Magnus Ulfarsson

University of Iceland | HI · Faculty of Electrical and Computer Engineering

In this letter, we propose a method using a three dimensional convolutional neural network (3-D-CNN) to fuse together multispectral (MS) and hyperspectral (HS) images to obtain a high resolution hyperspectral image. Dimensionality reduction of the hyperspectral image is performed prior to fusion in order to significantly reduce the computational ti...

In this paper, we present a deep learning based method for blind hyperspectral unmixing in the form of a neural network autoencoder. We show that the linear mixture model implicitly puts certain architectural constraints on the network, and it effectively performs blind hyperspectral unmixing. Several different architectural configurations of both...

Deep learning has shown to be a powerful tool and has heavily impacted the data-intensive field of remote sensing. As a result, the number of published deep learning-based spectral unmixing techniques is proliferating. Blind hyperspectral unmixing (HU) is the process of resolving the measured spectrum of a pixel into a combination of a set of spect...

Image fusion is utilized in remote sensing due to the limitation of the imaging sensor and the high cost of simultaneously acquiring high spatial and spectral resolution images. Optical remote sensing imaging systems usually provide images of high spatial resolution but low spectral resolution and vice versa. Therefore, fusing those images to obtai...

Mendelian Randomization studies indicate that BMI contributes to various diseases, but it’s unclear if this is entirely mediated by BMI itself. This study examines whether disease risk from BMI-associated sequence variants is mediated through BMI or other mechanisms, using data from Iceland and the UK Biobank. The associations of BMI genetic risk s...

Importance Understanding of the genetics of accessory atrioventricular pathways (APs) and affiliated arrhythmias is limited. Objective To investigate the genetics of APs and affiliated arrhythmias. Design, Setting, and Participants This was a genome-wide association study (GWAS) of APs, defined by International Classification of Diseases ( ICD )...

Sentinel-2 gervihnattaparið tekur myndir á 13 róftíðniböndum með þremur mismunandi upplausnum (10, 20, 60 m). Skerpingaraðferðir eru notaðar til að bæta gæði mynda með lægri upplausn. Hefðbundnar skerpingaraðferðir byggja á líkanagerð og regluðum kostnaðarföllum, en það getur verið erfitt að stilla þær fyrir ólíkar myndir. Framfarir í vélnámi og ta...

We report a multi-ancestry genome-wide association study on liver cirrhosis and its associated endophenotypes, alanine aminotransferase (ALT) and γ-glutamyl transferase. Using data from 12 cohorts, including 18,265 cases with cirrhosis, 1,782,047 controls, up to 1 million individuals with liver function tests and a validation cohort of 21,689 cases...

A comprehensive exploration of the transformer and dual receiver system-based direction-of-arrival (DOA) estimation is presented in the context of sea surface scattering, particularly under varying sea conditions. A bidirectional encoder representation from the transformer (BERT) with a physics-based loss function is utilized to process two individ...

Clonal hematopoiesis (CH) arises when a substantial proportion of mature blood cells is derived from a single hematopoietic stem cell lineage. Using whole-genome sequencing of 45,510 Icelandic and 130,709 UK Biobank participants combined with a mutational barcode method, we identified 16,306 people with CH. Prevalence approaches 50% in elderly part...

Multispectral remote sensing images are often have band-dependent image resolution due to cost and technical limitations. To address this, we developed a method that sharpens low-resolution (LR) images using high-resolution (HR) images. In this paper, we propose a novel unsupervised deep learning (DL) approach that involves unrolling an iterative a...

High-throughput proteomics platforms measuring thousands of proteins in plasma combined with genomic and phenotypic information have the power to bridge the gap between the genome and diseases. Here we performed association studies of Olink Explore 3072 data generated by the UK Biobank Pharma Proteomics Project¹ on plasma samples from more than 50,...

Importance: Whether protein risk scores derived from a single plasma sample could be useful for risk assessment for atherosclerotic cardiovascular disease (ASCVD), in conjunction with clinical risk factors and polygenic risk scores, is uncertain. Objective: To develop protein risk scores for ASCVD risk prediction and compare them to clinical ris...

Background: The TNM system is used to assess prognosis after colorectal cancer (CRC) diagnosis. Other prognostic factors reported include histopathological assessments of the tumour, tumour mutations and proteins in the blood. As some of these factors are strongly correlated, it is important to evaluate the independent effects they may have on sur...

In hyperspectral unmixing (HU), spectral variability in hyperspectral images (HSIs) is a major challenge which has received a lot of attention over the last few years. Here, we propose a method utilizing a generative adversarial network (GAN) for creating synthetic HSIs having a controllable degree of realistic spectral variability from existing HS...

This paper proposes a new loss function to train a convolutional neural network (CNN) for multispectral and hyperspectral (MS-HS) image fusion. The loss function is based on the relative dimensionless global error synthesis (ERGAS), where we exchange the mean squared error (MSE) for its unbiased estimate using Stein's risk unbiased estimate (SURE)....

Background Persistent symptoms are common after SARS-CoV-2 infection but correlation with objective measures is unclear. Methods We invited all 3098 adults who tested SARS-CoV-2 positive in Iceland before October 2020 to the deCODE Health Study. We compared multiple symptoms and physical measures between 1706 Icelanders with confirmed prior infect...

The genetic basis of the human vocal system is largely unknown, as are the sequence variants that give rise to individual differences in voice and speech. Here, we couple data on diversity in the sequence of the genome with voice and vowel acoustics in speech recordings from 12,901 Icelanders. We show how voice pitch and vowel acoustics vary across...

Microsatellites and drones are often equipped with digital cameras whose sensing system is based on color filter arrays (CFAs), which define a pattern of color filter overlaid over the focal plane. Recent commercial cameras have started implementing RGBW patterns, which include some filters with a wideband spectral response together with the more c...

Intracranial volume, measured through magnetic resonance imaging and/or estimated from head circumference, is heritable and correlates with cognitive traits and several neurological disorders. We performed a genome-wide association study (GWAS) meta-analysis of intracranial volume (N = 79,174) and found 64 associating sequence variants explaining 5...

Nonalcoholic fatty liver (NAFL) and its sequelae are growing health problems. We performed a genome-wide association study of NAFL, cirrhosis and hepatocellular carcinoma, and integrated the findings with expression and proteomic data. For NAFL, we utilized 9,491 clinical cases and proton density fat fraction extracted from 36,116 liver magnetic re...

Detailed knowledge of how diversity in the sequence of the human genome affects phenotypic diversity depends on a comprehensive and reliable characterization of both sequences and phenotypic variation. Over the past decade, insights into this relationship have been obtained from whole-exome sequencing or whole-genome sequencing of large cohorts wit...

Recent advances in deep learning (DL) reveal that the structure of a convolutional neural network (CNN) is a good image prior (called deep image prior (DIP)), bridging the model-based and DL-based methods in image restoration. However, optimizing a DIP-based CNN is prone to over-fitting leading to a poorly reconstructed image. This paper derives a...

Spectral and spatial correlation in hyperspectral images (HSIs) can be exploited in HSI processing because it directly induces a sparse and low-rank prior via linear transformations. Researchers have used the sparse and low-rank prior as an image prior for HSI restoration, such as denois-ing, deblurring, and super-resolution. This paper proposes a...

The Copernicus Sentinel-2 (S2) constellation comprises of two satellites in a sun-synchronous orbit. The S2 sensors have three spatial resolutions: 10, 20, and 60 m. The Landsat 8 (L8) satellite has sensors that provide seasonal coverage at spatial resolutions of 15, 30, and 60 m. Many remote sensing applications require the spatial resolutions of...

The classification of hyperspectral images (HSIs) is an essential application of remote sensing and it is addressed by numerous publications every year. A large body of these papers present new classification algorithms and benchmark them against established methods on public hyperspectral datasets. The metadata contained in these research papers (...

Persistent symptoms are common after SARS-CoV-2 infection but the correlation with objective measures is unclear. We utilized the deCODE Health Study to compare multiple symptoms and physical measures between 1,721 Icelanders with prior SARS-CoV-2 infection (cases) and 546 contemporary and 13,842 historical controls. Cases participated in the study...

Genomic copy number variants (CNVs) are associated with a high risk of neurodevelopmental disorders. A growing body of genetic studies suggests that these high-risk genetic variants converge in common molecular pathways, and that common pathways also exist across clinically distinct disorders, such as schizophrenia and autism spectrum disorder. A k...

High-throughput proteomics platforms measuring thousands of proteins in blood combined with genomic information have the power to bridge the gap between the genome and diseases and in that capture some of the environmental contributions to their risk and pathogenesis. Although such methods have already demonstrated their utility, the validation of...

This paper proposes a denoising method based on sparse spectral-spatial and low-rank representations (SSSLRR) using 3-D orthogonal transform (3-DOT). SSSLRR can be effectively used to remove Gaussian and mixed noise. SSSLRR uses 3-DOT to decompose noisy HSI to sparse transform coefficients. 3-D discrete orthogonal wavelet transform (3-D DWT) is a r...

The Enhancing NeuroImaging Genetics through Meta-Analysis copy number variant (ENIGMA-CNV) and 22q11.2 Deletion Syndrome Working Groups (22q-ENIGMA WGs) were created to gain insight into the involvement of genetic factors in human brain development and related cognitive, psychiatric and behavioral manifestations. To that end, the ENIGMA-CNV WG has...

Spectral variability in hyperspectral images (HSIs) has received lot of attention over the last years, especially in the field of hyperspectral unmixing (HU) where it is a major issue. In this letter, we propose a method utilizing a variational autoencoder (VAE) for creating synthetic HSIs having controllable degree of spectral variability from exi...

The plasma proteome can help bridge the gap between the genome and diseases. Here we describe genome-wide association studies (GWASs) of plasma protein levels measured with 4,907 aptamers in 35,559 Icelanders. We found 18,084 associations between sequence variants and levels of proteins in plasma (protein quantitative trait loci; pQTL), of which 19...

We describe the analysis of whole genome sequencing (WGS) of 150,119 individuals from the UK biobank (UKB). This yielded a set of high quality variants, including 585,040,410 SNPs, representing 7.0% of all possible human SNPs, and 58,707,036 indels. The large set of variants allows us to characterize selection based on sequence variation within a p...

The Sentinel-2 (S2) constellation provides multispectral images at 10 m, 20 m, and 60 m resolution bands. Obtaining all bands at 10 m resolution would benefit many applications. Recently, many model-based and deep learning (DL)-based sharpening methods have been proposed. However, the downside of those methods is that the DL-based methods need to b...

This paper proposes a novel method for sharpening the 20 m bands of the multispectral images acquired by the Sentinel-2 (S2) constellation. We formulate the S2 sharpening as an inverse problem and solve it using an unsupervised convolutional neural network (CNN), called S2UCNN. The proposed method extends the deep image prior provided by a CNN stru...

Predicting all-cause mortality risk is challenging and requires extensive medical data. Recently, large-scale proteomics datasets have proven useful for predicting health-related outcomes. Here, we use measurements of levels of 4,684 plasma proteins in 22,913 Icelanders to develop all-cause mortality predictors both for short- and long-term risk. T...

In the context of earth observation and remote sensing, super-resolution aims to enhance the resolution of a captured image by upscaling and enhancing its details. In recent years, numerous methods for super-resolution of Sentinel-2 (S2) multispectral images have been suggested. Most of those methods depend on various tuning parameters that affect...

Taking insights from the remote sensing field, in the recent decade, the advantages of hyperspectral imaging technology has been exploited for painting conservation. The estimation of pigment proportion in painting is a challenge that is due to the highly mixed nature of the paint layers, usually requiring to take into account the law of mixing col...

Low-frequency 1q21.1 distal deletion and duplication copy number variant (CNV) carriers are predisposed to multiple neurodevelopmental disorders, including schizophrenia, autism and intellectual disability. Human carriers display a high prevalence of micro- and macrocephaly in deletion and duplication carriers, respectively. The underlying brain st...

Background Copy-number variations at the 15q11.2 BP1-BP2 locus are present in 0.5 to 1.0% of the population, and the deletion is associated with several neurodevelopmental disorders. Previously, we showed a reciprocal effect of 15q11.2 copy-number variation on fractional anisotropy, with widespread increases in deletion carriers. We aim to expand t...

The Enhancing NeuroImaging Genetics through Meta‐Analysis copy number variant (ENIGMA‐CNV) and 22q11.2 Deletion Syndrome Working Groups (22q‐ENIGMA WGs) were created to gain insight into the involvement of genetic factors in human brain development and related cognitive, psychiatric and behavioral manifestations. To that end, the ENIGMA‐CNV WG has...

Low-frequency 1q21.1 distal deletion and duplication copy number variant (CNV) carriers are predisposed to multiple neurodevelopmental disorders including schizophrenia, autism and intellectual disability. Human carriers display a high prevalence of micro- and macrocephaly in deletion and duplication carriers, respectively. The underlying brain str...

The Enhancing Neuroimaging Genetics through Meta-Analysis copy number variant (ENIGMA-CNV) and 22q11.2 Deletion Syndrome Working Groups (22q-ENIGMA WGs) were created to gain insight into the involvement of genetic factors in human brain development and related cognitive, psychiatric and behavioral manifestations. To that end, the ENIGMA-CNV WG has...

Pansharpening refers to the fusion of a multispectral (MS) image and panchromatic (PAN) data aimed at generating an outcome with the same spatial resolution of the PAN data and the spectral resolution of the MS image. In the last 30 years, several approaches to deal with this issue have been proposed. However, the reproducibility of these methods i...

Olfactory receptor (OR) genes in humans form a special class characterized by unusually high DNA sequence diversity, which should give rise to differences in perception and behavior. In the largest genome-wide association study to date based on olfactory testing, we investigated odor perception and naming with smell tasks performed by 9,122 Iceland...

Background: Copy-number variations at the 15q11.2 BP1-BP2 locus are present in 0.5 to 1.0% of the population, and the deletion is associated with a range of neurodevelopmental disorders. Previously, we showed a reciprocal effect of 15q11.2 copy-number variation on fractional anisotropy, with widespread increases in deletion carriers. We aim to repl...

Supplementary Material for Semi-Supervised Mixture of Factor Analyzers Feature Extraction for Hyperspectral Images

This paper presents two hyperspectral image (HSI) denoising methods, mixtures of factor analyzers (MFA) and wavelet-based MFA (WMFA). MFA uses a Gaussian mixture model to segment the original HSI into different parts, where each part follows Gaussian distribution and then utilizes a factor ana-lyzer to get a low-rank factor loading matrix, and fina...

This paper presents a local spatial-spectral correlation based mixtures of factor analyzers (LSSC-MFA) denoising method for hyperspectral image (HSI). HSIs are usually degraded by different noise types such as missing lines (ML), missing pixels (MP), salt and pepper noise (SP), and Gaussian noise. The proposed method, hierarchically, removes the mi...

Sentinel-2 (S2) satellite constellations can provide multispec-tral images of 10 m, 20 m, and 60 m resolution for visible, near-infrared (NIR) and shortwave infrared (SWIR) in the electromagnetic spectrum. In this paper, we present a sharpening method based on a symmetric skipped connection con-volutional neural network, called SSC-CNN, to sharpen...

This paper addresses the hyperspectral image (HSI) denois-ing problem by using Stein's unbiased risk estimate (SURE) based convolutional neural network (CNN). Conventional deep learning denoising approaches often use supervised methods that minimize a mean-squared error (MSE) by training on noisy-clean image pairs. In contrast, our proposed CNN-bas...

Hyperspectral images (HSI) are composed of hundreds of spectral bands, covering a broad range of the electromagnetic spectrum. However, images can only be visualized using three spectral channels for red, green, and blue (RGB) colors. Generating realistic RGB images using HSI is seldom the main focus of remote sensing researchers, and is therefore...