مقالات پذیرفته شده در سومین کنگره بین المللی زیست پزشکی
Using machine- learning and Biological biomarker as new approaches in fMRI studies
Using machine- learning and Biological biomarker as new approaches in fMRI studies
Sina Zamani,1,*Sara Zamani,2
1. School of Medicine, Shiraz University of Medical Sciences, Shiraz, Iran 2. Department of microbiology and microbial biotechnology, faculty of life sciences and biotechnology, Shahid Beheshti University, Tehran, Iran
Introduction: Functional magnetic resonance imaging (fMRI) is a safe and non-invasive tool to estimate the brain functions by using the signal changes associated with functional brain activities.
Methods: We reviewed the papers without time limitation about using machine- learning and Biological biomarker in fMRI studies. Our research involved papers with the relevant keywords: fMRI, Machine learning and Biomarker. The data bases of Science Direct, Medline Scopus and Google Scholar were searched. The title and abstract of the papers were reviewed. Some of the papers that studied about this topic were used in this review.
Results: Advantages of fMRI
fMRI is a new procedure, which measures tiny metabolic changes, which occur in an active part of the brain using magnetic resonance imaging. Neuroscientists and physicians carry out fMRI for learning how normal, disease or injured brain performs, in addition to, for assessing the potential risk of surgery or other invasive treatment on the brain. fMRI experiments are used to determine which part of brain is active for critical functions, to assess the effects of stroke, trauma or degenerative diseases on brain function, to monitor the growth and the activity of the brain tumor regions and to plan surgery, radiotherapy or other surgical treatments of the brain.
The complexity of fMRI data and the importance of capturing spatiotemporal dependencies make it very desirable to find a level of abstraction at which inference can be performed. Graphs have the desirable property of being able to represent data at many spatial resolutions, meaning that the same mathematical models and algorithms can be applied at different scales.
Machine- learning is a new method for the analysis of the dynamic of brain networks:
There are new methods for the analysis of the dynamic of brain networks like dynamic causal modeling, independent component analysis, Bayesian statistics and machine-learning approaches. Recent advances in human neuroimaging, including functional and structural magnetic resonance imaging(fMRI/sMRI) combined with machine learning techniques, are bringing us closer to the goal of developing objective, brain-based markers of neural functions and neuropathology that underlie brain disorders.
In parallel with the rise of interest in brain networks, there has been an increase in the use and development of machine-learning techniques in neuroscience. Indeed, the high-dimensional nature of fMRI data hinders the application of many multivariate methods from classical statistics, prompting an increasing number of researchers to rely on regularization methods common in machine learning.
Biological biomarker is a major goal for many brain disorders studies:
Good biomarkers should produce high diagnostic performance in classification or prediction. Brain-based biomarkers should be meaningful and interpretable in terms of neuroscience, including prior neuroimaging studies and converging evidence from multiple sources (animal models, lesion studies). Once the classification or outcome prediction model has been developed as a neuroimaging biomarker, the model and the testing procedure should be precisely defined so that it can be properly applied to new data. Analyzing the reliability of fMRI related to different biomarkers (like hormonal assay, CBC and ESR) in different conditions such as healthy people, brain dx like CVA, meningitis, brain disorder like Parkinson and dementia that have structural change, with different analysis method like ICD and machine learning, in longitudinal study could be useful.
Conclusion: Precise characterization of dynamic connectivity is critical for understanding the intrinsic functional organization of the human brain. The mechanisms underlying brain function are not completely understood. fMRI can be used to develop and test models of brain function. Machine learning can be combined with fMRI technique to well- established mass-univariate analysis techniques. Neuroimaging biomarker allow us to develop a classification model based on brain functions and identify the regions that make the most important contributions to the classification. This classification can be done by using machine learning after establishing a reliable database. Such biomarkers have not yet made their way into clinical practice and this review serves as a starting point for biomarker discovery and validation. The investigation of brain lateralization is not only important from a neuroscientific perspective, but has also clinical implications, for instance to better assess the long-term effects of a stroke or of a neurosurgical intervention.