Online dictionary learning for single-subject fMRI data unmixing

Bhanot, Argheesh; Meillier, Céline; Heitz, Fabrice; Harsan, Laura

doi:10.23919/eusipco.2019.8902991

Cited by 1 publication

(4 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We apply the algorithm on a single subject rs-fMRI. The motivation to work on single subject has been detailed in [16]. Benefit of integrating a high resolution (HR) anatomical atlas in the single-subject case has also been demonstrated in this previous work.…”

Section: A Semi Real Rs-fmri Datasetmentioning

confidence: 88%

“…In practice, a good initialisation of A and the presence of pure pixels (as in remote sensing applications) in each region guarantee a good joint estimation of U and A. Previous work [16] has demonstrated the importance of well-defining the spatial constraint on abundance I M (Ã) (A) to ensure an acceptable estimate of abundances and spectral or temporal signatures.…”

Section: Estimation Of the Abundance / Mixing Matrix Amentioning

confidence: 99%

“…Reproduction of figure 21 from paper [32] with the authors' permission and pending approval from A&A. of sources spatial localisation must be precisely registered to the spectral/temporal data to unmix. As for the fMRI application, in [16] the authors showed that the estimated values were sensitive to the quality of the localisation maps provided. As these maps are used to define the regularisation term I M (ãi) , their precision influences the estimation of the temporal/spectral signatures and their abundances in strongly mixed voxels: the better the precision of the localisation map, the better the estimation of the spectral/temporal information associated to each source.…”

Section: Lyαmentioning

confidence: 99%

“…In the first one, for a given pixel/voxel, different sources contribute to the mix in the sense that the spatial resolution is not fine enough to allow spatial separation of the sources. This is the case, for example, with unmixing problems in remote sensing [10] or fMRI applications [15], [16]. In this case, the mixing matrix is a matrix of proportions where for a given pixel, the sum of the contributions of each source is equal to 1.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Spatially Constrained Online Dictionary Learning for Source Separation

Bhanot

Meillier

Heitz

et al. 2021

IEEE Trans. on Image Process.

Self Cite

View full text Add to dashboard Cite

Whether in medical imaging, astronomy or remote sensing, the data are increasingly complex. In addition to the spatial dimension, the data may contain temporal or spectral information that characterises the different sources present in the image. The compromise between spatial resolution and temporal/spectral resolution is often at the expense of spatial resolution, resulting in a potentially large mixing of sources in the same pixel/voxel. Source separation methods must incorporate spatial information to estimate the contribution and signature of each source in the image. We consider the particular case where the position of the sources is approximately known thanks to external information that may come from another imaging modality or from a priori knowledge. We propose a spatially constrained dictionary learning source separation algorithm that uses e.g. high resolution segmentation map or regions of interest defined by an expert to regularise the source contribution estimation. The originality of the proposed model is the replacement of the sparsity constraint classically expressed in the form of an 1 penalty on the localisation of sources by an indicator function exploiting the external source localisation information.The model is easily adaptable to different applications by adding or modifying the constraints on the sources properties in the optimisation problem. The performance of this algorithm has been validated on synthetic and quasi-real data, before being applied on real data previously analysed by other methods of the literature in order to compare the results. To illustrate the potential of the approach, different applications have been considered, from scintigraphic data to astronomy or fMRI data.

show abstract

Section: A Semi Real Rs-fmri Datasetmentioning

confidence: 88%

Section: Estimation Of the Abundance / Mixing Matrix Amentioning

confidence: 99%

Section: Lyαmentioning

confidence: 99%