Model Assessment Research Articles

A Comprehensive Review of Facial Beauty Prediction Using Multi-task Learning and Facial Attributes

Beauty multi-task prediction from facial attributes is a multidisciplinary challenge at the intersection of computer vision, machine learning, and psychology. Despite the centrality of beauty in human perception, its subjective nature—shaped by individual, social, and cultural influences—complicates its computational modeling. This review addresses the pressing need to develop robust and fair predictive models for facial beauty assessments by leveraging deep learning techniques. Using facial attributes such as symmetry, skin complexion, and hairstyle, we explore how these features influence perceptions of attractiveness. The study adopts advanced computational methodologies, including convolutional neural networks and multi-task learning frameworks, to capture nuanced facial cues. A comprehensive analysis of publicly available datasets reveals critical gaps in diversity, biases, and ground truth annotation for training effective models. We further examine the methodological challenges in defining and measuring beauty, such as data imbalances and algorithmic fairness. By synthesizing insights from psychology and machine learning, this work highlights the potential of interdisciplinary approaches to enhance the reliability and inclusivity of automated beauty prediction systems.

Hypoxia-selective prodrug restrains tumor cells through triggering mitophagy and inducing apoptosis.

Hypoxia is a common feature of various solid tumors, which reduces the sensitivity of tumor cells to both radiotherapy and chemotherapy. However, hypoxia also presents an opportunity for tumor-selective therapy. The prodrug strategy, leveraging the hypoxic nature of the tumor microenvironment, shows significant potential for clinical application. Here we present CHD-1, a hypoxia-activated antitumor prodrug that activates in hypoxic environments, effectively inhibiting hypoxic tumor cells while exhibiting no toxicity to normoxic cells. CHD-1 impairs mitochondrial morphology and membrane potential of hypoxic tumor cells, further triggers excessive mitophagy and induces apoptosis. Moreover, prodrug CHD-1 significantly inhibits HeLa xenograft growth in vivo, and shows lower toxicity than parent molecule in an acute toxicity assessment in animal models. This study introduces a promising hypoxia-activated antitumor prodrug with strong potential for further development in hypoxic tumor therapy.

Metabolomics in Pathogenic Pathways and Targeted Therapies for Diabetic Neuropathy: A Comprehensive Review

Introduction and objective: This literature review aims to provide an overview of the progress in metabolomic assessment in animal and cell models and in humans with diabetic neuropathy (DN). Methods: Metabolomics has emerged as an important approach for investigating, identifying, and describing biomarkers related to DN. None has yet been validated for use in clinical practice. Results: DN induced significant alterations in energy metabolism and carbohydrates, lipids, amino acids, peptides, and proteins. Several treatments for DN, evaluated using metabolomics, were proved to have promising results. Conclusions: The ideal metabolite or set of metabolites that could be used as biomarkers should identify patients with diabetes prone to develop DN or those prone to progress to severe forms of sensory loss, associated with risk of ulcerations and amputation. Another potential use of a metabolite might be as an indicator of treatment response in clinical trials using agents with potential disease-modifying properties.

Fluted pumpkin waste potential as a green solid bioalkaline catalyst for neem seed oil biodiesel synthesis

Developing green routes for biodiesel synthesis is essential for sustaining environmental benignity. Thus, this present study investigated the development of a solid catalyst using fluted pumpkin (Telfairia occidentalis) pod husk as a bio-based raw material via high-temperature furnace heating, ranging from 300 – 1000 oC. The catalyst properties were investigated by conducting surface functional group, surface crystallinity, surface area, and pore distribution analyses. The efficacy of the synthesized catalyst was investigated using 28 randomized experiments carried out on transesterification of esterified neem oil, which established 12:1, 3.5 wt.%, and 60 min as optimal conditions for methanol/neem oil ratio, catalyst dosage, and reaction time, respectively, with a maximum neem seed oil methyl esters (NSOME) yield of 96.71% through response surface methodology. The statistical assessment of the developed model gave a correlation coefficient (R) of 0.9907 and a coefficient determination of 0.9816 (R2). The physicochemical properties of the NSOME agree with American and European standard limits, making it suitable for transportation and energy purposes. The catalyst could be deployed for three reaction cycles with high turnover frequencies. Hence, fluted pumpkin waste and neem oil are suitable as a green route for biodiesel synthesis.

Shear Strengthening of RC Beams with U-Wrapped FRCM Composites: State of the Art and Assessment of Available Analytical Models

Shear Strengthening of RC Beams with U-Wrapped FRCM Composites: State of the Art and Assessment of Available Analytical Models

Development and Assessment of an AI-Based Machine Learning Model for Predicting Urinary Continence and Erectile Function Recovery After Robotic-Assisted Radical Prostatectomy: Insights from a Prostate Cancer Referral Center

IntroductionProstate cancer remains a significant health concern, with radical prostatectomy being a common treatment approach. However, predicting postoperative functional outcomes, particularly urinary continence and erectile function, poses challenges. Emerging artificial intelligence (AI) technologies offer promise in predictive modeling. This study aimed to develop and validate AI-based models to predict continence and potency following nerve-sparing robotic radical prostatectomy (RARP). MethodsA cohort of 8,524 patients undergoing RARP was analyzed. Preoperative variables were collected, and two separate machine-learning Artificial Neural Network (ANN) models were trained to predict continence and potency at 12 months post- surgery. Model performance was assessed using area under the curve (AUC) values, with comparisons made to other machine learning algorithms. Feature importance analysis was conducted to identify key predictors. ResultsThe ANN models demonstrated AUCs of 0.74 for potency and 0.68 for continence prediction, outperforming other algorithms. Feature importance analysis identified variables such as age, comorbidities, and preoperative scores as significant predictors for both outcomes. ConclusionAI-based models show potential in predicting postoperative functional outcomes following RARP. Continued efforts in optimizing models and exploring additional factors are needed to improve predictive accuracy and clinical applicability. Multi-center studies and larger datasets will further contribute to enhancing the value of AI in clinical decision-making for prostate cancer treatment.

Abstract TMP118: Validation and Reliability of the SPAN Score for Assessing Neurological Function in Mouse Stroke Models following MCAO

Introduction: Neuroanatomy-based functional examinations are crucial for understanding the extent and effects of infarcts in stroke studies. Despite the importance of correlating neurological function with brain anatomy, a comprehensive neurological score that correlates with infarction volume and location following middle cerebral artery occlusion (MCAO) in mice is lacking. During the pilot phase of the Stroke Preclinical Assessment Network (SPAN), a new scoring system—SPAN score—was developed. This study aims to evaluate the reliability and validity of the SPAN score, a 9-item tool rated on a 0-3 Likert scale. Methods: The SPAN 2.0 pilot database was utilized for analysis. Reliability was assessed using Cronbach's alpha (α) and intraclass correlation coefficient (ICC) for individual items and the total score. Multivariate regression was performed for item selection and to determine the weight of each item. Concurrent validity was examined by correlating post-operative day 2 infarction volume, as measured by MRI, with the SPAN score. Results: The total SPAN score demonstrated acceptable reliability, with Cronbach's α and ICC both at 0.751. Item-total correlation coefficients of each item ranged from 0.092 to 0.638. The Pearson correlation coefficient between the total SPAN score and infarction volume was 0.539. Notably, spontaneous activity showed a reversed relationship with other items; excluding this item increased Cronbach's α to 0.766. Conclusion: The SPAN score proves to be a reliable and valid tool for neuroanatomy-based functional assessment in mouse stroke models following MCAO. Ongoing analysis will focus on refining the score by reducing the number of items, adjusting the Likert scale, incorporating lesion area, and further exploring item relationships.

Quality Assessment of the Brazilian National Bathymetric Digital Terrain Model: a Comparative Study

Quality Assessment of the Brazilian National Bathymetric Digital Terrain Model: a Comparative Study

GDFold2: A fast and parallelizable protein folding environment with freely defined objective functions.

An important step of mainstream protein structure prediction is to model the 3D protein structure based on the predicted 2D inter-residue geometric information. This folding step has been integrated into a unified neural network to allow end-to-end training in state-of-the-art methods like AlphaFold2, but is separately implemented using the Rosetta folding environment in some traditional methods like trRosetta. Despite the inferiority in prediction accuracy, the conventional approach allows for the sampling of various protein conformations compatible with the predicted geometric constraints, partially capturing the dynamic information. Here, we propose GDFold2, a novel protein folding environment, to address the limitations of Rosetta. On the one hand, GDFold2 is highly computationally efficient, capable of accomplishing multiple folding processes in parallel within the time scale of minutes for generic proteins. On the other hand, GDFold2 supports freely defined objective functions to fulfill diversified optimization requirements. Moreover, we propose a quality assessment (QA) model to provide reliable prediction on the quality of protein structures folded by GDFold2, thus substantially simplifying the selection of structural models. GDFold2 and the QA model could be combined to investigate the transition path between protein conformational states, and the online server is available at https://structpred.life.tsinghua.edu.cn/server_gdfold2.html.

Heat transfer modeling and system performance assessment of an innovative thermosyphon radiator integrated with air source heat pumps for sustainable heating applications

Heat transfer modeling and system performance assessment of an innovative thermosyphon radiator integrated with air source heat pumps for sustainable heating applications

Fish migration modeling and habitat assessment in a complex fluvial system.

Fish migration patterns are driven by hydrodynamic factors, which are essential in aquatic ecology. This study investigated the hydrodynamic drivers of Gymnocypris przewalskii fish migration in two distinct river reaches-a straight reach (SR) and a confluence reach (CR)- in the area of Qinghai Lake, China, using a 3D numerical model, fish density field data, and four predictive models. Thirteen hydrodynamic factors, with a focus on water depth and velocity, were analyzed to identify their influence on fish migration. It was found that in the SR, linear factors of flow velocity and turbulent kinetic energy were most influential, while in the CR, nonlinear factors of water temperature and vortex intensity dominated.For CR, fish migration patterns are also important nonlinear factors. Methods that accurately reveal fish migration patterns, such as Random Forest, offer higher precision for habitat assessment. Our research also shows that fish swimming ability can, to some extent, reflect migration direction. Combining fish swimming ability with traditional linear habitat assessment methods can improve the adaptability of these methods in complex fluvial system. Based on our research findings, we propose a new workflow for fish habitat assessment that integrates both linear and nonlinear predictive methods. This framework provides valuable insights for enhancing fish conservation strategies in various fluvial systems.

Sustainable process modeling and holistic 4E assessment of an innovative CHP plant with renewable hydrogen production based on multi-heat recovery and PEM electrolyzer

Sustainable process modeling and holistic 4E assessment of an innovative CHP plant with renewable hydrogen production based on multi-heat recovery and PEM electrolyzer

Understanding heavy Metal pollution dynamics in Çandarlı Gulf, western Türkiye: Insights from spatial analysis, sediment transport modeling and ecological risk assessment

Understanding heavy Metal pollution dynamics in Çandarlı Gulf, western Türkiye: Insights from spatial analysis, sediment transport modeling and ecological risk assessment

Use of Nonhuman Primates in Periodontal Disease Research: Contribution of the Caribbean Primate Research Center and Cayo Santiago Rhesus Colony.

This review article provides a historical summary regarding the use, value, and validity of the nonhuman primate model of periodontal disease. The information provided cites results regarding the features of naturally occurring periodontitis in various nonhuman primate species, as well as the implementation of a model of experimental periodontitis. Clinical similarities to human disease are discussed, as well as the use of these models to document physiological and pathophysiological tissue changes in the periodontium related to the initiation and progression of the disease. Additionally, the use of these analytics in examination of the tissue characteristics of the disease, and the utility of nonhuman primates in testing and describing various therapeutic modalities are described. As periodontitis represents a disease of an oral microbiome dysbiosis, features of the altered microbiome in the disease in nonhuman primates are related to similar findings in the human condition. The review then provides a summary of the features of local and systemic host responses to a periodontal infection in an array of nonhuman primate species. This includes attributes of innate immunity, acute and chronic inflammation, and adaptive immune responses. Finally, extensive information is presented regarding the role of Macaca mulatta derived from the Cayo Santiago community in evaluating critical biologic details of disease initiation, progression, and resolution. This unique resource afforded the capacity to relate risk and expression of disease and traits of the responses to age, sex, and matriline derivation (e.g., heritability) of the animals. The Cayo Santiago colony continues to provide a critical preclinical model for assessment of molecular aspects of the disease process that can lead to both new targets for therapeutics and consideration of vaccine approaches to preventing and/or treating this global disease.

Insulin resistance links dysbiosis of gut microbiota with cognitive impairment in first-episode drug-naïve schizophrenia.

This study aimed to explore the relationship among gut microbiota imbalance, the homeostasis model of assessment of insulin resistance (HOMA-IR) and cognitive impairments in patients with schizophrenia (SCZ). We conducted a case-control study involving 189 first-episode, drug-naïve SCZ patients and 115 healthy controls (HCs). Main methods adopted included metagenomics analysis, glucose metabolism assessment, and cognitive function evaluation using the MATRICS Consensus Cognitive Battery (MCCB). Fecal microbiota composition was analyzed via high-throughput sequencing of 16S ribosomal RNA. Patients with SCZ showed a higher likelihood of developing IR (23 %), compared to HCs (12 %). The IR group exhibited significantly higher levels of fasting blood glucose (FBG), fasting insulin (FINS), HOMA-IR, and homeostasis model assessment-β (HOMA-β), while showing lower insulin sensitivity index (ISI) levels (all p < 0.05). Patients with IR demonstrated lower scores in working memories (WM), verbal learning (HVLT) and reasoning and problem solving (RPS), compared to those without IR. Additionally, microbiota analysis revealed that IR patients had higher abundance of Negativicutes, Streptococcaceae, Enterobacteriaceae, Lachnoclostridium, Dialister, Klebsiella, and Enterobacter, and lower abundance of Flavonifractor and Rikenellaceae. Notably, Negativicutes, Streptococcaceae, Lachnoclostridium, Flavonifractor, and Rikenellaceae were shared between SCZ and IR conditions. Mediation analysis indicated that the relative abundance of Streptococcaceae have an indirect effect on WM through HOMA-IR (β=-0.148, SE=0.067, 95 %CI=-0.280 to -0.020). The study suggests that IR may play a mediating role in the relationship between gut microbiota dysbiosis and cognitive impairments in patients with SCZ, which could point to potential new avenues for therapeutic interventions.

Fat-Free Mass: Friend or Foe to Metabolic Health?

Fat mass (FM) and fat-free mass (FFM) are body composition estimates commonly reported in research studies and clinical settings. Recently, fat-free mass indexed to height (fat-free mass index; FFMI) has been shown to be positively associated with impaired insulin sensitivity or insulin resistance. Consequently, hypertrophic resistance training which can increase FFM was also questioned. This paper sets out to evaluate these propositions. In this narrative review, we discuss possible reasons that link FFMI to adverse metabolic health outcomes including the limitations of the body composition model that utilizes FFM. The safety of resistance training is also briefly discussed. Approximately 50% of FFM is comprised of skeletal muscle (SM), with the other 50% being viscera, skin, and bone; FFM and SM cannot be conflated. FFM and fat mass (FM) can both rise with increasing body weight and adiposity, indicating a positive correlation between the two compartments. Risk assessment models not adequately adjusting for this correlation may cause erroneous conclusions, however which way FM and FFM are indexed. Adipose tissue accumulation with weight gain, measured by dual-energy X-ray absorptiometry or bioelectrical impedance, can inflate FFM estimates owing to increased connective tissue. Increased adiposity can also result in fat deposition within skeletal muscle disrupting metabolic health. Importantly, non-skeletal muscle components of the FFM, i.e., the liver and pancreas, both critical in metabolic health, can also be negatively affected by the same lifestyle factors that impact SM. The most frequently used body composition techniques used to estimate FM and FFM cannot detect muscle, liver or pancreas fat infiltration. Prospective evidence demonstrates that resistance training is a safe and effective exercise modality across all ages, especially in older adults experiencing age- or disease-related declines in muscle health. The association between FFM and insulin resistance is largely an artefact driven by inadequate assessment of skeletal muscle. If FM and FFM are used, at the minimum, they need to be evaluated in context with one another. Body composition methods, such as magnetic resonance imaging, which measures skeletal muscle rather than fat-free mass, and adipose tissue as well as muscle ectopic fat, are preferred methods. Resistance training is important in achieving and maintaining good health across the lifespan. While strength and power are critical components of resistance training, the reduction of skeletal mass through ageing or disease may require hypertrophic training to mitigate and slow down the progression of this often-inevitable process.

Integrating D-Dimer Thresholds into the Revised Caprini Risk Stratification to Predict Deep Vein Thrombosis Risk in Preoperative Knee Osteoarthritis Patients.

Preoperative patients with knee osteoarthritis have a significantly increased risk of venous thromboembolism (VTE). While the Caprini risk assessment model offers some clinical guidance in predicting deep vein thrombosis (DVT), it has a relatively low predictive accuracy. Enhancing the model by integrating biomarkers, such as D-dimers, can potentially improve its accuracy. In this study, we explored the effectiveness of combining the Caprini risk model with D-dimer levels for individualized DVT risk assessment in patients with knee osteoarthritis. This retrospective cohort study included 1605 knee osteoarthritis patients scheduled for total knee arthroplasty from Peking Union Medical College Hospital, screened between January 2015 and December 2018. A revised Caprini risk stratification model was developed, and a predictive DVT model was developed based on this revised system. The sensitivity, specificity, and the area under the curve (AUC) were used to determine predictive effectiveness of the model. In the revised Caprini risk stratification, the incidence of DVT increased with higher risk levels: 2.52% in the low-risk group (scores 0-2), 2.88% in the moderate-risk group (score 3), 6.47% in the high-risk group (score 4), and 9.09% in the highest-risk group (score ≥ 5). The incidence of DVT was 3.869-fold higher in the highest-risk group and 2.676-fold higher in the high-risk group compared to the low-risk group (p = 0.013 and p = 0.014, respectively). Combining the revised Caprini risk stratification with D-dimer level demonstrated an improved AUC of 0.792, compared to D-dimer level alone (AUC 0.774) and the revised Caprini model alone (AUC 0.598). Furthermore, applying specific D-dimer thresholds across the four Caprini risk stratifications outperformed the combination of the revised Caprini model and D-dimer level in terms of AUC, specificity, and reduction in unnecessary ultrasonography. Using the Youden index, the AUC for the threshold-based method was slightly higher (0.775 vs 0.754, p = 0.310), with significantly better specificity (76.8% vs 63.6%, p < 0.001) and a greater reduction in ultrasound use (74.1% vs 61.4%). At a sensitivity of 85.5%, the differences were modest but still favored the threshold-based approach. At a sensitivity of 100%, the specificity (36.0% vs 24.7%, p < 0.001) and ultrasound reduction (34.8% vs 23.9%) were significantly better. The revised Caprini risk stratification improves preoperative DVT prediction in patients with knee osteoarthritis. Incorporating specific D-dimer thresholds into the four-level Caprini risk model enhances specificity and reduces unnecessary ultrasonography, outperforming both the use of individual indicators and the combination of the revised Caprini model with D-dimer level.

Examining pathways for a climate neutral Europe by 2050; A model comparison analysis including integrated assessment models and energy system models.

Examining pathways for a climate neutral Europe by 2050; A model comparison analysis including integrated assessment models and energy system models.

Modeling the spatiotemporal distribution, bioaccumulation, and ecological risk assessment of microplastics in aquatic ecosystems: A review.

Microplastic (MP) pollution poses a significant threat to aquatic ecosystems. Numerical modeling has emerged as an effective tool for predicting the distribution, accumulation, and risk assessment of MPs in aquatic ecosystems. However, published work has not systematically assessed the strengths and weaknesses of various modeling approaches. Therefore, we conducted a thorough review of the main modeling approaches for MPs over the past six years. We classified the approaches into three categories as: spatial and temporal distribution, bioaccumulation, and systematic ecological risk assessment. The review analyzed application scenarios, modeling methods, and the advantages and disadvantages of models. The results indicate that the accurate simulation of MPs spatial and temporal distribution requires reasonable parameterization and comprehensive transport considerations. Meanwhile, it is important to focus on coupling process models with other types of models. To enhance risk assessment models, expanding the relevant evaluation indicators is essential.

Fast optimization of the emission reduction pathways of major air pollutants in China: From the perspective of different decision preferences.

To address the concern of optimization problem of China's PM2.5 control and the limitation of computational efficiencies for traditional air quality models, we developed an integrated analysis framework to efficiently establish the identification and cost-benefit assessment of PM2.5 control pathways in China by constructing a rapid PM2.5 exposure response method based on the high-order decoupled direct method (HDDM) and coupling the sequential least square algorithm (SLSQP) and health impact assessment model. Six emission reduction scenarios with varying decision preferences were analyzed. Our study provides a methodological approach for the rapid optimization of emission pathways of major air pollutants in China with flexible options in terms of objectives and constraints, fully considering the diverse differences in environmental, health, and economic impacts among different pollution sources simultaneously. Our findings based on the multi-scenario analysis strengthen the understanding of the importance of diverse species and regions for emission reduction among various decision preferences in China, confirm the necessity of implementing differentiated control strategies for distinct pollution sources from both cost and cost-effectiveness perspectives, and indicate that accelerating PM2.5 control process in the early stage is a beneficial choice for achieving more health benefits.