Journal: NeuroImage : Clinical

Article Title: 3D scattering transforms for disease classification in neuroimaging

doi: 10.1016/j.nicl.2017.02.004

Figure Lengend Snippet: A scheme of the main steps of the 3D scattering paradigm. A procedure to regress out the effect of age on the MRI scans, is performed in the beginning. The 3D scattering transform is subsequently applied to each MRI subject resulting in the scattering representation of the MRI data. Cross-validation is performed in all the experiments where data are split as follows: 60% training + 20% validation + 20% test. Supervised learning is performed using a linear support vector machines (SVM) classifier. The learning algorithm used in the last experiment, which is a semi-supervised learning experiment performed on the ADNI dataset, is a Laplacian SVM. Not shown is the introduced method of visualization of evidence for or against a class, on both the group and individual level, see Section 2.5 .

Article Snippet: Code for Laplacian SVM used in our experiments is available at http://sourceforge.net/projects/lapsvmp/

Techniques: Biomarker Discovery, Plasmid Preparation

Journal: NeuroImage : Clinical

Article Title: 3D scattering transforms for disease classification in neuroimaging

doi: 10.1016/j.nicl.2017.02.004

Figure Lengend Snippet: Results of the MCI experiments. The first experiment is a supervised learning experiment with two labels, pMCI and sMCI. Using a linear SVM on top of the scattering representation leads to a superior classification accuracy compared to state-of-the-art by Moradi et al. (2015) . The second experiment is an SSL experiment where the training data consist of unlabeled MCI MRI data in addition to data belonging to the pMCI and sMCI labels. Using a Laplacian SVM trained in the primal on top of the scattering representation leads to a higher accuracy than state-of-the-art by Moradi et al. (2015) .

Article Snippet: Code for Laplacian SVM used in our experiments is available at http://sourceforge.net/projects/lapsvmp/

Techniques:

Journal: Translational Vision Science & Technology

Article Title: Automated Focal Plane Merging From a Stack of Gonioscopic Photographs Using a Focus-Stacking Algorithm

doi: 10.1167/tvst.11.4.22

Figure Lengend Snippet: The example of image processing for objectively analyzing the sharpness of an image. ( A ) The pair of original gonio-photographs. ( B ) The pair of A images after conversion to grayscale. ( C ) The pair of results after the Laplacian filter processing with B image (|Δ I |). The focus-stacked image exhibited more edges ( white part ) than the best-focused one visually.

Article Snippet: The energy of the Laplacian, which is a measure of image sharpness was analyzed with the Laplacian filter (Nidek Co., Gamagori, Japan) and compared for each set.

Techniques:

Journal: Translational Vision Science & Technology

Article Title: Automated Focal Plane Merging From a Stack of Gonioscopic Photographs Using a Focus-Stacking Algorithm

doi: 10.1167/tvst.11.4.22

Figure Lengend Snippet: Comparisons Between the Assessments for the Gonio-Photos With and Without the Focus-Stacking Processing

Article Snippet: The energy of the Laplacian, which is a measure of image sharpness was analyzed with the Laplacian filter (Nidek Co., Gamagori, Japan) and compared for each set.

Techniques:

Journal: Translational Vision Science & Technology

Article Title: Automated Focal Plane Merging From a Stack of Gonioscopic Photographs Using a Focus-Stacking Algorithm

doi: 10.1167/tvst.11.4.22

Figure Lengend Snippet: The representative pairs of focus-stacked and best-focused images, in which the focal plane merging technique achieved the significant deepening of the depth of field and improvement of informativeness and increased energy of the Laplacian. The salient areas of interest are indicated by white arrows . ( A ) The images show tall tent-shaped peripheral anterior synechiae (PAS). The focus-stacked image has clearer PAS boundaries and more detail about iris damage. ( B ) The pair of images is an example of pseudoexfoliation glaucoma. The shape and shade of angle pigmentation can be observed in the focus-stacked image compared with the best-focused one. ( C ) The pair of images is an example after trabeculectomy. Compared with the best-focused image, the focus-stacked image clearly shows the peripheral iridectomy scar and iris damage. ( D ) As an example, after the EX-PRESS Glaucoma Filtration Device (Alcon Laboratories, Fort Worth, TX) procedure, the DOF in the focus-stacked image is deep enough to be observed the angle from the tip of the device to the insertion area compared with that of the best-focused one. ( E ) The pair of images is an example after the Ahmed glaucoma valve implantation (Model FP-7; New World Medical, Rancho Cucamonga, CA) into the anterior chamber. Compared with the best-focused image, the focus-stacked one has a wider range of focus, and it can be observed well from the tip of the device to the insertion part. Moreover, its lumen can be seen.

Article Snippet: The energy of the Laplacian, which is a measure of image sharpness was analyzed with the Laplacian filter (Nidek Co., Gamagori, Japan) and compared for each set.

Techniques: Filtration