Segmentation of mature human oocytes provides interpretable and improved blastocyst outcome predictions by a machine learning model.

Within the medical field of human assisted reproductive technology, a method for interpretable, non-invasive, and objective oocyte evaluation is lacking. To address this clinical gap, a workflow utilizing machine learning techniques has been developed involving automatic multi-class segmentation of two-dimensional images, morphometric analysis, and prediction of developmental outcomes of mature denuded oocytes based on feature extraction and clinical variables. Two separate models have been developed for this purpose-a model to perform multiclass segmentation, and a classifier model to classify oocytes as likely or unlikely to develop into a blastocyst (Day 5-7 embryo). The segmentation model is highly accurate at segmenting the oocyte, ensuring high-quality segmented images (masks) are utilized as inputs for the classifier model (mask model). The mask model displayed an area under the curve (AUC) of 0.63, a sensitivity of 0.51, and a specificity of 0.66 on the test set. The AUC underwent a reduction to 0.57 when features extracted from the ooplasm were removed, suggesting the ooplasm holds the information most pertinent to oocyte developmental competence. The mask model was further compared to a deep learning model, which also utilized the segmented images as inputs. The performance of both models combined in an ensemble model was evaluated, showing an improvement (AUC 0.67) compared to either model alone. The results of this study indicate that direct assessments of the oocyte are warranted, providing the first objective insights into key features for developmental competence, a step above the current standard of care-solely utilizing oocyte age as a proxy for quality.

An artificial intelligence tool predicts blastocyst development from static images of fresh mature oocytes.

RESEARCH QUESTION: Can a deep learning image analysis model be developed to assess oocyte quality by predicting blastocyst development from images of denuded mature oocytes? DESIGN: A deep learning model was developed utilizing 37,133 static oocyte images with associated laboratory outcomes from eight fertility clinics (six countries). A subset of data (n = 7807) was allocated to test model performance. External model validation was conducted to assess generalizability and robustness on new data (n = 12,357) from two fertility clinics (two countries). Performance was assessed by calculating area under the curve (AUC), balanced accuracy, specificity and sensitivity. Subgroup analyses were performed on the test dataset for age group, male factor and geographical location of the clinic. Model probabilities of the external dataset were converted to a 0-10 scoring scale to facilitate analysis of correlation with blastocyst development and quality. RESULTS: The deep learning model demonstrated AUC of 0.64, balanced accuracy of 0.60, specificity of 0.55 and sensitivity of 0.65 on the test dataset. Subgroup analyses displayed the highest performance for age group 38-39 years (AUC 0.68), a negligible impact of male factor, and good model generalizability across geographical locations. Model performance was confirmed on external data: AUC of 0.63, balanced accuracy of 0.58, specificity of 0.57 and sensitivity of 0.59. Analysis of the scoring scale revealed that higher scoring oocytes correlated with higher likelihood of blastocyst development and good-quality blastocyst formation. CONCLUSION: The deep learning model showed a favourable performance for the evaluation of oocytes in terms of competence to develop into a blastocyst, and when the predictions were converted into scores, they correlated with blastocyst quality. This represents a significant first step in oocyte evaluation for scientific and clinical applications.

Use of a Novel EEG-Based Objective Test, the Cognalyzer®, in Quantifying the Strength and Determining the Action Time of Cannabis Psychoactive Effects and Factors that May Influence Them Within an Observational Study Framework.

INTRODUCTION: Current methods to detect recent delta-9-tetrahydrocannabinol (THC) use cannot objectively quantify its psychoactive effects (PE). The Cognalyzer®, an electroencephalography (EEG)-based method, detects and quantifies the strength of THC-induced PE on a scale from 0 to 100%. This study assesses the relationship between the magnitude of Cognalyzer® PE predictions and reported subjective drug effects for 4-h post-cannabis inhalation. METHODS: Seventy-five participants were enrolled in the study. Prior to ad libitum cannabis inhalation, an EEG recording episode was completed. Immediately after inhalation, the Drug Effects Questionnaire (DEQ) was administered and another EEG recording performed. For 25 participants, the study ended. For 50 participants, assessments were repeated at 30-min intervals for 4 h. EEG files were blinded and analyzed using two versions of the Cognalyzer® algorithm. The relationship between the Cognalyzer® PE level results and the DEQ was assessed using generalized linear models and multiple regression. RESULTS: There were significant PE increases from pre-cannabis for up to 3.5 h. Mean reports of feeling drug effects were > 0 at all post-inhalation time points (p ≤ 0.024). Furthermore, there were significant relationships between the Cognalyzer® PE and self-reported perception of drug effects (p ≤ 0.001). Subgroup analysis showed that Cognalyzer® PE levels were impacted by cannabis use history, subjective ratings of drug effects, oral fluid THC concentration and the cannabis product inhaled. CONCLUSION: The findings show that the Cognalyzer® can be used to objectively determine the strength of cannabis psychoactive effects that cannabis products create on consumers and how it changes depending on their experience with cannabis. The Cognalyzer® can be used to conduct scientific consumer research to generate trustworthy informational material about the psychoactive experience of cannabis products. For clinical research, the Cognalyzer® can be used to study the pharmacodynamics of cannabinoids or delivery systems, such as nano-emulsifications.

Sensitivity, Specificity and Accuracy of a Novel EEG-Based Objective Test, the Cognalyzer®, in Detecting Cannabis Psychoactive Effects.

INTRODUCTION: Current standards for identifying recent cannabis use are based on body fluid testing. The Cognalyzer® is a novel electroencephalography (EEG) measurement device and algorithm designed to objectively characterize brainwave alterations associated with cannabis. The objective of this study was to assess the accuracy, sensitivity and specificity levels of the Cognalyzer® to characterize brainwave alterations following cannabis inhalation. METHODS: Seventy-five participants, aged 19-55 years, were enrolled, and oral fluid samples were collected pre-cannabis inhalation. EEG and subjective drug effects questionnaire (DEQ) were administered pre- and post-ad libitum cannabis inhalation. Fifty participants remained in the clinic for 4 h post-inhalation. Blinded analyses of the EEG files were conducted by Zentrela Inc. using two versions (V1 and V2) of the Cognalyzer® algorithm. Pre- vs. post-inhalation comparison status was characterized by the Cognalyzer® and summarized for: sensitivity, specificity, accuracy, percent false positive, percent false negative and positive and negative predictive value. The null hypothesis was tested using McNemar's test. Cognalyzer® results pre- and post-inhalation were combined with the oral fluid tetrahydrocannabinol (THC) concentration to evaluate potential to improve current drug testing. RESULTS: The two versions of the Cognalyzer® algorithm had similar diagnostic results. Diagnostic outcomes were improved when participants with missing EEG recordings or electrode placement errors were removed. The Cognalyzer® accuracy was 85.5% and 83.9%, sensitivity was 87.1% and 88.7%, and specificity was 83.9% and 79.0% for algorithm V1 and V2, respectively. Combining Cognalyzer® results with oral fluid concentrations reduced false-positive oral fluid test results by up to 49%. CONCLUSION: The Cognalyzer® characterized brainwave alterations associated with cannabis inhalation with high levels of accuracy in a population of participants with varied cannabis inhalation histories, relative to the comparison standard of pre- vs. post-inhalation status. The Cognalyzer® allows the results to be generalized to the larger population addressing a limitation in currently accepted standards.

Molecular Simulation of Chemical Reaction Equilibrium by Computationally Efficient Free Energy Minimization.

The molecular simulation of chemical reaction equilibrium (CRE) is a challenging and important problem of broad applicability in chemistry and chemical engineering. The primary molecular-based approach for solving this problem has been the reaction ensemble Monte Carlo (REMC) algorithm [Turner et al. Molec. Simulation2008, 34, (2), 119-146], based on classical force-field methodology. In spite of the vast improvements in computer hardware and software since its original development almost 25 years ago, its more widespread application is impeded by its computational inefficiency. A fundamental problem is that its MC basis inhibits the implementation of significant parallelization, and its successful implementation often requires system-specific tailoring and the incorporation of special MC approaches such as replica exchange, expanded ensemble, umbrella sampling, configurational bias, and continuous fractional component methodologies. We describe herein a novel CRE algorithm (reaction ensemble molecular dynamics, ReMD) that exploits modern computer hardware and software capabilities, and which can be straightforwardly implemented for systems of arbitrary size and complexity by exploiting the parallel computing methodology incorporated within many MD software packages (herein, we use GROMACS for illustrative purposes). The ReMD algorithm utilizes these features in the context of a macroscopically inspired and generally applicable free energy minimization approach based on the iterative approximation of the system Gibbs free energy function by a mathematically simple convex ideal solution model using the composition at each iteration as a reference state. Finally, we additionally describe a simple and computationally efficient a posteriori method to estimate the equilibrium concentrations of species present in very small amounts relative to others in the primary calculation. To demonstrate the algorithm, we show its application to two classic example systems considered previously in the literature: the N2-O2-NO system and the ammonia synthesis system.

Glass forming phase diagram and local structure of Kob-Andersen binary Lennard-Jones nanoparticles.

Molecular dynamics simulation is used to study glass formation in Kob-Andersen binary Lennard-Jones nanoparticles and determine the glass forming phase diagram for the system as a function of composition. The radial distribution function, a Steinhardt bond-orientational order parameter, and favored local structure analysis are used to distinguish between glassy and ordered systems. We find that surface enrichment of the large atoms alters the nanoparticle core composition, leading to an overall shift of the glass forming region to lower small atom mole fractions, relative to the bulk system. At small atom mole fraction, xB = 0.1, the nanoparticles form a solid with an amorphous core, enriched with the small atoms, surrounded by a partially ordered surface region, enriched with the large atom component. The most disordered glass nanoparticles occur at xB ≈ 0.3, but the surface-core enrichment leads to the crystallization of the nanoparticle to the CsCl crystal above xB ≈ 0.35, which is lower than observed in the bulk. The glass transition temperatures of the nanoparticles are also significantly reduced. This allows the liquid to remain dynamic to low temperatures and sample the low energy inherent structure minima on the potential energy surface containing a high abundance of favoured local structures.

Molecular simulation of caloric properties of fluids modelled by force fields with intramolecular contributions: Application to heat capacities.

The calculation of caloric properties such as heat capacity, Joule-Thomson coefficients, and the speed of sound by classical force-field-based molecular simulation methodology has received scant attention in the literature, particularly for systems composed of complex molecules whose force fields (FFs) are characterized by a combination of intramolecular and intermolecular terms. The calculation of a thermodynamic property for a system whose molecules are described by such a FF involves the calculation of the residual property prior to its addition to the corresponding ideal-gas property, the latter of which is separately calculated, either using thermochemical compilations or nowadays accurate quantum mechanical calculations. Although the simulation of a volumetric residual property proceeds by simply replacing the intermolecular FF in the rigid molecule case by the total (intramolecular plus intermolecular) FF, this is not the case for a caloric property. We describe the correct methodology required to perform such calculations and illustrate it in this paper for the case of the internal energy and the enthalpy and their corresponding molar heat capacities. We provide numerical results for cP, one of the most important caloric properties. We also consider approximations to the correct calculation procedure previously used in the literature and illustrate their consequences for the examples of the relatively simple molecule 2-propanol, CH3CH(OH)CH3, and for the more complex molecule monoethanolamine, HO(CH2)2NH2, an important fluid used in carbon capture.

Vapor Condensed and Supercooled Glassy Nanoclusters.

We use molecular simulation to study the structural and dynamic properties of glassy nanoclusters formed both through the direct condensation of the vapor below the glass transition temperature, without the presence of a substrate, and via the slow supercooling of unsupported liquid nanodroplets. An analysis of local structure using Voronoi polyhedra shows that the energetic stability of the clusters is characterized by a large, increasing fraction of bicapped square antiprism motifs. We also show that nanoclusters with similar inherent structure energies are structurally similar, independent of their history, which suggests the supercooled clusters access the same low energy regions of the potential energy landscape as the vapor condensed clusters despite their different methods of formation. By measuring the intermediate scattering function at different radii from the cluster center, we find that the relaxation dynamics of the clusters are inhomogeneous, with the core becoming glassy above the glass transition temperature while the surface remains mobile at low temperatures. This helps the clusters sample the highly stable, low energy structures on the potential energy surface. Our work suggests the nanocluster systems are structurally more stable than the ultrastable glassy thin films, formed through vapor deposition onto a cold substrate, but the nanoclusters do not exhibit the superheating effects characteristic of the ultrastable glass states.

Nonclassical Nucleation in a Solid-Solid Transition of Confined Hard Spheres.

A solid-solid phase transition of colloidal hard spheres confined between two planar hard walls is studied using a combination of molecular dynamics and Monte Carlo simulation. The transition from a solid consisting of five crystalline layers with square symmetry (5□) to a solid consisting of four layers with triangular symmetry (4△) is shown to occur through a nonclassical nucleation mechanism that involves the initial formation of a precritical liquid cluster, within which the cluster of the stable 4△ phase grows. Free-energy calculations show that the transition occurs in one step, crossing a single free-energy barrier, and that the critical nucleus consists of a small 4△ solid cluster wetted by a metastable liquid. In addition, the liquid cluster and the solid cluster are shown to grow at the planar hard walls. We also find that the critical nucleus size increases with supersaturation, which is at odds with classical nucleation theory. The △-solid-like cluster is shown to contain both face-centered-cubic and hexagonal-close-packed ordered particles.

A novel chiral phase of achiral hard triangles and an entropy-driven demixing of enantiomers.

We investigate the phase behavior of a system of hard equilateral and right-angled triangles in two dimensions using Monte Carlo simulations. Hard equilateral triangles undergo a continuous isotropic-triatic liquid crystal phase transition at packing fraction ϕ = 0.7. Similarly, hard right-angled isosceles triangles exhibit a first-order phase transition from an isotropic fluid phase to a rhombic liquid crystal phase with a coexistence region ϕ ∈ [0.733, 0.782]. Both these liquid crystals undergo a continuous phase transition to their respective close-packed crystal structures at high pressures. Although the particles and their close-packed crystals are both achiral, the solid phases of equilateral and right-angled triangles exhibit spontaneous chiral symmetry breaking at sufficiently high packing fractions. The colloidal triangles rotate either in the clockwise or anti-clockwise direction with respect to one of the lattice vectors for packing fractions higher than ϕχ. As a consequence, these triangles spontaneously form a regular lattice of left- or right-handed chiral holes which are surrounded by six triangles in the case of equilateral triangles and four or eight triangles for right-angled triangles. Moreover, our simulations show a spontaneous entropy-driven demixing transition of the right- and left-handed "enantiomers".

Destabilisation of the hexatic phase in systems of hard disks by quenched disorder due to pinning on a lattice.

We investigate the effect of quenched disorder on the melting mechanism of two-dimensional hard disks using large-scale event-driven molecular dynamics simulations. The two-stage melting scenario of a continuous solid-hexatic and a first-order hexatic-liquid transition for a 2D system of hard disks does not persist in the case of quenched disorder, which arises by pinning less than one percent of the particles on a triangular lattice. Based on the Halperin-Nelson-Young (HNY) renormalization group equation, we observe that a first-order solid-liquid transition preempts the Kosterlitz-Thouless-type solid-hexatic transition in a 2D system of hard disks with quenched disorder as the stiffness of the crystal is increased by the presence of pinned particles.

Highly cooperative stress relaxation in two-dimensional soft colloidal crystals.

Stress relaxation in crystalline solids is mediated by the formation and diffusion of defects. Although it is well established how externally generated stresses relax, through the proliferation and motion of dislocations in the lattice, it remains relatively unknown how crystals cope with internal stresses. We investigate, both experimentally and in simulations, how highly localized stresses relax in 2D soft colloidal crystals. When a single particle is actively excited, by means of optical tweezing, a rich variety of highly collective stress relaxation mechanisms results. These relaxation processes manifest in the form of open strings of cooperatively moving particles through the motion of dissociated vacancy-interstitial pairs, and closed loops of mobile particles, which either result from cooperative rotations in transiently generated circular grain boundaries or through the closure of an open string by annihilation of a vacancy-interstitial pair. Surprisingly, we find that the same collective events occur in crystals that are excited by thermal fluctuations alone; a large thermal agitation inside the crystal lattice can trigger the irreversible displacements of hundreds of particles. Our results illustrate how local stresses can induce large-scale cooperative dynamics in 2D soft colloidal crystals and shed light on the stabilization mechanisms in ultrasoft crystals.

Two-stage melting induced by dislocations and grain boundaries in monolayers of hard spheres.

Melting in two-dimensional systems has remained controversial as theory, simulations, and experiments show contrasting results. One issue that obscures this discussion is whether or not theoretical predictions on strictly 2D systems describe those on quasi-2D experimental systems, where out-of-plane fluctuations may alter the melting mechanism. Using event-driven molecular dynamics simulations, we find that the peculiar two-stage melting scenario of a continuous solid-hexatic and a first-order hexatic-liquid transition as observed for a truly 2D system of hard disks [Bernard and Krauth, Phys. Rev. Lett., 2011, 107, 155704] persists for a quasi-2D system of hard spheres with out-of-plane particle motions as high as half the particle diameter. By calculating the renormalized Young's modulus, we show that the solid-hexatic transition is of the Kosterlitz-Thouless type and occurs via dissociation of bound dislocation pairs. In addition, we find a first-order hexatic-liquid transition that seems to be driven by spontaneous proliferation of grain boundaries.

Evolution of topological defects in two-dimensional quenched colloidal systems.

We investigated the evolution of topological defects in two-dimensional (2D) quenched colloidal systems using topological current theory. As a singularity of topological currents in order parameter fields, a topological defect is associated with three cases of solutions of zero points: the isolation solution, the limit point, and the bifurcation point. At the limit point, the defects represent a generation or annihilation process, and the number of defects satisfies a power law time-dependent scaling behaviour N d ∝ t (-1) . At the bifurcation point, a merging or splitting process appears and N d ∝ t (-2) . These properties are in agreement with the results from Brownian dynamics simulations of the quenching processes in 2D colloidal systems with a Yukawa pair interaction.

Phase diagram of octapod-shaped nanocrystals in a quasi-two-dimensional planar geometry.

Recently, we reported the formation of crystalline monolayers consisting of octapod-shaped nanocrystals (so-called octapods) that had arranged in a square-lattice geometry through drop deposition and fast evaporation on a substrate [W. Qi, J. de Graaf, F. Qiao, S. Marras, L. Manna, and M. Dijkstra, Nano Lett. 12, 5299 (2012)]. In this paper we give a more in-depth exposition on the Monte Carlo simulations in a quasi-two-dimensional (quasi-2D) geometry, by which we modelled the experimentally observed crystal structure formation. Using a simulation model for the octapods consisting of four hard interpenetrating spherocylinders, we considered the effect of the pod length-to-diameter ratio on the phase behavior and we constructed the full phase diagram. The methods we applied to establish the nature of the phase transitions between the various phases are discussed in detail. We also considered the possible existence of a Kosterlitz-Thouless-type phase transition between the isotropic liquid and hexagonal rotator phase for certain pod length-to-diameter ratios. Our methods may prove instrumental in guiding future simulation studies of similar anisotropic nanoparticles in confined geometries and monolayers.

Electron tomography resolves a novel crystal structure in a binary nanocrystal superlattice.

The self-assembly of different nanocrystals into a binary superlattice is of interest for both colloidal science and nanomaterials science. New properties may emerge from the interaction between the nanocrystal building blocks that are ordered in close contact in three dimensions. Identification of the superlattice structure including its defects is of key interest in understanding the electrical and optical properties of these systems. Transmission electron microscopy (TEM) has been very instrumental to reach this goal but fails for complex crystal structures and buried defects. Here, we use electron tomography to resolve the three-dimensional crystal structure of a binary superlattice that could not be resolved by TEM only. The structure with a [PbSe]6[CdSe]19 stoichiometry has no analogue in the atomic world. Moreover we will show how tomography can overcome the clouding effects of planar defects on structure identification by TEM.

Ordered two-dimensional superstructures of colloidal octapod-shaped nanocrystals on flat substrates.

We studied crystal structures in a monolayer consisting of anisotropic branched colloidal (nano)octapods. Experimentally, octapods were observed to form a monolayer on a substrate with a square-lattice crystal structure by drop-casting and fast evaporation of solvent. The experimental results were analyzed by Monte Carlo simulations using a hard octapod model consisting of four interpenetrating spherocylinders. We confirmed by means of free-energy calculations that crystal structures with a (binary-lattice) square morphology are indeed thermodynamically stable at high densities. The effect of the pod length-to-diameter ratio on the crystal structures was also considered and we used this to constructed the phase diagram for these hard octapods. In addition to the (binary-lattice) square crystal phase, a rhombic crystal and a hexagonal plastic-crystal (rotator) phase were obtained. Our phase diagram may prove instrumental in guiding future experimental studies.

Two features at the two-dimensional freezing transitions.

We studied the two-dimensional freezing transitions in monolayers of microgel colloidal spheres with short-ranged repulsions in video-microscopy experiments, and monolayers of hard disks, and Yukawa particles in simulations. These systems share two common features at the freezing points: (1) the bimodal distribution profile of the local orientational order parameter; (2) the two-body excess entropy, s(2), reaches -4.5±0.5 k(B). Both features are robust and sensitive to the freezing points, so that they can potentially serve as empirical freezing criteria in two dimensions. Compared with the conventional freezing criteria, the first feature has no finite-size ambiguities and can be resolved adequately with much less statistics; and the second feature can be directly measured in macroscopic experiments without the need for microscopic information.

Melting in two-dimensional Yukawa systems: a Brownian dynamics simulation.

We studied the melting behavior of two-dimensional colloidal crystals with a Yukawa pair potential by Brownian dynamics simulations. The melting follows the Kosterlitz-Thouless-Halperin-Nelson-Young (KTHNY) scenario with two continuous phase transitions and a middle hexatic phase. The two phase-transition points were accurately identified from the divergence of the translational and orientational susceptibilities. Configurational temperatures were employed to monitor the equilibrium of the overdamped system and the strongest temperature fluctuation was observed in the hexatic phase. The inherent structure obtained by rapid quenching exhibits three different behaviors in the solid, hexatic, and liquid phases. The measured core energy of the free dislocations, E(c) = 7.81 ± 0.91 k(B)T, is larger than the critical value of 2.84 k(B)T, which consistently supports the KTHNY melting scenario.

Four-state rock-paper-scissors games in constrained Newman-Watts networks.

We study the cyclic dominance of three species in two-dimensional constrained Newman-Watts networks with a four-state variant of the rock-paper-scissors game. By limiting the maximal connection distance Rmax in Newman-Watts networks with the long-range connection probability p , we depict more realistically the stochastic interactions among species within ecosystems. When we fix mobility and vary the value of p or Rmax, the Monte Carlo simulations show that the spiral waves grow in size, and the system becomes unstable and biodiversity is lost with increasing p or Rmax. These results are similar to recent results of Reichenbach et al. [Nature (London) 448, 1046 (2007)], in which they increase the mobility only without including long-range interactions. We compared extinctions with or without long-range connections and computed spatial correlation functions and correlation length. We conclude that long-range connections could improve the mobility of species, drastically changing their crossover to extinction and making the system more unstable.