Artificial intelligence in neuroimaging: four challenges to improve interpretation of brain images

Neuroimaging is a growing interdisciplinary field at the intersection of image processing, computer science, physics, statistics and neurosciences. It gathers research on imaging techniques used to investigate mental activities in healthy as well as diseased situations. Two classes of images are acquired: anatomical or structural images, which give information about brain structures (e.g. using magnetic resonance imaging (MRI) or computed tomography), blood vessels (e.g. using MR angiography) or brain connectivity (e.g. using MR diffusion imaging); and functional images, which give information about the neuronal populations involved in specific experimental conditions. Several methods are available to form such images. For functional brain imaging, changes in neuronal population activity, provoked by sensory stimulations or during cognitive tasks, are measured directly, by the means of evoked electric (ERP: event-related potentials) or magnetic activity (magneto-encephalography: MEG) or indirectly, via the metabolic and hemodynamic responses induced by the neuronal activation (functional MRI: fMRI and positron emission tomography: PET). Each method reflects a specific aspect of the neuronal activity and produces data with a particular spatial and temporal resolution. Neuroimaging with its palette of methods, especially with non-invasive fMRI, offers new windows on the brain and has a considerable impact on neurosciences as a whole. Interestingly, some connexions between artificial intelligence (AI) and the couple neuroimaging/neurosciences can be underlined. As mentioned by [12] in p. 73: ‘‘There are two complementary views of AI: one as an engineering discipline concerned with the creation of intelligent machines, the other as an empirical science concerned with the computational modelling of human intelligence’’. Clearly, the latter view, cognitive sciences oriented, enters directly in resonance with neurosciences. The open problem of understanding how distributed systems of brain areas work altogether, how mental capabilities emerge from the neuronal networks or how knowledge and its representation by the brain are related, can be attacked in computational terms [12] or information processing terms [1] by the two disciplines. It is also interesting to relate the engineering view of AI with neuroimaging. In practice, the huge development of neuroimaging poses unique challenges in term of Artificial Intelligence in Medicine 30 (2004) 91–95

[1]  Robert A. Wilson,et al.  Book Reviews: The MIT Encyclopedia of the Cognitive Sciences , 2000, CL.

[2]  David C. Van Essen,et al.  Windows on the brain: the emerging role of atlases and databases in neuroscience , 2002, Current Opinion in Neurobiology.

[3]  Balakrishnan Chandrasekaran,et al.  What are ontologies, and why do we need them? , 1999, IEEE Intell. Syst..

[4]  R. Kötter,et al.  Neuroscience databases: tools for exploring brain structure-function relationships. , 2001, Philosophical transactions of the Royal Society of London. Series B, Biological sciences.

[5]  R. Kikinis,et al.  An Automated Registration Algorithm for Measuring MRI Subcortical Brain Structures , 1997, NeuroImage.

[6]  John R. Josephson,et al.  What Are They? Why Do We Need Them? , 1999 .

[7]  R. Woods,et al.  Mathematical/computational challenges in creating deformable and probabilistic atlases of the human brain , 2000, Human brain mapping.

[8]  Ray L. Somorjai,et al.  Exploratory data analysis in functional neuroimaging , 2002, Artif. Intell. Medicine.

[9]  A. Toga.  Neuroimage databases: The good, the bad and the ugly , 2002, Nature Reviews Neuroscience.

[10]  N Lange.  Statistical approaches to human brain mapping by functional magnetic resonance imaging. , 1996, Statistics in medicine.

[11]  Richard Lepage,et al.  Knowledge-Based Image Understanding Systems: A Survey , 1997, Comput. Vis. Image Underst..

[12]  Timothy F. Cootes,et al.  Active Shape Models-Their Training and Application , 1995, Comput. Vis. Image Underst..

[13]  Alan L. Rector,et al.  The Interface between Information, Terminology, and Inference Models , 2001, MedInfo.

[14]  Pierre Hellier,et al.  Coupling dense and landmark-based approaches for nonrigid registration , 2003, IEEE Transactions on Medical Imaging.

[15]  J B Woodward,et al.  The Functional Magnetic Resonance Imaging Data Center (fMRIDC): the challenges and rewards of large-scale databasing of neuroimaging studies. , 2001, Philosophical transactions of the Royal Society of London. Series B, Biological sciences.

[16]  D. Stuss,et al.  Cognitive neuroscience. , 1993, Current opinion in neurobiology.

[17]  B. Horwitz,et al.  Predicting human functional maps with neural net modeling , 1999, Human brain mapping.

[18]  F Makedon,et al.  Statistical Methods in Medical Research Data Mining in Brain Imaging , 2022 .

[19]  Karl J. Friston,et al.  Statistical parametric maps in functional imaging: A general linear approach , 1994 .

[20]  Jacques Gautrais,et al.  SpikeNET: A simulator for modeling large networks of integrate and fire neurons , 1999, Neurocomputing.

[21]  Nigel H. Goddard,et al.  Towards NeuroML: model description methods for collaborative modelling in neuroscience. , 2001, Philosophical transactions of the Royal Society of London. Series B, Biological sciences.

[22]  M. Musen.  Scalable Software Architectures for Decision Support , 1999, Methods of Information in Medicine.

[23]  Christopher J. Taylor,et al.  A cooperative framework for segmentation of MRI brain scans , 2000, Artif. Intell. Medicine.

[24]  D. V. van Essen,et al.  Windows on the brain: the emerging role of atlases and databases in neuroscience , 2002, Current Opinion in Neurobiology.