Fitting large-scale structured additive regression models using Krylov subspace methods

Abstract

Fitting regression models can be challenging when regression coefficients are high-dimensional. Especially when large spatial or temporal effects need to be taken into account the limits of computational capacities of normal working stations are reached quickly. The analysis of images with several million pixels, where each pixel value can be seen as an observation on a new spatial location, represent such a situation. A Markov chain Monte Carlo (MCMC) framework for the applied statistician is presented that allows to fit models with millions of parameters with only low to moderate computational requirements. The method combines a modified sampling scheme with novel accomplishments in iterative methods for sparse linear systems. This way a solution is given that eliminates potential computational burdens such as calculating the log-determinant of massive precision matrices and sampling $ Data used in preparation of this article were obtained from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database (adni.loni.usc.edu).

Standardization of T1w/T2w Ratio Improves Detection of Tissue Damage in Multiple Sclerosis

“…Co-registration and registration to standard space was performed using a spline interpolation with FSL FLIRT (26, 27). Lesion segmentation was done semi-automatically on FLAIR using the lesion prediction algorithm [LPA (28)] as implemented in the Lesion Segmentation Toolbox version 2.0.15 (www.statistical-modelling.de/lst.html). Lesion masks were subsequently manually corrected using ITK-SNAP (29) (www.itksnap.org).…”

Characterization of normal-appearing white matter in multiple sclerosis using quantitative susceptibility mapping in conjunction with diffusion tensor imaging

“…The diffusion-weighted, multi-echo GRE and FLAIR images were registered to the T1w anatomic image using the boundary based registration tool in FreeSurfer. White matter lesion segmentation was performed using the lesion prediction algorithm as implemented in the Lesion Segmentation Toolbox (v. 2.0.13; www.statistical-modelling.de/lst.html) for SPM12 [25]. The generated lesion masks were then manually edited by an experienced neuroradiologist (F.Y., 7 years of experience) using the FLAIR images for reference.…”

SegAE: Unsupervised white matter lesion segmentation from brain MRIs using a CNN autoencoder

“…The WMHs in a total of 25 subjects were manually delineated by a neuroradiologist to be used as ground truth lesion segmentations for evaluation of the proposed method. We compared the proposed method with three state-of-the-art methods; two publicly available WMH segmentation methods, i.e., the Lesion Growth Algorithm (LGA) (Schmidt et al., 2012) and the Lesion Prediction Algorithm (LPA) (Schmidt, 2017) as implemented in the LST toolbox2 version 2.0.15, and one method developed previously for the AGES-Reykjavik data set based on an artificial neural network classifier (ANNC) (Sigurdsson et al., 2012): LGA segments WMHs from T1-w and FLAIR images. A CSF, GM and WM segmentation is first obtained from the T1-w image and combined with FLAIR image intensities for calculation of WMH belief maps.…”