مقالات پذیرفته شده در نهمین کنگره بین المللی زیست پزشکی
Classification of active and non-active multiple sclerosis plaques using mp-MRI based on Radiomics
Classification of active and non-active multiple sclerosis plaques using mp-MRI based on Radiomics
Mahmoud Mohammadi-Sadr,1Farzaneh Kashi,2,*
1. Department of Medical Physics, School of Medicine, Isfahan University of Medical Sciences, Isfahan, Iran 2. Department of Medical Physics and Engineering, School of Medicine, Tehran University of Medical Sciences, Tehran, Iran
Introduction: Multiple sclerosis (MS) is a chronic autoimmune disease characterized by the immune system attacking the central nervous system, leading to the formation of plaques that can be classified as active or non-active based on their inflammatory status. Accurate classification of these plaques is crucial for effective diagnosis, treatment planning, and monitoring disease progression, as active lesions indicate ongoing inflammation and require different management strategies compared to non-active lesions. Recent advancements in magnetic resonance imaging (MRI), particularly multiparametric MRI (mpMRI), have enhanced the visualization of MS lesions, allowing for more detailed analysis of their characteristics.
Methods: This review focuses on the application of radiomics and machine learning (ML) techniques to classify active and non-active MS plaques using mpMRI data, emphasizing the extraction of quantitative features from imaging. Radiomics involves the high-throughput extraction of numerous quantitative features from medical images, which can be analyzed alongside ML algorithms to improve classification accuracy. Various ML models, including deep learning (DL) approaches, have been employed to analyze T2-weighted MRI images, with feature selection techniques enhancing the predictive performance of these models. The review synthesizes findings from multiple studies that have utilized radiomics and ML to assess lesion activity, highlighting the importance of specific features such as texture, shape, and intensity in distinguishing between active and non-active plaques.
Results: Studies have demonstrated that radiomics-based models can achieve significant accuracy in classifying MS lesions, with some models outperforming traditional methods. For instance, a deep learning model utilizing T2-weighted MRI images achieved high sensitivity and specificity in predicting active lesions, indicating the potential of these advanced techniques in clinical practice. The integration of radiomic features, such as maximum intensity, elongation, and sphericity, has been shown to enhance the robustness of predictions, providing valuable insights into lesion characteristics. Additionally, the use of non-contrast imaging techniques has been highlighted as a safer alternative to gadolinium-based contrast agents, further supporting the feasibility of radiomics in routine clinical assessments.
Conclusion: The application of radiomics and machine learning in classifying active and non-active MS plaques from mpMRI represents a significant advancement in the field of neuroimaging and MS management. These techniques not only improve diagnostic accuracy but also inform treatment decisions and enhance patient outcomes by enabling timely interventions based on lesion activity. Future research should focus on refining these models, validating their performance across diverse populations, and integrating them into clinical workflows to optimize the management of multiple sclerosis.