- dimensional (3D) view. The ability of MRI scan in converting the 3D brain images into multiple two- dimensional (2D) image slices in a various way may lead to certain necessary information being lost. This problem has attracted researchers to continuously seek towards the 3D reconstruction. The availability of 3D imaging has made disease diagnosis and surgical planning successful, which resulted in saving precious patient lives [12]. The proposed analytical framework of this study is based Automated Stroke Lesion Detection and Diagnosis System N. Mohd Saad, N. S. M. Noor, A.R. Abdullah, Sobri Muda, A. F. Muda and N. N. S. Abdul Rahman S Proceedings of the International MultiConference of Engineers and Computer Scientists 2017 Vol I, IMECS 2017, March 15 - 17, 2017, Hong Kong ISBN: 978-988-14047-3-2 ISSN: 2078-0958 (Print); ISSN: 2078-0966 (Online) IMECS 2017