Fisher information for smart sampling in time-domain spectroscopy.

Time-domain spectroscopy encompasses a wide range of techniques, such as Fourier-transform infrared, pump-probe, Fourier-transform Raman, and two-dimensional electronic spectroscopies. These methods enable various applications, such as molecule characterization, excited state dynamics studies, or spectral classification. Typically, these techniques rarely use sampling schemes that exploit the prior knowledge scientists typically have before the actual experiment. Indeed, not all sampling coordinates carry the same amount of information, and a careful selection of the sampling points may notably affect the resulting performance. In this work, we rationalize, with examples, the various advantages of using an optimal sampling scheme tailored to the specific experimental characteristics and/or expected results. We show that using a sampling scheme optimizing the Fisher information minimizes the variance of the desired parameters. This can greatly improve, for example, spectral classifications and multidimensional spectroscopy. We demonstrate how smart sampling may reduce the acquisition time of an experiment by one to two orders of magnitude, while still providing a similar level of information.

Adaptive sparse estimation of nonlinear chirp signals using Laplace priors.

The identification of nonlinear chirp signals has attracted notable attention in the recent literature, including estimators such as the variational mode decomposition and the nonlinear chirp mode estimator. However, most presented methods fail to process signals with close frequency intervals or depend on user-determined parameters that are often non-trivial to select optimally. In this work, we propose a fully adaptive method, termed the adaptive nonlinear chirp mode estimation. The method decomposes a combined nonlinear chirp signal into its principal modes, accurately representing each mode's time-frequency representation simultaneously. Exploiting the sparsity of the instantaneous amplitudes, the proposed method can produce estimates that are smooth in the sense of being piecewise linear. Furthermore, we analyze the decomposition problem from a Bayesian perspective, using hierarchical Laplace priors to form an efficient implementation, allowing for a fully automatic parameter selection. Numerical simulations and experimental data analysis show the effectiveness and advantages of the proposed method. Notably, the algorithm is found to yield reliable estimates even when encountering signals with crossed modes. The method's practical potential is illustrated on a whale whistle signal.

BERT based natural language processing for triage of adverse drug reaction reports shows close to human-level performance.

Post-marketing reports of suspected adverse drug reactions are important for establishing the safety profile of a medicinal product. However, a high influx of reports poses a challenge for regulatory authorities as a delay in identification of previously unknown adverse drug reactions can potentially be harmful to patients. In this study, we use natural language processing (NLP) to predict whether a report is of serious nature based solely on the free-text fields and adverse event terms in the report, potentially allowing reports mislabelled at time of reporting to be detected and prioritized for assessment. We consider four different NLP models at various levels of complexity, bootstrap their train-validation data split to eliminate random effects in the performance estimates and conduct prospective testing to avoid the risk of data leakage. Using a Swedish BERT based language model, continued language pre-training and final classification training, we achieve close to human-level performance in this task. Model architectures based on less complex technical foundation such as bag-of-words approaches and LSTM neural networks trained with random initiation of weights appear to perform less well, likely due to the lack of robustness that a base of general language training provides.

[The Swecrit Biobank, associated clinical registries, and machine learning (artificial intelligence) improve critical care knowledge]. / Provsamlingen Swecrit och AI ger ny kunskap vid intensivvård.

The unique Swecrit Biobank and its associated clinical registries for sepsis, ARDS, cardiac arrest, trauma, and COVID-19 include more than 150,000 blood samples and descriptions of critically ill patients. These assets provide a unique opportunity to research and improve the care of the most seriously ill patients through biomarker analyses, proteomic studies, and genetic and epigenetic studies using modern machine learning techniques (artificial intelligence). Interested researchers are invited to submit their proposals and participate.

Comparison of dyskinesia profiles after L-DOPA dose challenges with or without dopamine agonist coadministration.

Many patients with Parkinson's disease (PD) experiencing l-DOPA-induced dyskinesia (LID) receive adjunct treatment with dopamine agonists, whose functional impact on LID is unknown. We set out to compare temporal and topographic profiles of abnormal involuntary movements (AIMs) after l-DOPA dose challenges including or not the dopamine agonist ropinirole. Twenty-five patients with PD and a history of dyskinesias were sequentially administered either l-DOPA alone (150% of usual morning dose) or an equipotent combination of l-DOPA and ropinirole in random order. Involuntary movements were assessed by two blinded raters prior and every 30 min after drug dosing using the Clinical Dyskinesia Rating Scale (CDRS). A sensor-recording smartphone was secured to the patients' abdomen during the test sessions. The two raters' CDRS scores were highly reliable and concordant with models of hyperkinesia presence and severity trained on accelerometer data. The dyskinesia time curves differed between treatments as the l-DOPA-ropinirole combination resulted in lower peak severity but longer duration of the AIMs compared with l-DOPA alone. At the peak of the AIMs curve (60-120 min), l-DOPA induced a significantly higher total hyperkinesia score, whereas in the end phase (240-270 min), both hyperkinesia and dystonia tended to be more severe after the l-DOPA-ropinirole combination (though reaching statistical significance only for the item, arm dystonia). Our results pave the way for the introduction of a combined l-DOPA-ropinirole challenge test in the early clinical evaluation of antidyskinetic treatments. Furthermore, we propose a machine-learning method to predict CDRS hyperkinesia severity using accelerometer data.

Sparse optimization for nonlinear group delay mode estimation.

Nonlinear group delay signals with frequency-varying characteristics are common in a wide variety of fields, for instance, structural health monitoring and fault diagnosis. For such applications, the signal is composed of multiple modes, where each mode may overlap in the frequency-domain. The resulting decomposition and forming of time-frequency representations of the nonlinear group delay modes is a challenging task. In this study, the nonlinear group delay signal is modelled in the frequency-domain. Exploiting the sparsity of the signal, we present the nonlinear group delay mode estimation technique, which forms the demodulation dictionary from the group delay. This method can deal with crossed modes and transient impulse signals. Furthermore, an augmented alternating direction multiplier method is introduced to form an efficient implementation. Numerical simulations and experimental data analysis show the effectiveness and advantages of the proposed method. In addition, the included analysis of Lamb waves as well as of a bearing signal show the method's potential for structural health monitoring and fault diagnosis.

Tremor evaluation using smartphone accelerometry in standardized settings.

Tremor can be highly incapacitating in everyday life and typically fluctuates depending on motor state, medication status as well as external factors. For tremor patients being treated with deep-brain stimulation (DBS), adapting the intensity and pattern of stimulation according the current needs therefore has the potential to generate better symptomatic relief. We here describe a procedure for how patients independently could perform self-tests in their home to generate sensor data for on-line adjustments of DBS parameters. Importantly, the inertia sensor technology needed exists in any standard smartphone, making the procedure widely accessible. Applying this procedure, we have characterized detailed features of tremor patterns displayed by both Parkinson's disease and essential tremor patients and directly compared measured data against both clinical ratings (Fahn-Tolosa-Marin) and finger-attached inertia sensors. Our results suggest that smartphone accelerometry, when used in a standardized testing procedure, can provide tremor descriptors that are sufficiently detailed and reliable to be used for closed-loop control of DBS.

Rapid measurement of heteronuclear transverse relaxation rates using non-uniformly sampled R1ρ accordion experiments.

Multidimensional, heteronuclear NMR relaxation methods are used extensively to characterize the dynamics of biological macromolecules. Acquisition of relaxation datasets on proteins typically requires significant measurement time, often several days. Accordion spectroscopy offers a powerful means to shorten relaxation rate measurements by encoding the "relaxation dimension" into the indirect evolution period in multidimensional experiments. Time savings can also be achieved by non-uniform sampling (NUS) of multidimensional NMR data, which is used increasingly to improve spectral resolution or increase sensitivity per unit time. However, NUS is not commonly implemented in relaxation experiments, because most reconstruction algorithms are inherently nonlinear, leading to problems when estimating signal intensities, relaxation rate constants and their error bounds. We have previously shown how to avoid these shortcomings by combining accordion spectroscopy with NUS, followed by data reconstruction using sparse exponential mode analysis, thereby achieving a dramatic decrease in the total length of longitudinal relaxation experiments. Here, we present the corresponding transverse relaxation experiment, taking into account the special considerations required for its successful implementation in the framework of the accordion-NUS approach. We attain the highest possible precision in the relaxation rate constants by optimizing the NUS scheme with respect to the Cramér-Rao lower bound of the variance of the estimated parameter, given the total number of sampling points and the spectrum-specific signal characteristics. The resulting accordion-NUS R1ρ relaxation experiment achieves comparable precision in the parameter estimates compared to conventional CPMG (Carr-Purcell-Meiboom-Gill) R2 or spin-lock R1ρ experiments while saving an order of magnitude in experiment time.

Development and evaluation of a deep learning based artificial intelligence for automatic identification of gold fiducial markers in an MRI-only prostate radiotherapy workflow.

Identification of prostate gold fiducial markers in magnetic resonance imaging (MRI) images is challenging when CT images are not available, due to misclassifications from intra-prostatic calcifications. It is also a time consuming task and automated identification methods have been suggested as an improvement for both objectives. Multi-echo gradient echo (MEGRE) images have been utilized for manual fiducial identification with 100% detection accuracy. The aim is therefore to develop an automatic deep learning based method for fiducial identification in MRI images intended for MRI-only prostate radiotherapy. MEGRE images from 326 prostate cancer patients with fiducials were acquired on a 3T MRI, post-processed with N4 bias correction, and the fiducial center of mass (CoM) was identified. A 9 mm radius sphere was created around the CoM as ground truth. A deep learning HighRes3DNet model for semantic segmentation was trained using image augmentation. The model was applied to 39 MRI-only patients and 3D probability maps for fiducial location and segmentation were produced and spatially smoothed. In each of the three largest probability peaks, a 9 mm radius sphere was defined. Detection sensitivity and geometric accuracy was assessed. To raise awareness of potential false findings a 'BeAware' score was developed, calculated from the total number and quality of the probability peaks. All datasets, annotations and source code used were made publicly available. The detection sensitivity for all fiducials were 97.4%. Thirty-six out of thirty-nine patients had all fiducial markers correctly identified. All three failed patients generated a user notification using the BeAware score. The mean absolute difference between the detected fiducial and ground truth CoM was 0.7 ± 0.9 [0 3.1] mm. A deep learning method for automatic fiducial identification in MRI images was developed and evaluated with state-of-the-art results. The BeAware score has the potential to notify the user regarding patients where the proposed method is uncertain.

Artificial neural networks improve early outcome prediction and risk classification in out-of-hospital cardiac arrest patients admitted to intensive care.

BACKGROUND: Pre-hospital circumstances, cardiac arrest characteristics, comorbidities and clinical status on admission are strongly associated with outcome after out-of-hospital cardiac arrest (OHCA). Early prediction of outcome may inform prognosis, tailor therapy and help in interpreting the intervention effect in heterogenous clinical trials. This study aimed to create a model for early prediction of outcome by artificial neural networks (ANN) and use this model to investigate intervention effects on classes of illness severity in cardiac arrest patients treated with targeted temperature management (TTM). METHODS: Using the cohort of the TTM trial, we performed a post hoc analysis of 932 unconscious patients from 36 centres with OHCA of a presumed cardiac cause. The patient outcome was the functional outcome, including survival at 180 days follow-up using a dichotomised Cerebral Performance Category (CPC) scale with good functional outcome defined as CPC 1-2 and poor functional outcome defined as CPC 3-5. Outcome prediction and severity class assignment were performed using a supervised machine learning model based on ANN. RESULTS: The outcome was predicted with an area under the receiver operating characteristic curve (AUC) of 0.891 using 54 clinical variables available on admission to hospital, categorised as background, pre-hospital and admission data. Corresponding models using background, pre-hospital or admission variables separately had inferior prediction performance. When comparing the ANN model with a logistic regression-based model on the same cohort, the ANN model performed significantly better (p = 0.029). A simplified ANN model showed promising performance with an AUC above 0.852 when using three variables only: age, time to ROSC and first monitored rhythm. The ANN-stratified analyses showed similar intervention effect of TTM to 33 °C or 36 °C in predefined classes with different risk of a poor outcome. CONCLUSION: A supervised machine learning model using ANN predicted neurological recovery, including survival excellently, and outperformed a conventional model based on logistic regression. Among the data available at the time of hospitalisation, factors related to the pre-hospital setting carried most information. ANN may be used to stratify a heterogenous trial population in risk classes and help determine intervention effects across subgroups.

Compressed Sensing for Reconstructing Coherent Multidimensional Spectra.

We apply two sparse reconstruction techniques, the least absolute shrinkage and selection operator (LASSO) and the sparse exponential mode analysis (SEMA), to two-dimensional (2D) spectroscopy. The algorithms are first tested on model data, showing that both are able to reconstruct the spectra using only a fraction of the data required by the traditional Fourier-based estimator. Through the analysis of the sparsely sampled experimental fluorescence-detected 2D spectra of LH2 complexes, we conclude that both SEMA and LASSO can be used to significantly reduce the required data, still allowing one to reconstruct the multidimensional spectra. Of the two techniques, it is shown that SEMA offers preferable performance, providing more accurate estimation of the spectral line widths and their positions. Furthermore, SEMA allows for off-grid components, enabling the use of a much smaller dictionary than that of the LASSO, thereby improving both the performance and the lowering of the computational complexity for reconstructing coherent multidimensional spectra.

Artificial neural networks improve and simplify intensive care mortality prognostication: a national cohort study of 217,289 first-time intensive care unit admissions.

PURPOSE: We investigated if early intensive care unit (ICU) scoring with the Simplified Acute Physiology Score (SAPS 3) could be improved using artificial neural networks (ANNs). METHODS: All first-time adult intensive care admissions in Sweden during 2009-2017 were included. A test set was set aside for validation. We trained ANNs with two hidden layers with random hyper-parameters and retained the best ANN, determined using cross-validation. The ANNs were constructed using the same parameters as in the SAPS 3 model. The performance was assessed with the area under the receiver operating characteristic curve (AUC) and Brier score. RESULTS: A total of 217,289 admissions were included. The developed ANN (AUC 0.89 and Brier score 0.096) was found to be superior (p <10-15 for AUC and p <10-5 for Brier score) in early prediction of 30-day mortality for intensive care patients when compared with SAPS 3 (AUC 0.85 and Brier score 0.109). In addition, a simple, eight-parameter ANN model was found to perform just as well as SAPS 3, but with better calibration (AUC 0.85 and and Brier score 0.106, p <10-5). Furthermore, the ANN model was superior in correcting mortality for age. CONCLUSION: ANNs can outperform the SAPS 3 model for early prediction of 30-day mortality for intensive care patients.

Rapid NMR Relaxation Measurements Using Optimal Nonuniform Sampling of Multidimensional Accordion Data Analyzed by a Sparse Reconstruction Method.

Nonuniform sampling (NUS) of multidimensional NMR data offers significant time savings while improving spectral resolution or increasing sensitivity per unit time. However, NUS has not been widely used for quantitative analysis because of the nonlinearity of most methods used to model NUS data, which leads to problems in estimating signal intensities, relaxation rate constants, and their error bounds. Here, we present an approach that avoids these limitations by combining accordion spectroscopy and NUS in the indirect dimensions of multidimensional spectra and then applying sparse exponential mode analysis, which is well suited for analyzing accordion-type relaxation data in a NUS context. By evaluating the Cramér-Rao lower bound of the variances of the estimated relaxation rate constants, we achieve a robust benchmark for the underlying reconstruction model. Furthermore, we design NUS schemes optimized with respect to the information theoretical lower bound of the error in the parameters of interest, given a specified number of sampling points. The accordion-NUS method compares favorably with conventional relaxation experiments in that it produces identical results, within error, while shortening the length of the experiment by an order of magnitude. Thus, our approach enables rapid acquisition of NMR relaxation data for optimized use of spectrometer time or accurate measurements on samples of limited lifetime.

Sparse estimation of backscattered echoes from underwater object using integrated dictionaries.

The problem of time-delays estimation of backscattered echoes from underwater targets is presented using a sparse reconstruction framework employing an integrated dictionary. To achieve high resolution, the used dictionary is usually defined over a finely spaced grid over the region of interest. Such a procedure may result in problems of being computational cumbersome or suffering from basis mismatch. In addition, the shape of the backscattered echoes may differ significantly from the expected waveforms used to form the dictionary, causing further mismatch problems. To alleviate such problems, the use of an integrated dictionary framework is introduced. Unlike traditional dictionaries that are defined over a set of grid points, the elements in an integrated dictionary are formed by integrating the expected waveform over bands of the parameter space. The resulting dictionary may be used to find initial regions of the parameters of interest using a smaller dictionary than otherwise required, without suffering a loss of performance. The elements can also better match with the backscattered echoes, even if these differ from their expected shape. Simulated results of the backscattered echoes from a cylindrical shell, as well as results from experimental measurements, illustrate the performance of the proposed method.

Batch-specific discrimination using nuclear quadrupole resonance spectroscopy.

In this paper, we report on the identification of batches of analgesic paracetamol (acetaminophen) tablets using nitrogen-14 nuclear quadrupole resonance spectroscopy ((14)N NQR). The high sensitivity of NQR to the electron charge distribution surrounding the quadrupolar nucleus enables the unique characterization of the crystal structure of the material. Two hypothesis were tested on batches of the same brand: the within the same batch variability and the difference between batches that varied in terms of their batch number and expiry date. The multivariate analysis of variance (MANOVA) did not provide any within-batches variations, indicating the natural deviation of a medicine manufactured under the same conditions. Alternatively, the statistical analysis revealed a significant discrimination between the different batches of paracetamol tablets. Therefore, the NQR signal is an indicator of factors that influence the physical and chemical integrity of the material. Those factors might be the aging of the medicine, the manufacturing, or storage conditions. The results of this study illustrate the potential of NQR as promising technique in applications such as detection and authentication of counterfeit medicines.

Spectral estimation of irregularly sampled exponentially decaying signals with applications to RF spectroscopy.

The problem of estimating the spectral content of exponentially decaying signals from a set of irregularly sampled data is of considerable interest in several applications, for example in various forms of radio frequency spectroscopy. In this paper, we propose a new nonparametric iterative adaptive approach that provides a solution to this estimation problem. As opposed to commonly used methods in the field, the damping coefficient, or linewidth, is explicitly modeled, which allows for an improved estimation performance. Numerical examples using both simulated data and data from NQR experiments illustrate the benefits of the proposed estimator as compared to currently available nonparametric methods.

Adaptive spectral doppler estimation.

In this paper, 2 adaptive spectral estimation techniques are analyzed for spectral Doppler ultrasound. The purpose is to minimize the observation window needed to estimate the spectrogram to provide a better temporal resolution and gain more flexibility when designing the data acquisition sequence. The methods can also provide better quality of the estimated power spectral density (PSD) of the blood signal. Adaptive spectral estimation techniques are known to provide good spectral resolution and contrast even when the observation window is very short. The 2 adaptive techniques are tested and compared with the averaged periodogram (Welch's method). The blood power spectral capon (BPC) method is based on a standard minimum variance technique adapted to account for both averaging over slow-time and depth. The blood amplitude and phase estimation technique (BAPES) is based on finding a set of matched filters (one for each velocity component of interest) and filtering the blood process over slow-time and averaging over depth to find the PSD. The methods are tested using various experiments and simulations. First, controlled flow-rig experiments with steady laminar flow are carried out. Simulations in Field II for pulsating flow resembling the femoral artery are also analyzed. The simulations are followed by in vivo measurement on the common carotid artery. In all simulations and experiments it was concluded that the adaptive methods display superior performance for short observation windows compared with the averaged periodogram. Computational costs and implementation details are also discussed.

Steroid hormones control circadian Elovl3 expression in mouse liver.

The Elovl3 gene belongs to the Elovl gene family, which encodes for enzymes involved in the elongation of very long chain fatty acids. The recognized role for the enzyme is to control the elongation of saturated and monounsaturated fatty acids up to 24 carbons in length. Elovl3 was originally identified as a highly expressed gene in brown adipose tissue on cold exposure. Here we show that hepatic Elovl3 mRNA expression follows a distinct diurnal rhythm exclusively in mature male mice, with a sharp increase early in the morning Zeitgeber time (ZT) 20, peaks around ZT2, and is back to basal level at the end of the light period at ZT10. In female mice and sexually immature male mice, the Elovl3 expression was constantly low. Fasting and refeeding mice with chow or high-fat diet did not alter the Elovl3 mRNA levels. However, animals that were exclusively fed during the day for 9 d displayed an inverted expression profile. In addition, we show that Elovl3 expression is transcriptionally controlled and significantly induced by the exposure of the synthetic glucocorticoid dexamethasone. Taken together, these data suggest that Elovl3 expression in mouse liver is under strict diurnal control by circulating steroid hormones such as glucocorticoids and androgens. Finally, Elovl3 expression was found to be elevated in peroxisomal transporter ATP-binding cassette, subfamily D(ALD), member 2 ablated mice and suppressed in ATP-binding cassette subfamily D(ALD) member 2 overexpressing mice, implying a tight cross talk between very long chain fatty acid synthesis and peroxisomal fatty acid oxidation.

Fatty acid elongases in mammals: their regulation and roles in metabolism.

A significant amount of the fatty acids synthesized by the cytosolic enzyme complex fatty acid synthase (FAS) or taken up by the diet are further elongated into very long chain fatty acids (VLCFA) in a four-step reaction cycle by membrane-bound enzymes predominantly located in the endoplasmic reticulum. Members of the Elovl (elongation-of-very-long-chain-fatty acids) gene family encode for enzymes (elongases), which are believed to perform the first, regulatory, step (condensation) in the elongation cycle in mammals. The family of enzymes consists of at least six members in mouse and human, believed to carry out substrate-specific elongation with fatty acids of different lengths and degrees of unsaturation. The ability to synthesize VLCFA is a ubiquitous system found in different organs and cell types. However, VLCFAs seldom occur unesterified. Instead, they are joined either by an ester or amide linkage to a broad variety of different lipid species. VLCFA are most commonly found as building blocks in sphingolipids, although they are also important constituents of glycerophospholipids, triacylglycerols, sterol- and wax-esters. To generalize, the fatty acid elongases can be divided into two major groups: (a) enzymes which are suggested to be involved in the elongation of saturated and monounsaturated VLCFA (ELOVL1, 3 and 6) and (b) enzymes which are elongases of polyunsaturated fatty acids (PUFA) (ELOVL2, 4 and 5). All the elongases exhibit specific spatial and temporal expression. In this review, we will present and discuss the regulation of the mammalian fatty acid elongases and their potential role in lipid metabolism. We will consider both the biochemical functions of the proteins, as well as their role in a more physiological context.

Differential regulation of fatty acid elongation enzymes in brown adipocytes implies a unique role for Elovl3 during increased fatty acid oxidation.

The expression of the Elovl3 gene, which belongs to the Elovl gene family coding for microsomal enzymes involved in very long-chain fatty acid (VLCFA) elongation, is dramatically increased in mouse brown adipose tissue upon cold stimulation. In the present study, we show that the cold-induced Elovl3 expression is under the control of peroxisome proliferator-activated receptor-alpha (PPARalpha) and that this regulation is part of a fundamental divergence in the regulation of expression for the different members of the Elovl gene family. In cultured brown adipocytes, a mixture of norepinephrine, dexamethasone, and the PPARalpha ligand Wy-14643, which rendered the adipocytes a high oxidative state, was required for substantial induction of Elovl3 expression, whereas the same treatment suppressed Elovl1 mRNA levels. The nuclear liver X receptor (LXR) has been implicated in the control of fatty acid synthesis and subsequent lipogenic processes in several tissues. This regulation is also exerted in part by sterol regulatory element-binding protein (SREBP-1), which is a target gene of LXR. We found that stimulation of Elovl3 expression was independent of LXR and SREBP-1 activation. In addition, exposure to the LXR agonist TO-901317 increased nuclear abundance of LXR and mature SREBP-1 as well as expression of the elongases Lce and Elovl1 in a lipogenic fashion but repressed Elovl3 expression. A functional consequence of this was seen on the level of esterified saturated fatty acids, such as C22:0, which was coupled to Elovl3 expression. These data demonstrate differential transcriptional regulation and concomitantly different functional roles for fatty acid elongases in lipid metabolism of brown adipocytes, which reflects the metabolic status of the cells.