CNN-based fully automatic wrist cartilage volume quantification in MR images: A comparative analysis between different CNN architectures.

PURPOSE: Automatic measurement of wrist cartilage volume in MR images. METHODS: We assessed the performance of four manually optimized variants of the U-Net architecture, nnU-Net and Mask R-CNN frameworks for the segmentation of wrist cartilage. The results were compared to those from a patch-based convolutional neural network (CNN) we previously designed. The segmentation quality was assessed on the basis of a comparative analysis with manual segmentation. The best networks were compared using a cross-validation approach on a dataset of 33 3D VIBE images of mostly healthy volunteers. Influence of some image parameters on the segmentation reproducibility was assessed. RESULTS: The U-Net-based networks outperformed the patch-based CNN in terms of segmentation homogeneity and quality, while Mask R-CNN did not show an acceptable performance. The median 3D DSC value computed with the U-Net_AL (0.817) was significantly larger than DSC values computed with the other networks. In addition, the U-Net_AL provided the lowest mean volume error (17%) and the highest Pearson correlation coefficient (0.765) with respect to the ground truth values. Of interest, the reproducibility computed using U-Net_AL was larger than the reproducibility of the manual segmentation. Moreover, the results indicate that the MRI-based wrist cartilage volume is strongly affected by the image resolution. CONCLUSIONS: U-Net CNN with attention layers provided the best wrist cartilage segmentation performance. In order to be used in clinical conditions, the trained network can be fine-tuned on a dataset representing a group of specific patients. The error of cartilage volume measurement should be assessed independently using a non-MRI method.

Crystal nucleation in a glass during relaxation well below Tg.

Until quite recently, in almost all papers on crystal nucleation in glass-forming substances, it was assumed that nucleation proceeds in a completely relaxed supercooled liquid and, hence, at constant values of the critical parameters determining the nucleation rate for any given set of temperature, pressure, and composition. Here, we analyze the validity of this hypothesis for a model system by studying nucleation in a lithium silicate glass treated for very long times (up to 250 days) in deeply supercooled states, reaching 60 K below the laboratory glass transition temperature, Tg. At all temperatures in the considered range, T < Tg, we observed an enormous difference between the experimental number of nucleated crystals, N(t), and its theoretically expected value computed by assuming the metastable state of the relaxing glass has been reached. Analyzing the origin of this discrepancy, we confirmed that the key parameters determining the nucleation rates change with time as a result of the glass relaxation process. Finally, we demonstrate that, for temperatures below 683 K, this particular glass almost fully crystallizes prior to reaching the ultimate steady-state nucleation regime (e.g., at 663 K, it would take 176 years for the glass to reach 99% crystallization, while 2600 years would be needed for complete relaxation). This comprehensive study proves that structural relaxation strongly affects crystal nucleation in deeply supercooled states at temperatures well below Tg; hence, this phenomenon has to be accounted for in any crystal nucleation model.

Crystallization of Supercooled Liquids: Self-Consistency Correction of the Steady-State Nucleation Rate.

Crystal nucleation can be described by a set of kinetic equations that appropriately account for both the thermodynamic and kinetic factors governing this process. The mathematical analysis of this set of equations allows one to formulate analytical expressions for the basic characteristics of nucleation, i.e., the steady-state nucleation rate and the steady-state cluster-size distribution. These two quantities depend on the work of formation, Δ G ( n ) = - n Δ µ + γ n 2 / 3 , of crystal clusters of size n and, in particular, on the work of critical cluster formation, Δ G ( n c ) . The first term in the expression for Δ G ( n ) describes changes in the bulk contributions (expressed by the chemical potential difference, Δ µ ) to the Gibbs free energy caused by cluster formation, whereas the second one reflects surface contributions (expressed by the surface tension, σ : γ = Ω d 0 2 σ , Ω = 4 π ( 3 / 4 π ) 2 / 3 , where d 0 is a parameter describing the size of the particles in the liquid undergoing crystallization), n is the number of particles (atoms or molecules) in a crystallite, and n = n c defines the size of the critical crystallite, corresponding to the maximum (in general, a saddle point) of the Gibbs free energy, G. The work of cluster formation is commonly identified with the difference between the Gibbs free energy of a system containing a cluster with n particles and the homogeneous initial state. For the formation of a "cluster" of size n = 1 , no work is required. However, the commonly used relation for Δ G ( n ) given above leads to a finite value for n = 1 . By this reason, for a correct determination of the work of cluster formation, a self-consistency correction should be introduced employing instead of Δ G ( n ) an expression of the form Δ G ˜ ( n ) = Δ G ( n ) - Δ G ( 1 ) . Such self-consistency correction is usually omitted assuming that the inequality Δ G ( n ) ≫ Δ G ( 1 ) holds. In the present paper, we show that: (i) This inequality is frequently not fulfilled in crystal nucleation processes. (ii) The form and the results of the numerical solution of the set of kinetic equations are not affected by self-consistency corrections. However, (iii) the predictions of the analytical relations for the steady-state nucleation rate and the steady-state cluster-size distribution differ considerably in dependence of whether such correction is introduced or not. In particular, neglecting the self-consistency correction overestimates the work of critical cluster formation and leads, consequently, to far too low theoretical values for the steady-state nucleation rates. For the system studied here as a typical example (lithium disilicate, Li 2 O · 2 SiO 2 ), the resulting deviations from the correct values may reach 20 orders of magnitude. Consequently, neglecting self-consistency corrections may result in severe errors in the interpretation of experimental data if, as it is usually done, the analytical relations for the steady-state nucleation rate or the steady-state cluster-size distribution are employed for their determination.

Effects of Glass Transition and Structural Relaxation on Crystal Nucleation: Theoretical Description and Model Analysis.

In the application of classical nucleation theory (CNT) and all other theoretical models of crystallization of liquids and glasses it is always assumed that nucleation proceeds only after the supercooled liquid or the glass have completed structural relaxation processes towards the metastable equilibrium state. Only employing such an assumption, the thermodynamic driving force of crystallization and the surface tension can be determined in the way it is commonly performed. The present paper is devoted to the theoretical treatment of a different situation, when nucleation proceeds concomitantly with structural relaxation. To treat the nucleation kinetics theoretically for such cases, we need adequate expressions for the thermodynamic driving force and the surface tension accounting for the contributions caused by the deviation of the supercooled liquid from metastable equilibrium. In the present paper, such relations are derived. They are expressed via deviations of structural order parameters from their equilibrium values. Relaxation processes result in changes of the structural order parameters with time. As a consequence, the thermodynamic driving force and surface tension, and basic characteristics of crystal nucleation, such as the work of critical cluster formation and the steady-state nucleation rate, also become time-dependent. We show that this scenario may be realized in the vicinity and below the glass transition temperature, and it may occur only if diffusion (controlling nucleation) and viscosity (controlling the alpha-relaxation process) in the liquid decouple. Analytical estimates are illustrated and confirmed by numerical computations for a model system. The theory is successfully applied to the interpretation of experimental data. Several further consequences of this newly developed theoretical treatment are discussed in detail. In line with our previous investigations, we reconfirm that only when the characteristic times of structural relaxation are of similar order of magnitude or longer than the characteristic times of crystal nucleation, elastic stresses evolving in nucleation may significantly affect this process. Advancing the methods of theoretical analysis of elastic stress effects on nucleation, for the first time expressions are derived for the dependence of the surface tension of critical crystallites on elastic stresses. As the result, a comprehensive theoretical description of crystal nucleation accounting appropriately for the effects of deviations of the liquid from the metastable states and of relaxation on crystal nucleation of glass-forming liquids, including the effect of simultaneous stress evolution and stress relaxation on nucleation, is now available. As one of its applications, this theoretical treatment provides a new tool for the explanation of the low-temperature anomaly in nucleation in silicate and polymer glasses (the so-called "breakdown" of CNT at temperatures below the temperature of the maximum steady-state nucleation rate). We show that this anomaly results from much more complex features of crystal nucleation in glasses caused by deviations from metastable equilibrium (resulting in changes of the thermodynamic driving force, the surface tension, and the work of critical cluster formation, in the necessity to account of structural relaxation and stress effects) than assumed so far. If these effects are properly accounted for, then CNT appropriately describes both the initial, the intermediate, and the final states of crystal nucleation.

Deep learning-based fully automatic segmentation of wrist cartilage in MR images.

The study objective was to investigate the performance of a dedicated convolutional neural network (CNN) optimized for wrist cartilage segmentation from 2D MR images. CNN utilized a planar architecture and patch-based (PB) training approach that ensured optimal performance in the presence of a limited amount of training data. The CNN was trained and validated in 20 multi-slice MRI datasets acquired with two different coils in 11 subjects (healthy volunteers and patients). The validation included a comparison with the alternative state-of-the-art CNN methods for the segmentation of joints from MR images and the ground-truth manual segmentation. When trained on the limited training data, the CNN outperformed significantly image-based and PB-U-Net networks. Our PB-CNN also demonstrated a good agreement with manual segmentation (Sørensen-Dice similarity coefficient [DSC] = 0.81) in the representative (central coronal) slices with a large amount of cartilage tissue. Reduced performance of the network for slices with a very limited amount of cartilage tissue suggests the need for fully 3D convolutional networks to provide uniform performance across the joint. The study also assessed inter- and intra-observer variability of the manual wrist cartilage segmentation (DSC = 0.78-0.88 and 0.9, respectively). The proposed deep learning-based segmentation of the wrist cartilage from MRI could facilitate research of novel imaging markers of wrist osteoarthritis to characterize its progression and response to therapy.

Volumetric wireless coil based on periodically coupled split-loop resonators for clinical wrist imaging.

PURPOSE: Design and characterization of a new inductively driven wireless coil (WLC) for wrist imaging at 1.5 T with high homogeneity operating due to focusing the B1 field of a birdcage body coil. METHODS: The WLC design has been proposed based on a volumetric self-resonant periodic structure of inductively coupled split-loop resonators with structural capacitance. The WLC was optimized and studied regarding radiofrequency fields and interaction to the birdcage coil (BC) by electromagnetic simulations. The manufactured WLC was characterized by on-bench measurements and in vivo and phantom study in comparison to a standard cable-connected receive-only coil. RESULTS: The WLC placed into BC gave the measured B1+ increase of the latter by 8.6 times for the same accepted power. The phantom and in vivo wrist imaging showed that the BC in receiving with the WLC inside reached equal or higher signal-to-noise ratio than the conventional clinical setup comprising the transmit-only BC and a commercial receive-only flex-coil and created no artifacts. Simulations and on-bench measurements proved safety in terms of specific absorption rate and reflected transmit power. CONCLUSIONS: The results showed that the proposed WLC could be an alternative to standard cable-connected receive coils in clinical magnetic resonance imaging. As an example, with no cable connection, the WLC allowed wrist imaging on a 1.5 T clinical machine using a full-body BC for transmitting and receive with the desired signal-to-noise ratio, image quality, and safety.

Ultrasound-assisted focused open necrosectomy in the treatment of necrotizing pancreatitis.

CONTEXT: The surgical treatment of necrotizing pancreatitis includes percutaneous drainage of acute necrotic collections and sequestrectomy in the late phase of the disease. OBJECTIVE: The aim of the study was to compare the conventional open necrosectomy (CON) approach with the alternative focused open necrosectomy (FON) approach in patients with infected necrosis and progression of sepsis. METHODS: Patients with acute necrotizing pancreatitis were included in the study prospectively from January 2004 to July 2014. All patients had been admitted with the first or a new episode of disease. Symptomatic large fluid collections were drained percutaneously. The step-up approach was used in patients with several distant localizations of infected necrosis. The methods were analysed by comparing the individual severity according to the ASA, APACHE II and SOFA scores, infection rate, postoperative complication rate and mortality. RESULTS: A total of 31 patients were included in the FON group and 39 in the CON group. The incidence of infection was similar in groups. More ASA III comorbid conditions, a higher APACHE II score, a more frequent need for renal replacement therapy was observed in the CON group. The postoperative complication rate was in the range of 32% to 44%; mortality reached 6.5% in the FON group and 12.8% in the CON group. CONCLUSIONS: Comorbid conditions, organ failure, and infection are the main risk factors in patients with necrotizing pancreatitis. The step-up approach and perioperative ultrasonography navigation improves the clinical outcome and reduces the extent of invasive surgical intervention in patients unsuited to other minimally invasive procedures.

Focused open necrosectomy in necrotizing pancreatitis.

BACKGROUND: The control of sepsis is the primary goal of surgical intervention in patients with infected necrosis. Simple surgical approaches that are easy to reproduce may improve outcomes when specialists in endoscopy are not available. The aim of the present study was to describe the experience with a focused open necrosectomy (FON) in patients with infected necrosis. METHOD: A prospective pilot study conducted to compare a semi-open/closed drainage laparotomy and FON with the assistance of peri-operative ultrasound. The incidence of sepsis, dynamics of C-reactive protein (CRP), intensive care unit (ICU)/hospital stay, complication rate and mortality were compared and analysed. RESULTS: From a total of 58 patients, 36 patients underwent a conventional open necrosectomy and 22 patients underwent FON. The latter method resulted in a faster resolution of sepsis and a significant decrease in mean CRP on Day 3 after FON, P = 0.001. Post-operative bleeding was in 1 versus 7 patients and the incidence of intestinal and pancreatic fistula was 2 versus 8 patients when comparing FON to the conventional approach. The median ICU stay was 11.6 versus 23 days and the hospital stay was significantly shorter, 57 versus 72 days, P = 0.024 when comparing FON versus the conventional group. One patient died in the FON group and seven patients died in the laparotomy group, P = 0.139. DISCUSSION: FON can be an alternative method to conventional open necrosectomy in patients with infected necrosis and unresolved sepsis.

Dynamic processes in a silicate liquid from above melting to below the glass transition.

We collect and critically analyze extensive literature data, including our own, on three important kinetic processes--viscous flow, crystal nucleation, and growth--in lithium disilicate (Li(2)O·2SiO(2)) over a wide temperature range, from above T(m) to 0.98T(g) where T(g) ≈ 727 K is the calorimetric glass transition temperature and T(m) = 1307 K, which is the melting point. We found that crystal growth mediated by screw dislocations is the most likely growth mechanism in this system. We then calculated the diffusion coefficients controlling crystal growth, D(eff)(U), and completed the analyses by looking at the ionic diffusion coefficients of Li(+1), O(2-), and Si(4+) estimated from experiments and molecular dynamic simulations. These values were then employed to estimate the effective volume diffusion coefficients, D(eff)(V), resulting from their combination within a hypothetical Li(2)Si(2)O(5) "molecule". The similarity of the temperature dependencies of 1/η, where η is shear viscosity, and D(eff)(V) corroborates the validity of the Stokes-Einstein/Eyring equation (SEE) at high temperatures around T(m). Using the equality of D(eff)(V) and D(eff)(η), we estimated the jump distance λ ~ 2.70 Å from the SEE equation and showed that the values of D(eff)(U) have the same temperature dependence but exceed D(eff)(η) by about eightfold. The difference between D(eff)(η) and D(eff)(U) indicates that the former determines the process of mass transport in the bulk whereas the latter relates to the mobility of the structural units on the crystal/liquid interface. We then employed the values of η(T) reduced by eightfold to calculate the growth rates U(T). The resultant U(T) curve is consistent with experimental data until the temperature decreases to a decoupling temperature T(d)(U) ≈ 1.1-1.2T(g), when D(eff)(η) begins decrease with decreasing temperature faster than D(eff)(U). A similar decoupling occurs between D(eff)(η) and D(eff)(τ) (estimated from nucleation time-lags) but at a lower temperatureT(d)(τ) ≈ T(g). For T > T(g) the values of D(eff)(τ) exceed D(eff)(η) only by twofold. The different behaviors of D(eff)(τ)(T) and D(eff)(U)(T) are likely caused by differences in the mechanisms of critical nuclei formation. Therefore, we have shown that at low undercoolings, viscosity data can be employed for quantitative analyses of crystal growth rates, but in the deeply supercooled liquid state, mass transport for crystal nucleation and growth are not controlled by viscosity. The origin of decoupling is assigned to spatially dynamic heterogeneity in glass-forming melts.

Statistical design and analysis of label-free LC-MS proteomic experiments: a case study of coronary artery disease.

This chapter presents a case study, which applies statistical design and analysis to an LC-MS-based -investigation of subjects with coronary artery disease. First, we discuss the principles of statistical -experimental design, and the specification of an Analysis of Variance (ANOVA) model that describes the major sources of variation in the data. Second, we discuss procedures for detecting differentially abundant proteins, estimating protein abundance in individual samples, testing predefined groups of proteins for enrichment in differential abundance, and calculating sample size for a future experiment. The discussion is accompanied by examples of computer code implemented in the open-source statistical software R, which can be followed for an independent implementation of a similar investigation.

Diffusion coefficients for crystal nucleation and growth in deeply undercooled glass-forming liquids.

We calculate, employing the classical theory of nucleation and growth, the effective diffusion coefficients controlling crystal nucleation of nanosize clusters and the subsequent growth of micron-size crystals at very deep undercoolings, below and above Tg, using experimental nucleation and growth data obtained for stoichiometric Li2O.2SiO2 and Na2O.2CaO.3SiO2 glasses. The results show significant differences in the magnitude and temperature dependence of these kinetic coefficients. We explain this difference showing that the composition and/or structure of the nucleating critical clusters deviate from those of the stable crystalline phase. These results for diffusion coefficients corroborate our previous conclusion for the same glasses, based on different experiments, and support the view that, even for the so-called case of stoichiometric (polymorphic) crystallization, the nucleating phase may have a different composition and/or structure as compared to the parent glass and the evolving macroscopic crystalline phase. This finding gives a key to explain the discrepancies between calculated (by classical nucleation theory) and experimentally observed nucleation rates in these systems, in particular, and in deeply undercooled glass-forming liquids, in general.

Pressure dependence of viscosity.

We reanalyze the pressure dependence of viscosity of liquids of constant composition under isothermal conditions. Based exclusively on very general considerations concerning the relationship between viscosity and "free volume," we show that, at moderate values of pressure, viscosity increases, as a rule, with increasing pressure, provided the liquid is in stable or metastable (undercooled) equilibrium states. However, even if the behavior of the viscosity is governed by free volume effects, deviations from a positive pressure dependence are possible, when the liquid's thermal expansion coefficient is negative. We derive an equation that allows one to quantitatively determine the pressure dependence of viscosity, which requires, in the simplest case, only the knowledge of the temperature dependence of viscosity at constant pressure, the thermal expansion coefficient, and the isothermal compressibility of the liquid. As an example, the negative pressure dependence of water in the range of temperatures 0-4 degrees C and of several silicate liquids, such as albite, jadeite, dacite, basalts, etc., could be explained in such a way. Other glass-forming liquids initially (for moderate pressures) show a positive pressure dependence of viscosity that changes to a negative one when subjected to high (approximately GPa) isostatic pressure. A detailed analysis of water and already mentioned silicate melts at GPa pressures shows that, in addition to free volume effects, other pressure induced structural transformations may have to be accounted for in a variety of cases. By this reason, the theoretical analysis is extended (i) in order to describe the pressure dependence of viscosity for systems that are in frozen-in thermodynamic nonequilibrium states (glasses, i.e., undercooled liquids below the glass transition temperature Tg) and (ii) to systems which undergo, in addition to variations of the free volume, pressure induced changes of other structural parameters. In such cases a decrease of viscosity with increasing pressure may occur, in principle, even if the thermal expansion coefficient is positive. In this way, the present analysis grants a general tool to estimate the pressure dependence of viscosity and supposedly settles the controversy in the current literature.

Dynamics of first-order phase transitions in multicomponent systems: a new theoretical approach.

In the theoretical description of nucleation-growth processes, currently Gibbs's classical thermodynamic theory of heterogeneous systems is predominantly employed for the description of the properties of the clusters. However, Gibbs's approach does not make it possible to describe, in general, the properties of critical clusters (determining the rate of nucleation) in a sufficiently correct way. Moreover, Gibbs's approach is restricted by its applicability to thermodynamic equilibrium states exclusively. For this reason, it does not give a theoretically founded prescription for a determination of the possible states of clusters of sub- and supercritical sizes in dependence on supersaturation and size of the clusters. In order to overcome these shortcomings, in recent years a generalization of Gibbs's classical approach has been developed and employed for the description of nucleation processes. This generalization of Gibbs' classical method leads, for a variety of different applications, to dependencies of the work of critical cluster formation on supersaturation, which are qualitatively and widely even quantitatively in agreement with density-functional computations. The theoretical methods and results are summarized in the first part of the present paper. They are then extended for the first time to a description of processes of growth of single clusters and ensembles of clusters. In order to fulfill this task, a new method for determination of the parameters of sub- and supercritical clusters is developed. It turns out as the result of the analysis that a variety of thermodynamic and kinetic parameters, determining cluster growth, become dependent on cluster size as well. The results are illustrated for a model system (segregation in regular solutions) and applied to the interpretation of experimental results on segregation processes in solutions and crystallization processes in glass-forming melts. It is shown that the newly developed approach resolves a variety of problems in the interpretation of experimental data on the kinetics of phase formation processes which could not be given a satisfactory explanation so far.