An artificial intelligence approach for segmenting and classifying brain lesions caused by stroke

ABSTRACTBrain injuries caused by strokes are one of the leading causes of disability worldwide. Current procedures require a specialised physician to analyse MRI images before diagnosing and deciding on the specific treatment. However, the procedure can be costly and time-consuming. Artificial intelligence techniques are becoming a game-changer for analysing MRI images. This work proposes an end-to-end approach in three stages: Pre-processing techniques for normalising the images to the standard MNI space, as well as inhomogeneities and bias corrections; lesion segmentation using a CNN network, trained for cerebrovascular accidents and feature extraction; and, classification for determining the vascular territory within which the lesion occurred. A CLCI-Net was used for stroke segmentation. Four Deep Learning (DL) and four Shallow Machine Learning (ML) network architectures were evaluated to assess the strokes’ territory localisation. All models’ architectures were designed, analysed, and compared based on their performance scores, reaching an accuracy of 84% with the DL models and 95% with the Shallow ML models. The proposed methodology may be helpful for rapid and accurate stroke assessment for an acute treatment to minimise patient complications.KEYWORDS: Artificial intelligencelesion segmentationMRI preprocessingstroke assessment AcknowledgementWe would like to thank Carlos Jimenez, Alisson Constantine and Edwin Valarezo for their helpful contribution in perfecting the text and debugging the scripts.Disclosure statementAll authors have seen and agreed with the content of the manuscript; there is no financial interest to report, or declare any conflicts of interest, neither there are funding sources involved. We certify that the submission is original work and is not under review at any other publication.Additional informationNotes on contributorsRoberto MenaRoberto Alejandro Mena is a graduate student in Computer Science Engineering from Escuela Superior Politécnica del Litoral – ESPOL University. Throughout his career, he has played a leading role as a data analyst in various research projects, mainly centered on system development for magnetic resonance imaging (MRI) processing and visualization.Enrique PelaezDr. Enrique Peláez earned his Ph.D. in Computer Engineering from the University of South Carolina, USA, in 1994. Currently, he is a Professor at ESPOL University where he leads the AI research in Computational Intelligence. Over recent years, Dr. Pelaez has been engaged in applied research on Parkinson's Disease, leveraging machine and deep learning techniques. His academic contributions showcased in leading publications and forums, with papers presented in several conferences and symposia. Dr. Pelaez's work has been published in journals, including the IEEE and Nature Communications. His research topics encompass EEG signal classification, deep learning for medical imaging, and behavioral signal processing using AI.Francis LoayzaDr. Francis Loayza serves as a Full Professor in the Mechanical Engineering Department (FIMCP) at ESPOL University. He was conferred with a Ph.D. in Neurosciences from the University of Navarra, Spain, in 2010. With a deep-rooted expertise in image data analysis, Dr. Loayza utilizes statistical methods such as functional Magnetic Resonance Imaging and Voxel Based Morphometry. Furthermore, his application of machine and deep learning methodologies is contributing to the growing knowledge of neurodegenerative disorders.Alex MacasAlex Macas Alcocer is a graduate student in Computer Science Engineering from Escuela Superior Politécnica del Litoral – ESPOL University. He has been working as a Data Scientist, analyzing magnetic resonance images using artificial intelligence techniques, as well as in web development.Heydy Franco-MaldonadoDr. Heydy Franco Maldonado is a distinguished specialist in Imagenología, trained at the Universidad de Cuenca. She further advanced her specialization in Magnetic Resonance at UNAM, Mexico, then pursued a diploma in Breast Pathology Imaging from the Universidad de Barcelona. Presently, she serves as a Medical Radiologist at both Hospital Luis Vernaza in Guayaquil and SOLCA Guayaquil - Ecuador. Besides her clinical roles, Dr. Maldonado is an active member of ESPOL's Artificial Intelligence Research Group and coordinates the postgraduate program in Imagenología at Universidad Espíritu Santo in conjunction with Hospital Luis Vernaza. Additionally, she is a recognized speaker for Bayer.