Terms Of Loss Research Articles

Predicting transformer temperature field based on physics‐informed neural networks

AbstractThe safe operation of oil‐immersed transformers is critical to the safety and stability of the power grid. As the operating time increases, the failure rate of oil‐immersed transformers shows an increasing trend, posing serious challenges to safe operation. It is necessary to investigate the internal state of the oil‐immersed transformer to improve the digital degree and achieve digitalisation and intelligent operation and maintenance. A physics‐informed neural network (PINN) for oil‐immersed transformers was introduced to reconstruct the temperature distribution inside the transformer. According to the approach, the loss function of the network would be optimised by incorporating physical constraint loss terms including heat transfer equations, initial conditions and boundary conditions. The results show that the method proposed can be used to reconstruct and predict the temperature field of transformers in a few seconds with satisfactory accuracy. In conclusion, the PINN proposed outperforms deep neural networks in terms of accuracy, reliability and interpretability, especially in data‐poor cases.

A Novel Consolidation Analysis Framework: Universal Function Approximators Regularized by Physical Principles

Analytical and numerical techniques are widely used to analyse and interpret soil consolidation problems. An important limitation is the requirement for significant geotechnical knowledge and expertise to derive ‘true’ solutions. This study proposes an alternative: a universal function approximator regularized with known physical principles. The proposed approach here advances previous work to solve one-dimensional (1D) consolidation considering both self-weight and large strains, as well as two-dimensional (2D) consolidation by vertical drains. Initial and boundary conditions of the studied consolidation problems are first strongly and weakly enforced for comparison. A neural network is adopted here as the function approximator. To boost prediction accuracy, a novel strategy is proposed to adaptively sample data points for training. An estimate of epistemic uncertainty is achieved using the confidence interval of ensembled multiple outputs. The results show that the proposed approach accurately predicts the behaviours of complex consolidation processes. Results also indicate that regularization using weak physical constraints can alleviate the imbalance of back-propagated gradients of different loss terms and, in turn, achieve higher accuracy. The proposed method is generic, mesh-free, more robust and can be applied to a wide range of geotechnical problems.

Influence of learned landmark correspondences on lung CT registration.

Disease or injury may cause a change in the biomechanical properties of the lungs, which can alter lung function. Image registration can be used to measure lung ventilation and quantify volume change, which can be a useful diagnostic aid. However, lung registration is a challenging problem because of the variation in deformation along the lungs, sliding motion of the lungs along the ribs, and change indensity. Landmark correspondences have been used to make deformable image registration robust to largedisplacements. To tackle the challenging task of intra-patient lung computed tomography (CT) registration, we extend the landmark correspondence prediction model deep convolutional neural network-Match by introducing a soft mask loss term to encourage landmark correspondences in specific regions and avoid the use of a mask during inference. To produce realistic deformations to train the landmark correspondence model, we use data-driven synthetic transformations. We study the influence of these learned landmark correspondences on lung CT registration by integrating them into intensity-based registration as a distance-basedpenalty. Our results on the public thoracic CT dataset COPDgene show that using learned landmark correspondences as a soft constraint can reduce median registration error from approximately 5.46 to 4.08mm compared to standard intensity-based registration, in the absence of lungmasks. We show that using landmark correspondences results in minor improvements in local alignment, while significantly improving globalalignment.

Insects as radar targets: size, form, density and permittivity

ABSTRACT To make biologically useful interpretations of the echo signals from an entomological radar (a radar designed and operated to observe insects in flight), it is necessary to relate basic properties of insects (i.e. their mass, size, and form) to their radio-wave scattering characters – e.g. their radar cross-section (RCS) and its variation with aspect or beam polarization. Measurements of RCSs of live or freshly dead specimens in laboratory rigs have played an important role in establishing such relationships, but it is desirable to develop and validate methods employing numerical computation, using electromagnetic calculation software, as a more widely applicable and generalizable alternative. In such computations, the insect is represented by a model with specified size, form, and composition; composition appears as the complex relative permittivity, with both polarizability (real) and loss (imaginary) terms. As a preliminary to investigating how adequately different types of electromagnetic model perform, we have collected and analysed morphometric data on adult, flight-capable specimens of a wide range of insect species, including many known to be migratory and to fly at heights where they would be detectable by radar. Our analyses reveal that appropriate densities for prolate-spheroid models of flying insects are in the range 0.60–0.76 g cm−3, which is about half that measured for air-free samples of the solids and liquids from which insect bodies are formed. From this, appropriate permittivities for the material filling the model are found to be in the range 7–11 (real component) and 2.0–3.5 (imaginary); again, these are much lower than measured values for pure insect-body material. These results enable construction of simple electromagnetic models of insects in which both spatial dimensions and masses are accurate. Computation of model RCSs and their aspect and polarization variations, and comparison of these with laboratory measurements, will be the topic of a follow-up validation study.

Equivariant neural operators for gradient-consistent topology optimization

Abstract Most traditional methods for solving partial differential equations (PDEs) require the costly solving of large linear systems. Neural operators (NOs) offer remarkable speed-ups over classical numerical PDE solvers. Here, we conduct the first exploration and comparison of NOs for three-dimensional topology optimization. Specifically, we propose replacing the PDE solver within the popular Solid Isotropic Material with Penalization (SIMP) algorithm, which is its main computational bottleneck. For this, the NO not only needs to solve the PDE with sufficient accuracy but also has the additional challenge of providing accurate gradients which are necessary for SIMP’s density updates. To realize this, we do three things: (i) We introduce a novel loss term to promote gradient-consistency. (ii) We guarantee equivariance in our NOs to increase the physical correctness of predictions. (iii) We introduce a novel NO architecture called U-Net Fourier neural operator (U-Net FNO), which combines the multi-resolution properties of U-Nets with the Fourier neural operator (FNO)’s focus on local features in frequency space. In our experiments we demonstrate that the inclusion of the novel gradient loss term is necessary to obtain good results. Furthermore, enforcing group equivariance greatly improves the quality of predictions, especially on small training datasets. Finally, we show that in our experiments the U-Net FNO outperforms both a standard U-Net, as well as other FNO methods.

A physics-informed neural network framework to investigate nonlinear and heterogenous shrinkage of drying plant cells

This paper introduces a novel Physics-Informed Neural Network-based (PINN-based) multi-domain computational framework to analyse nonlinear and heterogeneous morphological variations of plant cells during drying. Here, two distinct models are involved: PINN-MT to simulate mass transfer; and PINNNS to simulate nonlinear shrinkage. The models are coupled to examine cellular morphological changes resulting from moisture loss during drying. Firstly, the coupled framework, in tandem with homogeneous conditions, operates in parallel, allowing the mutual parameters to update between models. This approach demonstrates ability to approximate homogeneous cellular shrinkage within a tissue, factoring in the influence of surrounding plant cells. Secondly, non-uniform cell wall properties and heterogeneous boundary conditions are incorporated into this computational framework through domain decomposition. Inherent capabilities of neural networks allow for seamless integration of multiple domains, with additional loss terms introduced at interfaces. The framework shows capacity to account for drastic and non-uniform morphological variations of plant cells even under extreme drying conditions, which is the key novelty and has been a challenging task for existing traditional computational methods. Hence, the proposed computational approach offers an innovative avenue for understanding nonlinear and heterogeneous morphological variations not only for plant cells, but also for soft matter in general.

Logic-Informed Graph Neural Networks for Structural Form-Finding

Computational form-finding methods hold great potential concerning resource-efficient structural design. The Combinatorial Equilibrium Modeling (CEM), an equilibrium-based form-finding method based on graphic statics and graph theory, allows the design of cross-typological tension–compression structures starting from an input topology diagram in the form of a graph. This paper presents a novel Logic-Informed Graph Neural Network (LIGNN) that integrates the validity conditions of CEM topology diagrams into the learning process through semantic loss terms. A Primary-LIGNN (P-LIGNN) and a Modification-LIGNN (M-LIGNN) are introduced and incorporated together with the CEM into a general form-finding-based computational structural design workflow that transforms input topologies into parametric models of equilibrium structures. An implementation of this computational design workflow for the conceptual design of pedestrian bridge structures is made, and presented through a case study, for which a synthetic training dataset of topology diagrams for the LIGNNs has been developed.

Ion flow and dust charging at the sheath boundary in dusty plasma with an electron-emitting surface: applications to laboratory and lunar dusty plasmas

In laboratory and space plasmas, the emission of electrons from the surface significantly affects the characteristics of the plasma sheath that forms at that surface, which is crucial to understanding the overall plasma-wall interaction mechanism. In this work, the collisional fluid model is used for laboratory dusty plasma, whereas the collisionless model is used for lunar dusty plasma. We have extended the Bohm sheath criterion for the formation of the stable plasma sheath due to electron emission from the surface, loss of ion flux, and the gas pressure of the collisional laboratory dusty plasmas. It is found that ion flow at the sheath boundary is considerably influenced by the concentration of electron emission, the ion loss term, and gas pressure. The evolution of the dust charge explicitly determines the magnitude of the ion flow at the sheath boundary. The plasma parameters adopted in the present case are reliable in laboratory and space dusty plasmas, especially the dusty plasma environment on the lunar surface. The lunar surface and dust grains on the Moon become electrically charged as a result of the interaction between solar wind plasma and photoemission electrons emitted from the lunar surface. In addition, the lunar plasma sheath characteristics, dust-charging process, and stable dust levitation in the sheath region have been studied.

Semi-Supervised Medical Image Segmentation Using Cross-Style Consistency With Shape-Aware and Local Context Constraints.

Despite the remarkable progress in semi-supervised medical image segmentation methods based on deep learning, their application to real-life clinical scenarios still faces considerable challenges. For example, insufficient labeled data often makes it difficult for networks to capture the complexity and variability of the anatomical regions to be segmented. To address these problems, we design a new semi-supervised segmentation framework that aspires to produce anatomically plausible predictions. Our framework comprises two parallel networks: shape-agnostic and shape-aware networks. These networks learn from each other, enabling effective utilization of unlabeled data. Our shape-aware network implicitly introduces shape guidance to capture shape fine-grained information. Meanwhile, shape-agnostic networks employ uncertainty estimation to further obtain reliable pseudo-labels for the counterpart. We also employ a cross-style consistency strategy to enhance the network's utilization of unlabeled data. It enriches the dataset to prevent overfitting and further eases the coupling of the two networks that learn from each other. Our proposed architecture also incorporates a novel loss term that facilitates the learning of the local context of segmentation by the network, thereby enhancing the overall accuracy of prediction. Experiments on three different datasets of medical images show that our method outperforms many excellent semi-supervised segmentation methods and outperforms them in perceiving shape. The code can be seen at https://github.com/igip-liu/SLC-Net.

A Hybrid PWM Strategy for Switching Loss Reduction in Three-Level Inverters

The advanced bus-clamping PWM (ABCPWM) sequences are well known to reduce line current THD in a three-level neutral point-clamped inverter. This paper deals with ABCPWM sequences employed optimally to achieve minimum switching loss throughout the linear operating region in a three-level inverter. The work extensively analyzes the performance of these sequences in terms of inverter switching loss and compares them with centered space vector modulation PWM (CSVPWM). Based on the analysis, a hybrid PWM scheme termed as <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">switching loss reduction</i> is proposed by optimally employing ABCPWM sequences in conjunction with variation in power factors. At a given average switching frequency, the proposed hybrid scheme reduces inverter switching loss compared to CSVPWM and the recently proposed discontinuous PWM-based optimization strategy at all power factors and modulation indices. With this scheme, switching loss reduces to 22% to 36% compared to CSVPWM and a maximum of 12% reduction than the existing optimization strategy. Furthermore, the scheme gives lower harmonic distortion compared to these two existing PWM schemes near high modulation indices and near unity power factor load. The theoretical, simulated, and real-time simulation results are presented with experimental results using TMS320F28335 DSP controller for validating the proposed work.

Continuous Refinement-based Digital Pathology Image Assistance Scheme in Medical Decision-Making Systems.

Digital pathology images' extensive cellular information provide a trustworthy foundation for tumor diagnosis. With the aid of computer-aided diagnostics, pathologists can locate crucial information more quickly. The cascade structure refines the segmentation results by utilizing its multi-task and multi-stage characteristics. However, cascade-based models require downsampling and cropping of patches during the inference process due to the ultra-high resolution and complex structure of pathology images. This not only increases the cost and computation time but also results in the loss of cellular details and corrupts the global contextual information. This study proposes a Digital Pathology Image Assistance Program (CRSDPI) for medical decision-making systems that is based on continuous improvement. After locating the region of interest using the maximum inter-class variance method, the pictures are preprocessed to account for the impacts of staining inconsistencies and sensitivity variations on the model's performance. Ultimately, we create a two-phase continuously refined segmentation network (TCRNet) by combining an enhanced continuous refinement model with a coarse segmentation network built on a pyramid scene parsing network. The coarse segmentation network introduces an auxiliary loss term to speed up convergence, and the refined model introduces an implicit function to reduce computational cost and reconstruct more details. The TCRNet model refines the target by successively aligning the features without the need to take cascading decoder operations after encoder. Experiments conducted on digital pathology images of breast cancer and osteosarcoma demonstrate the superior prediction accuracy and computational speed of our strategy.

A Qualitative and Comparative Performance Assessment of Logically Centralized SDN Controllers via Mininet Emulator

An alternative networking approach called Software Defined Networking (SDN) enables dynamic, programmatically efficient network construction, hence enhancing network performance. It splits a traditional network into a centralized control plane and a configurable data plane. Because the core component overseeing every data plane action is the controller in the control plane, which may contain one or more controllers and is thought of as the brains of the SDN network, controller functionality and performance are crucial to achieve optimal performances. There is much controller research available in the existing literature. Nevertheless, no qualitative comparison study of OpenFlow-enabled distributed but logically centralized controllers exists. This paper includes a quantitative investigation of the performance of several distributed but logically centralized SDN controllers in custom network scenarios using Mininet, as well as a thorough qualitative comparison of them. More precisely, we give a qualitative evaluation of their attributes and classify and categorize 13 distributed but logically centralized SDN controllers according to their capabilities. Additionally, we offer a comprehensive SDN emulation tool, called Mininet-based SDN controller performance assessment, in this study. Using six performance metrics—bandwidth, round-trip time, delay, jitter, packet loss, and throughput—this work also assesses five distributed but logically centralized controllers within two custom network scenarios (uniform and non-uniform host distribution). Our analysis reveals that the Ryu controller outperforms the OpenDayLight controller in terms of latency, packet loss, and round-trip time, while the OpenDayLight controller performs well in terms of throughput, bandwidth, and jitter. Throughout the entire experiment, the HyperFlow and ONOS controllers performed worst in all performance metrics. Finally, we discuss detailed research findings on performance. These experimental results provide decision-making guidelines when selecting a controller.

Batch Normalization Is Blind to the First and Second Derivatives of the Loss

We prove that when we do the Taylor series expansion of the loss function, the BN operation will block the influence of the first-order term and most influence of the second-order term of the loss. We also find that such a problem is caused by the standardization phase of the BN operation. We believe that proving the blocking of certain loss terms provides an analytic perspective for potential detects of a deep model with BN operations, although the blocking problem is not fully equivalent to significant damages in all tasks on benchmark datasets. Experiments show that the BN operation significantly affects feature representations in specific tasks.

Physics informed contour selection for rapid image segmentation

Effective training of deep image segmentation models is challenging due to the need for abundant, high-quality annotations. To facilitate image annotation, we introduce Physics Informed Contour Selection (PICS)—an interpretable, physics-informed algorithm for rapid image segmentation without relying on labeled data. PICS draws inspiration from physics-informed neural networks (PINNs) and an active contour model called snake. It is fast and computationally lightweight because it employs cubic splines instead of a deep neural network as a basis function. Its training parameters are physically interpretable because they directly represent control knots of the segmentation curve. Traditional snakes involve minimization of the edge-based loss functionals by deriving the Euler–Lagrange equation followed by its numerical solution. However, PICS directly minimizes the loss functional, bypassing the Euler Lagrange equations. It is the first snake variant to minimize a region-based loss function instead of traditional edge-based loss functions. PICS uniquely models the three-dimensional (3D) segmentation process with an unsteady partial differential equation (PDE), which allows accelerated segmentation via transfer learning. To demonstrate its effectiveness, we apply PICS for 3D segmentation of the left ventricle on a publicly available cardiac dataset. We also demonstrate PICS’s capacity to encode the prior shape information as a loss term by proposing a new convexity-preserving loss term for left ventricle. Overall, PICS presents several novelties in network architecture, transfer learning, and physics-inspired losses for image segmentation, thereby showing promising outcomes and potential for further refinement.

Abstract 7393: Tumor model to tumor treatment: Applying deep learning approaches to map multimodal data from cancer model systems to patients

Abstract Despite large-scale efforts to measure the effect of drug screens in cancer cell lines, mapping the effects of drugs to patient samples has been a challenge. Biological differences between cell lines and patients, such as lack of immune system or microbiome, in-vitro survival adaptations, and biases in measurement technologies create differences across sample modalities that can confound analysis including prediction with machine learning. In this work, we propose a multiway batch correction strategy to enable algorithmic prediction of tumor drug response across model systems and patient data.Recent advances in batch correction algorithms have been motivated by the need to correct for batch effects in single-cell omics and include diverse approaches such as variational autoencoders (VAEs) and generative adversarial networks (GANs). Given the successes of these generative deep learning methods in single cell sequencing analysis, we worked to employ similar approaches to correct large omics measurements across various cancer datasets. Here, we describe mapping of datasets from diverse data sources and model systems to the same space, so that a predictive model of drug response built in a system such as cell lines can be used in biologically relevant models such as organoids, patient derived xenografts, and tumor data. Specifically, we introduce a modified loss function in a VAE using cosine similarity distance to minimize the effect of different cancer model systems in predicting cancer types. We evaluate the method on standard data types for drug response prediction - gene expression, copy number variation, and protein abundance. For this method, the cosine similarity is added as an additional term to the VAE reconstruction and Kullback-Leibler divergence loss terms. This injects a quantification of the dissimilarity between the tumor and tumor model distributions into the backpropagation and gradient descent for updating the model parameters resulting in an encoded representation of the data where the effect of data source has been attenuated while preserving the phenotypic signal. We evaluate our approach for biological signal preservation while reducing model system-specific noise with logistic regression and Euclidean distance. Our results show that the proposed VAE can effectively correct for platform effects and improve the accuracy of downstream integrative analyses. This study has the potential to improve the accuracy and translatability of proteogenomic drug response studies. The proposed modified VAE could be used to correct for platform effects in a variety of datasets, including those from different studies, different platforms, and different cancer types. This could lead to new insights into cancer biology, calibration of cancer patient digital twins, and the development of new diagnostic and therapeutic strategies. Citation Format: Brian Karlberg, Raphael Kirchgaessner, Jeremy R. Jacobson, Kyle Ellrott, Sara J. Gosline. Tumor model to tumor treatment: Applying deep learning approaches to map multimodal data from cancer model systems to patients [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 7393.

An OpenFOAM multi-region solver for tritium transport modeling in fusion systems

The transport, permeation and retention of tritium inside nuclear fusion reactors is a topic of great interest due to the scarcity of the isotope, its ease of diffusion through materials and its radioactivity. An accurate balance of tritium is needed in all the fuel cycle, and each loss term needs to be evaluated in detail. In this context, reliable and flexible tools to evaluate tritium permeation and retention are a necessity for the development of fusion technologies. This work presents the OpenFOAM PermeAtion Solver for Tritium Analysis Foam (pastaFoam) solver and two custom boundary conditions, along with a series of verification and validation cases. The solver inherits all capabilities of the base solver, chtMultiRegionFoam, and is capable to simulate hydrogen transport in coupled fluid–solid systems in the presence of hydrogen traps, under diffusion limited regime or accounting for surface effects. All features are tested against analytical solutions and results are compared with other tritium transport codes. Agreement with experimental data is aligned with results of numerical benchmarks from literature.

Displacement Reconstruction Based on Physics-Informed DeepONet Regularizing Geometric Differential Equations of Beam or Plate

Physics-informed DeepONet (PI_DeepONet) is utilized for the reconstruction task of structural displacement based on measured strain. For beam and plate structures, the PI_DeepONet is built by regularizing the strain–displacement relation and boundary conditions, referred to as geometric differential equations (GDEs) in this paper, and the training datasets are constructed by modeling strain functions with mean-zero Gaussian random fields. For the GDEs with more than one Neumann boundary condition, an algorithm is proposed to balance the interplay between different loss terms. The algorithm updates the weight of each loss term adaptively using the back-propagated gradient statistics during the training process. The trained network essentially serves as a solution operator of GDEs, which directly maps the strain function to the displacement function. We demonstrate the application of the proposed method in the displacement reconstruction of Euler–Bernoulli beams and Kirchhoff plates, without any paired strain–displacement observations. The PI_DeepONet exhibits remarkable precision in the displacement reconstruction, with the reconstructed results achieving a close proximity, surpassing 99%, to the finite element calculations.

HyperPredict: Estimating Hyperparameter Effects for Instance-Specific Regularization in Deformable Image Registration

Methods for medical image registration infer geometric transformations that align pairs, or groups, of images by maximising an image similarity metric. This problem is ill-posed as several solutions may have equivalent likelihoods, also optimising purely for image similarity can yield implausible deformable transformations. For these reasons regularization terms are essential to obtain meaningful registration results. However, this requires the introduction of at least one hyperparameter, often termed λ, which serves as a trade-off between loss terms. In some approaches and situations, the quality of the estimated transformation greatly depends on hyperparameter choice, and different choices may be required depending on the characteristics of the data. Analyzing the effect of these hyperparameters requires labelled data, which is not commonly available at test-time. In this paper, we propose a novel method for evaluating the influence of hyperparameters and subsequently selecting an optimal value for given pair of images. Our approach, which we call HyperPredict, implements a Multi-Layer Perceptron that learns the effect of selecting particular hyperparameters for registering an image pair by predicting the resulting segmentation overlap and measures of deformation smoothness. This approach enables us to select optimal hyperparameters at test time without requiring labelled data, removing the need for a one-size-fits-all cross-validation approach. Furthermore, the criteria used to define optimal hyperparameter is flexible post-training, allowing us to efficiently choose specific properties (e.g. overlap of specific anatomical regions of interest, smoothness/plausibility of the final displacement field). We evaluate our proposed method on the OASIS brain MR standard benchmark dataset using a recent deep learning approach (cLapIRN) and an algorithmic method (Niftyreg). Our results demonstrate good performance in predicting the effects of regularization hyperparameters and highlight the benefits of our image-pair specific approach to hyperparameter selection.

Multidimensional Q‐compensated reverse time migration using a high‐efficient decoupled viscoacoustic wave equation

AbstractSeismic waves propagating through attenuating media induce amplitude loss and phase dispersion. Neglecting the attenuation effects during seismic processing results in the imaging profiles with weakened energy, mispositioned interfaces and reduced resolution. To obtain high‐quality imaging results, Q‐compensated reverse time migration is developed. The kernel of the Q‐compensated reverse time migration algorithm is a viscoacoustic wave equation with decoupled amplitude loss and phase dispersion terms. However, the majority of current decoupled viscoacoustic wave equations are solved using the computationally expensive pseudo‐spectral method. To enhance computational efficiency, we initiate our approach from the dispersion relation of a single standard linear solid model. Subsequently, we derive a novel decoupled viscoacoustic wave equation by separating the amplitude loss and phase dispersion terms, previously coupled in the memory variable. The newly derived decoupled viscoacoustic wave equation can be efficiently solved using the finite‐difference method. Then, we reverse the sign of the amplitude loss term of the newly derived viscoacoustic wave equation to implement high‐efficient Q‐compensated reverse time migration based on the finite‐difference method. In addition, we design a regularization term to suppress the high‐frequency noise for stabilizing the wavefield extrapolation. Forward modelling tests validate the decoupled amplitude loss and phase dispersion characteristics of the newly derived viscoacoustic wave equation. Numerical examples in both two‐dimensional and three‐dimensional confirm the effectiveness of the Q‐compensated reverse time migration based on the finite‐difference algorithm in mitigating the attenuation effects in subsurface media and providing high‐quality imaging results.

The impact of metformin on weight and metabolic parameters in patients with obesity: A systematic review and meta-analysis of randomized controlled trials.

There are conflicting data on the weight-reducing potential of metformin (MTF) in nondiabetic patients with obesity. The purpose of this systematic review and meta-analysis was to evaluate the effect of MTF on weight and cardiometabolic parameters in adults with overweight/obesity with or without nonalcoholic fatty liver disease (NAFLD) (CRD42018085512). We included randomized controlled trials (RCTs) in adults without diabetes mellitus, with mean body mass index (BMI) ≥ 25 kg/m2, with or without NAFLD, comparing MTF to placebo/control, lifestyle modification (LSM) or a US Food and Drug Administration-approved anti-obesity drug, reporting on weight or metabolic parameters, and extending over at least 3 months. We conducted a systematic search in MEDLINE, EMBASE, PubMed and the Cochrane Library without time limitation (until March 2022). We screened and selected eligible articles, abstracted relevant data, and assessed the risk of bias. All steps were in duplicate and independently. We conducted a random-effects model meta-analysis using Review Manager version 5.3, with prespecified subgroup analyses in case of heterogeneity. We identified 2650 citations and included 49 trials (55 publications). Compared to placebo, MTF was associated with a significant reduction in BMI (mean difference [MD] -0.56 [-0.74, -0.37] kg/m2; p < 0.0001), at doses ranging from 500 to 2550 mg/day, and with a significant percentage change in BMI of -2.53% (-2.90, -2.17) at the dose 1700 mg/day. There was no interaction by baseline BMI, MTF dose or duration, nor presence or absence of NAFLD. There was no significant difference between MTF and LSM. Orlistat was more effective than MTF (at doses of 1000-1700 mg/day) in terms of weight loss, with an MD in BMI of -3.17 (-5.88; -0.47) kg/m2, favouring the former. Compared to placebo/control, MTF improved insulin parameters, while no effect was detected when compared to LSM. A few small trials showed heterogenous effects on liver parameters in patients with NAFLD treated with MTF compared to placebo/control. There was a large variability in the expression of outcome measures and RCTs were of low quality. In conclusion, MTF was associated with a modest weight reduction in obese nondiabetic patients. Further high-quality and better powered studies are needed to examine the impact of MTF in patients with insulin resistance and NAFLD.