Getinge (Sweden)

Although Machine Learning can be seen as a promising tool to improve clinical decision-making, it remains limited by access to healthcare data. Healthcare data is sensitive, requiring strict privacy practices, and typically stored in data silos, making traditional Machine Learning challenging. Federated Learning can counteract those limitations by training Machine Learning models over data silos while keeping the sensitive data localized. This study proposes a Federated Learning workflow for Intensive Care Unit mortality prediction. Hereby, the applicability of Federated Learning as an alternative to Centralized Machine Learning and Local Machine Learning is investigated by introducing Federated Learning to the binary classification problem of predicting Intensive Care Unit mortality. We extract multivariate time series data from the MIMIC-III database (lab values and vital signs), and benchmark the predictive performance of four deep sequential classifiers (FRNN, LSTM, GRU, and 1DCNN) varying the patient history window lengths (8h, 16h, 24h, and 48h) and the number of Federated Learning clients (2, 4, and 8). The experiments demonstrate that both Centralized Machine Learning and Federated Learning are comparable in terms of AUPRC and F1-score. Furthermore, the federated approach shows superior performance over Local Machine Learning. Thus, Federated Learning can be seen as a valid and privacy-preserving alternative to Centralized Machine Learning for classifying Intensive Care Unit mortality when the sharing of sensitive patient data between hospitals is not possible.

This paper focuses on the computation of joint spectral radii (JSR), when the involved matrices are sparse. We provide a sparse variant of the procedure proposed by Parrilo and Jadbabaie in [24], to compute upper bounds of the JSR by means of sum-of-squares (SOS) relaxations. Our resulting iterative algorithm, called SparseJSR, is based on the term sparsity SOS (TSSOS) framework, developed by Wang, Magron and Lasserre in [37], yielding SOS decompositions of polynomials with arbitrary sparse support. SparseJSR exploits the sparsity of the input matrices to significantly reduce the computational burden associated with the JSR computation. Our algorithmic framework is then successfully applied to compute upper bounds for JSR, on randomly generated benchmarks as well as on problems arising from stability proofs of controllers, in relation with possible hardware and software faults.

From an early age, humans intuitively expect physical objects to obey core principles, including continuity (objects follow spatiotemporally continuous paths) and solidity (two solid objects cannot occupy the same space at the same time). These 2 principles are sometimes viewed as deriving from a single overarching "persistence" principle. Indeed, violations of solidity where one solid object seemingly passes through another could theoretically be interpreted as a violation of continuity, with an object "teleporting" to switch places rather than passing through a solid obstacle. However, it is an empirical issue whether the two principles are processed distinctly or identically to one another. Here, adult participants tracked objects during dynamic events in a novel location detection task, which sometimes involved violations of the principles of continuity or solidity. Although participants explicitly noticed both types of violations and reported being equally surprised at both, they made more errors and answered more slowly after continuity violations than after solidity violations. Our results demonstrate that the two principles show different signature patterns and are thus represented distinctly in the mind. (PsycInfo Database Record (c) 2020 APA, all rights reserved).

International audience

Cigarette smoking is one of the most important health concerns worldwide. Even though the rate of smoking is declining in developed countries, it is still experiencing growth in developing regions. Many studies have examined the relationship between smoking, as an established risk factor, and cardiovascular diseases. We provide an updated review of the underlying mechanisms of smokinginduced cardiovascular diseases, with a focus on the relationship between smoking and oxidative stress, particularly from the perspective of endothelial cell dysfunction. We review smoking-induced oxidative stress as a trigger for a generalized vascular inflammation associated with cytokine release, adhesion of inflammatory cells and, ultimately, disruption of endothelial integrity as a protective barrier layer. We also briefly discuss the harms related to the vaping of electronic cigarettes, which many erroneously consider as a safe alternative to smoking. We conclude that even though e-cigarette could be a helpful device during the transition period of cigarette quitting, it is by no means a safe substitute.

ABSTRACTWT1 is a transcription factor with a DNA binding N-terminal domain containing four C2H2-type zinc fingers. In order to perform its role as a transcription factor, WT1 needs to specifically recognize and properly bind to its target DNA. How this is done is still not completely clear. Two of WT1’s major isoforms are distinguished by the presence or absence of a 3 amino acid insert, Lysine-Threonine-Serine (KTS) in the linker between zinc-fingers 3 and 4. This KTS insert is conserved throughout all life forms expressing WT1. The –KTS isoform, which acts as a transcription factor, binds DNA with higher affinity than the +KTS isoform, which is thought to participate in RNA splicing and interaction with partner proteins. This study was aims at elucidating the effect of the KTS insert on DNA binding. Here we present the crystal structure of WT1 zinc fingers 2-4, with and without the KTS insert, bound to the WT1 9-base pair cognate DNA sequence, refined to 1.9 Å and 2.5 Å respectively. The structures show that the +KTS isoform of WT1 recognizes DNA with the same specificity as the –KTS isoform. The only differences in the DNA bound conformation of the two isoforms are found within the linker containing the KTS, and these mainly involve the loss of the C-capping interactions thought to stabilize the complex. These structures provide the molecular detail necessary for the interpretation of the WT1 transcriptional DNA recognition and validation of its transcriptional targets.

The BOADICEA breast cancer (BC) risk assessment model and its associated Web Application v3 (BWA) tool are being extended to incorporate additional genetic and non-genetic BC risk factors. From an online survey through the BOADICEA website and UK, Dutch, French and Swedish national genetic societies, we explored the relationships between the usage frequencies of the BWA and six other common BC risk assessment tools and respondents’ perceived importance of BC risk factors. Respondents (N = 443) varied in age, country and clinical seniority but comprised mainly genetics health professionals (82%) and BWA users (93%). Oncology professionals perceived reproductive, hormonal (exogenous) and lifestyle BC risk factors as more important in BC risk assessment compared to genetics professionals (p values < 0.05 to 0.0001). BWA was used more frequently by respondents who gave high weight to breast tumour pathology and low weight to personal BC history as BC risk factors. BWA use was positively related to the weight given to hormonal BC risk factors. The importance attributed to lifestyle and BMI BC risk factors was not associated with the use of BWA or any of the other tools. Next version of the BWA encompassing additional BC risk factors will facilitate more comprehensive BC risk assessment in genetics and oncology practice.

The question of registering two images (or image volumes) acquired with different modalities, and thus exhibiting different contrast, at different positions is addressed based on an extension of global digital image (or volume) correlation. A specific comparison metric is introduced allowing the signature of the different phases to be related. A first solution consists of a Gaussian mixture to describe the joint distribution of gray levels, which not only provides a matching of both images, but also offers a natural segmentation indicator. A second ‘self-adapting’ solution does not include any postulated a priori model for the joint histogram and leads to a registration of the images based on their initial histograms. The algorithm is implemented with a pyramidal multiscale framework for the sake of robustness. The proposed multiscale technique is tested on two 3D images obtained from x-ray and neutron tomography respectively. The proposed approach brings the two images to coincidence with a sub-pixel accuracy and allows for a ‘natural’ segmentation of the different phases.

An ab initio methodology for the first-principle derivation of crystal-field (CF) parameters for lanthanides is described. The methodology is applied to the analysis of CF parameters in [Tb(Pc)2 ]- (Pc=phthalocyanine) and Dy4 K2 ([Dy4 K2 O(OtBu)12 ]) complexes, and compared with often used approximate and model descriptions. It is found that the application of geometry symmetrization, and the use of electrostatic point-charge and phenomenological CF models, lead to unacceptably large deviations from predictions based on ab initio calculations for experimental geometry. It is shown how the predictions of standard CASSCF (Complete Active Space Self-Consistent Field) calculations (with 4f orbitals in the active space) can be systematically improved by including effects of dynamical electronic correlation (CASPT2 step) and by admixing electronic configurations of the 5d shell. This is exemplified for the well-studied Er-trensal complex (H3 trensal=2,2',2"-tris(salicylideneimido)trimethylamine). The electrostatic contributions to CF parameters in this complex, calculated with true charge distributions in the ligands, yield less than half of the total CF splitting, thus pointing to the dominant role of covalent effects. This analysis allows the conclusion that ab initio crystal field is an essential tool for the decent description of lanthanides.