Orzan Filip Ștefania D Iancu Laura Dioșan Zoltán Bálint ^*

• Babeș-Bolyai University, Cluj-Napoca, Romania

Conventionally, Magnetic Resonance Imaging (MRI) is used to detect and diagnose multiple sclerosis. The lumbar puncture, a very invasive method is often combined with MRI, to validate the diagnosis of Multiple Sclerosis (MS). In addition, to follow the evolution of the disease and the effectiveness of the treatment used, MRI is periodically repeated. Research in the last decade has been focused on the use of artificial intelligence (AI) together with radiomics, with direct application in medical image processing, medical diagnosis and treatment processes. The variation in models underscores a fragmented research landscape. There is no consensus on the optimal model for MS lesion segmentation or classification, where models like U-Net, Support Vector Machine, Random Forest, and k-nearest neighbors are frequently used. Performance evaluation varies, commonly relying on metrics such as Accuracy, Dice score, and Sensitivity. While some models demonstrate robustness across multi-center datasets, most studies lack clinical scenario validation. Clinical trials are thus a key area for future research. In conclusion, our article provides an insight into the contemporary landscape of decision support tools that leverage texture analysis and artificial intelligence for the analysis and monitoring of emerging multiple sclerosis lesions in MRI images.