This Research Topic is part of the Machine Learning in Imaging Neurodevelopment and Neurodegeneration series:
Machine Learning in Imaging Neurodevelopment and Neurodegeneration, Volume IAdvanced neuroimaging techniques, such as Magnetic Resonance Imaging (MRI) and Position Emission Tomography (PET), have provided great insights into the anatomy and function of the human brain in health and pathology. These techniques have been used to compare normal brains to brains affected by a disease in order to identify possible biomarkers for early diagnosis, treatment planning, and monitoring of disease progression. Extensive computational neuroimaging analyses, using statistical hypothesis testing approaches, have revealed structural, functional, and connectomic brain abnormalities and alterations in a number of neurological or psychiatric disorders. However, so far, the group-level findings of these studies have had quite little impact on clinical practice; clinical decisions on individuals still depend on traditional diagnostic and prognostic methods.
During the recent few years, there has been growing interest within the neuroimaging community in the use of machine learning approaches. These extract information hidden in an example data set and make accurate predictions on future data using the learned patterns, to conduct inference about individual samples. Machine learning techniques are powerful tools both for group studies in a multivariate fashion and for making individual-level prediction, as they are instrumental for mining big data and are discriminative against high noise levels. Numerous studies have shown the significant scientific promise of machine learning techniques to predict disease outcomes and impacts on the brain and to help clinical diagnosis and treatment prognosis for individual subjects. There are, however, still challenges for the translational implementation of the results in clinical settings. For example, training highly accurate learning machines requires large volumes of data, while most existing studies are carried out on only small samples; it is hence not possible to draw definite conclusions. Implementation, interpretation, and visualization of machine learning studies remains a difficult task, especially for most clinical units that have little data science expertise and limited computing resources. Integrative methods for multiple imaging modalities are also desired to improve classification or regression accuracy using more complex data.
This Research Topic intends to help advance the scientific research within the field of machine learning in neuroimaging so as to realize the potential of machine learning to translate neuroimaging data into tools that can directly aid clinicians in diagnosis and therapy for neurodevelopmental and neurodegenerative disorders, such as autism, schizophrenia, Alzheimer’s disease, and Parkinson’s disease, amongst others. For such purposes, this Research Topic welcomes original research and review articles that highlight major trends and challenges in the field and aim to identify new cutting-edge machine learning approaches and their applications in imaging neurodevelopment and neurodegeneration. Potential topics of interests include, but are not limited to:
• Machine learning analyses using large-scale neuroimaging data
• Computer-aided diagnosis/prognosis for neurodevelopmental and neurodegenerative disorders
• Analytical tools and software for computer-aided diagnosis/prognosis
• Multimodal fusion for outcome prediction, diagnosis/prognosis, image analysis
• Computer-aided lesion detection for neurodegeneration
• Machine learning analyses combining neuroimaging and genetics