Incomplete Data Sets Research Articles

Updating Nonlinear Stochastic Dynamics of an Uncertain Nozzle Model Using Probabilistic Learning With Partial Observability and Incomplete Dataset

Abstract This article introduces a methodology for updating the nonlinear stochastic dynamics of a nozzle with uncertain computational model. The approach focuses on a high-dimensional nonlinear computational model constrained by a small target dataset. Challenges include the large number of degrees-of-freedom, geometric nonlinearities, material uncertainties, stochastic external loads, underobservability, and high computational costs. A detailed dynamic analysis of the nozzle is presented. An updated statistical surrogate model relating the observations of interest to the control parameters is constructed. Despite small training and target datasets and partial observability, the study successfully applies probabilistic learning on manifolds (PLoM) to address these challenges. PLoM captures geometric nonlinear effects and uncertainty propagation, improving conditional mean statistics compared to training data. The conditional confidence region demonstrates the ability of the methodology to accurately represent both observed and unobserved output variables, contributing to advancements in modeling complex systems.

Predictive capability of rough set machine learning in tetracycline adsorption using biochar

Machine learning algorithms investigate relationships in data to deliver useful outputs. However, past models required complete datasets as a prerequisite. In this study, rough set-based machine learning was applied using real-world incomplete datasets to generate a prediction model of biochar’s adsorption capacity based on key attributes. The predictive model consists of if–then rules classifying properties by fulfilling certain conditions. The rules generated from both complete and incomplete datasets exhibit high certainty and coverage, along with scientific coherence. Based on the complete dataset model, optimal pyrolysis conditions, biomass characteristics and adsorption conditions were identified to maximize tetracycline adsorption capacity (> 200 mg/g) by biochar. This study demonstrates the capabilities of rough set-based machine learning using incomplete practical real-world data without compromising key features. The approach can generate valid predictive models even with missing values in datasets. Overall, the preliminary results show promise for applying rough set machine learning to real-world, incomplete data for generating biomass and biochar predictive models. However, further refinement and testing are warranted before practical implementation.

Does Early Conversion to Below-elbow Casting for Pediatric Diaphyseal Both-bone Forearm Fractures Adversely Affect Patient-reported Outcomes and ROM?

For distal forearm fractures in children, it has been shown that a below-elbow cast is an adequate treatment that overcomes the discomfort of an above-elbow cast and unnecessary immobilization of the elbow. For reduced diaphyseal both-bone forearm fractures, our previous randomized controlled trial (RCT)-which compared an above-elbow cast with early conversion to a below-elbow cast-revealed no differences in the risk of redisplacement or functional outcomes at short-term follow-up. Although studies with a longer follow-up after diaphyseal both-bone forearm fractures in children are scarce, they are essential, as growth might affect the outcome. In this secondary analysis of an earlier RCT, we asked: (1) Does early conversion from an above-elbow to a below-elbow cast in children with reduced, stable diaphyseal forearm fractures result in worse clinical and radiological outcome? (2) Does a malunion result in inferior clinical outcomes at 7.5 years of follow-up? In this study, we evaluated children at a minimum of 5 years of follow-up who were included in a previous RCT. The median (range) duration of follow-up was 7.5 years (5.2 to 9.9). The patients for this RCT were included from the emergency departments of four different urban hospitals. Between January 2006 and August 2010, we treated 128 patients for reduced diaphyseal both-bone forearm fractures. All 128 patients were eligible; 24% (31) were excluded because they were lost before the minimum study follow-up or had incomplete datasets, leaving 76% (97) for secondary analysis. The loss in the follow-up group was comparable to the included population. Eligible patients were invited for secondary functional and radiographic assessment. The primary outcome was the difference in forearm rotation compared with the uninjured contralateral arm. Secondary outcomes were the ABILHAND-kids and QuickDASH questionnaire, loss of flexion and extension of the elbow and wrist compared with the contralateral forearm, JAMAR grip strength ratio, and radiological assessment of residual deformity. The study was not blinded regarding the children, parents, and clinicians. At 7.5-year follow-up, there were no differences in ABILHAND-kids questionnaire score (above-elbow cast: 41 ± 2.4 versus above/below-elbow cast: 41.7 ± 0.7, mean difference -0.7 [95% confidence interval (CI) -1.4 to 0.04]; p = 0.06), QuickDASH (above-elbow cast: 5.8 ± 9.6 versus 2.9 ± 6.0 for above-/below-elbow cast, mean difference 2.9 [95% CI -0.5 to 6.2]; p = 0.92), and grip strength (0.9 ± 0.2 for above-elbow cast versus 1 ± 0.2 for above/below-elbow cast, mean difference -0.04 [95% CI -1 to 0.03]; p = 0.24). Functional outcomes showed no difference (loss of forearm rotation: above-elbow cast 7.9 ± 17.7 versus 4.1 ± 6.9 for above-/below-elbow cast, mean difference 3.8 [95% CI -1.7 to 9.4]; p = 0.47; arc of motion: above-elbow cast 152° ± 21° versus 155° ± 11° for the above/below-elbow cast group, mean difference -2.5 [95% CI -9.3 to -4.4]; p = 0.17; loss of wrist flexion-extension: above-elbow cast group 1.0° ± 5.0° versus 0.6° ± 4.2° for above/below-elbow cast, mean difference 0.4° [95% CI -1.5° to 2.2°]; p = 0.69). The secondary follow-up showed improvement in forearm rotation in both groups compared with the rotation at 7 months. For radiographical analysis, the only difference was in AP ulna (above-elbow cast: 6° ± 3° versus above/below-elbow cast: 5° ± 2°, mean difference 1.8° [0.7° to 3°]; p = 0.003), although this is likely not clinically relevant. There were no differences in the other parameters. Thirteen patients with persistent malunion at 7-month follow-up showed no clinically relevant differences in functional outcomes at 7.5-year follow-up compared with children without malunion. The loss of forearm rotation was 5.5ׄ° ± 9.1° for the malunion group compared with 6.0° ± 13.9° in the no malunion group, with a mean difference of 0.4 (95% CI of -7.5 to 8.4; p = 0.92). In light of these results, we suggest that surgeons perform an early conversion to a below-elbow cast for reduced diaphyseal both-bone forearm fractures in children. This study shows that even in patients with secondary fracture displacement, remodeling occurred. And even in persistent malunion, these patients mostly showed good-to-excellent final results. Future studies, such as a meta-analysis or a large, prospective observational study, would help to establish the influence of skeletal age, sex, and the severity and direction of malunion angulation of both the radius and ulna on clinical result. Furthermore, a similar systematic review could prove beneficial in clarifying the acceptable angulation for pediatric lower extremity fractures. Level I, therapeutic study.

Detracking Autoencoding Conditional Generative Adversarial Network: Improved Generative Adversarial Network Method for Tabular Missing Value Imputation.

Due to various reasons, such as limitations in data collection and interruptions in network transmission, gathered data often contain missing values. Existing state-of-the-art generative adversarial imputation methods face three main issues: limited applicability, neglect of latent categorical information that could reflect relationships among samples, and an inability to balance local and global information. We propose a novel generative adversarial model named DTAE-CGAN that incorporates detracking autoencoding and conditional labels to address these issues. This enhances the network's ability to learn inter-sample correlations and makes full use of all data information in incomplete datasets, rather than learning random noise. We conducted experiments on six real datasets of varying sizes, comparing our method with four classic imputation baselines. The results demonstrate that our proposed model consistently exhibited superior imputation accuracy.

The estimation of b-value of the frequency–magnitude distribution and of its 1σ intervals from binned magnitude data

SUMMARY The estimation of the slope (b-value) of the frequency–magnitude distribution of earthquakes is based on a formula derived by Aki decades ago, assuming a continuous exponential distribution. However, as the magnitude is usually provided with a limited resolution, its distribution is not continuous but discrete. In the literature, this problem was initially solved by an empirical correction (due to Utsu) to the minimum magnitude, and later by providing an exact formula such as that by Tinti and Mulargia, based on the geometric distribution theory. A recent paper by van der Elst showed that the b-value can be estimated also by considering the magnitude differences (which are proven to follow an exponential discrete Laplace distribution) and that in this case the estimator is more resilient to the incompleteness of the magnitude data set. In this work, we provide the complete theoretical formulation including (i) the derivation of the means and standard deviations of the discrete exponential and Laplace distributions; (ii) the estimators of the decay parameter of the discrete exponential and trimmed Laplace distributions and (iii) the corresponding formulas for the parameter b. We deduce (iv) the standard 1σ intervals for the estimated b. Moreover, we are able (v) to quantify the error associated with the Utsu minimum-magnitude correction. Furthermore, we have discussed the formulas to produce statistically independent magnitude differences. We tested extensively the b-value estimators on simulated synthetic data sets including complete catalogues as well as catalogues affected by a strong incompleteness degree such as aftershock sequences where the incompleteness is made to vary from one event to the next. We have also analysed the real aftershock sequence of the 30/10/2016 Norcia (central Italy) to integrate the finding of the simulations. To judge the performance of the various estimators we have introduced an index p that can be seen as a non-parametric extension of the Student's t index. The main outcomes of this paper are that (1) the b-value estimators devised for continuous magnitude data are not adequate for binned magnitudes, (2) for complete data sets, estimators based on magnitudes and on magnitude differences provide substantially equivalent results, (3) for incomplete magnitude data sets, estimators based on magnitude differences provide better results and (4) for incomplete aftershock sequences there is no evidence that methods based on positive magnitude differences are superior than other methods using differences. This conclusion is further confirmed by our analysis of the above-mentioned Norcia seismic sequence. This last finding contrasts with the van der Elst's claim that the so called ${{b}_ + }$ method is the most adequate to treat real aftershock sequences.

On the factor ambiguity of MCR problems for blockwise incomplete data sets

Multivariate curve resolution (MCR) methods are sometimes faced with missing or erroneous data, e.g., due to sensor saturation. In some cases, an estimation of the missing data is possible, but often MCR works with the largest submatrix without missing entries. This ignores all rows and columns of the data matrix that contain missing values. A successful approach to deal with incomplete data multisets has been proposed by Alier and Tauler (2013), but it does not include a factor ambiguity analysis. Here, the missing data problem is addressed in combination with a factor ambiguity analysis. An approach is presented that minimizes the factor ambiguity by extracting a maximum of spectral information even from incomplete rows and columns of the spectral data matrix. The method requires a high signal-to-noise ratio. Applications are presented for UV/Vis and HSI data.

Static Versus Articulating Spacer: Does Infectious Pathogen Type Affect Treatment Success?

Treatment with a static or an articulating antibiotic-containing spacer is a common strategy for treating periprosthetic joint infection (PJI), yet many patients have persistent infections after spacer treatment. Although previous studies have compared the efficacy of a static and articulating spacer for treating PJI, few studies have assessed infection control from the time of spacer implantation, or they defined treatment failure as including reinfection, reoperation, or chronic suppressive therapy. Additionally, few studies have examined whether there is an interaction between spacer and pathogen type with respect to treatment success. (1) Is there a difference in failure-free survival (defined as no reoperation, reinfection, or suppressive antibiotic therapy) between static and articulating spacers after spacer implantation for PJI? (2) Did the relationship between spacer type and failure-free survival differ by pathogen type (staphylococcal versus nonstaphylococcal and difficult-to-treat [including methicillin-resistant Staphylococcus aureus, methicillin-susceptible S. aureus , Corynebacterium, Mycobacterium, Enterococcus spp , and other gram-negative bacterium] versus not-difficult-to-treat organisms)? Between January 2014 and January 2022, a convenience sample of 277 patients was identified as having knee PJIs treated with an articulating (75% [208 of 277]) or static (25% [69 of 277]) antibiotic spacer and potentially eligible for this study. During that time, providers at our institution generally used spacers for later-presenting or chronic infections. Spacer choice was determined by surgeon preference, with static spacers used more often in instances of higher bone loss and poor soft tissue coverage. Thirty-one patients (8 static and 23 articulating spacers) were considered lost to follow-up or had incomplete datasets and were excluded from the analysis, resulting in a final analysis cohort of 246 patients: 25% (61 of 246) received a static spacer and 75% (185 of 246) received an articulating spacer. The mean ± standard deviation age of patients was 66 ± 9.9 years, BMI was 33.3 ± 6.9 kg/m 2 , and Elixhauser score was 18.1 ± 16.9. Demographic and clinical characteristics were similar between the two groups. Pathogen type was collected and categorized as staphylococcal versus nonstaphylococcal , and difficult-to-treat (including methicillin-resistant Staphylococcus aureus , methicillin-susceptible S. aureus , Corynebacterium, Mycobacterium, Enterococcus spp , and other gram-negative bacterium) versus not-difficult-to-treat, as defined by an infectious disease physician. Other variables we collected included sex, age, American Society of Anesthesiologists classification, BMI, and Elixhauser score. The primary outcome of interest was failure-free survival, which was a composite time-to-event outcome, with failure defined as reoperation, reinfection, death owing to infection, or chronic antibiotic use at a minimum of 1 year after the completion of the patient's Stage 1 postoperative antibiotic course, whichever came first. Reinfection was determined by the treating physicians in accordance with the Musculoskeletal Infection Society guidelines and included an evaluation of infectious laboratory values, cultures, and clinical signs of infection. We compared static and articulating spacers using a Cox proportional hazards model, with spacer type as the primary predictor variable. We compared staphylococcal versus nonstaphylococcal and difficult-to-treat versus not-difficult-to-treat infections by running additional models with interaction terms between spacer type and pathogen type. No difference was observed in the cause-specific hazard ratio for static versus articulating (reference) spacers (HR 1.45 [95% confidence interval 0.94 to 2.22]; p = 0.09), after adjusting for covariates. Additionally, no difference in the association between spacer type and failure-free survival was found between pathogen types or treatment difficulty after evaluating interactions (staphylococcal HR 0.37 [95% CI 0.15 to 0.91], nonstaphylococcal HR 0.79 [95% CI 0.49 to 1.28]; p value for interaction = 0.14; difficult-to-treat HR 0.37 [95% CI 0.14 to 0.99], not-difficult-to-treat HR 0.75 [95% CI 0.47 to 1.20]; p value for interaction = 0.20). The lack of a difference in failure-free survival and insufficient evidence of a difference in the association between spacer type and treatment failure by pathogen type suggests that infectious organism may not be an important consideration in the decision about spacer treatment type. Further studies should aim to elucidate which patient factors are the most influential in surgeon decision-making when choosing a spacer type in patients with PJI of the knee.Level of Evidence Level III, therapeutic study.

A novel Adaptive Zone-fusion network for precise waxberry semantic segmentation to improve automated-harvesting in a complex orchard environment

In China, waxberries are a specialty fruit that require large harvesting efforts each season. To ameliorate these issues, automated fruit-picking equipment has been widely developed within the last decade. However, most image analysis research used for waxberry segmentation contains issues such as incomplete datasets, insufficient segmentation accuracy, and susceptibility to interference. Here, we developed the Adaptive Zone-fusion Network (AZNet) that enables automated fruit harvesting and is designed to precisely segmentate waxberries in an orchard environment. Initially, we collected a diverse range of images to train and evaluate a better model, and found that AZNet demonstrated exceptional performance and achieved a mean Accuracy (mAcc) of 99.83%, mean Intersection over Union (mIoU) of 96.57%, and mean Dice coefficient (mDice) of 98.20%. To further validate AZNet’s effectiveness, we compared several state-of-the-art (SOTA) models, including DeepLab V3, SegNet, PSPNet, U-Net, RefineNet, FCN, DANet, and FastSCNN. Along with these comparisons, we also performed an ablation study to substantiate the individual impact of each module within AZNet. Through these experiments, we found that AZNet exhibits the capability to achieve accurate semantic segmentation of waxberries in complex orchard environments, and is suitable for automated fruit harvesting devices. Ultimately, AZNet improves upon previous models by constructing information fusion modules through spatial attention, channel attention, Atrous Spatial Pyramid Pooling (ASPP), and graph convolutional methods, which can be utilized in the future to enhance other fruit-picking software to reduce fruit loss and the investment of human resources.

Do We Really Need Imputation in AutoML Predictive Modeling?

Numerous real-world data contain missing values, while in contrast, most Machine Learning (ML) algorithms assume complete datasets. For this reason, several imputation algorithms have been proposed to predict and fill in the missing values. Given the advances in predictive modeling algorithms tuned in an Automated Machine Learning context (AutoML) setting, a question that naturally arises is to what extent sophisticated imputation algorithms (e.g., Neural Network based) are really needed, or we can obtain a descent performance using simple methods like Mean/Mode (MM). In this article, we experimentally compare six state-of-the-art representatives of different imputation algorithmic families from an AutoML predictive modeling perspective, including a feature selection step and combined algorithm and hyper-parameter selection. We used a commercial AutoML tool for our experiments, in which we included the selected imputation methods. Experiments ran on 25 binary classification real-world incomplete datasets with missing values and 10 binary classification complete datasets in which synthetic missing values are introduced according to different missingness mechanisms, at varying missing frequencies. The main conclusion drawn from our experiments is that the best method on average is the Denoise AutoEncoder on real-world datasets and the MissForest in simulated datasets, followed closely by MM. In addition, binary indicator variables encoding missingness patterns actually improve predictive performance, on average. Last, although there are cases where Neural-Network-based imputation significantly improves predictive performance, this comes at a great computational cost and requires measuring all feature values to impute new samples.

Electron density changes accompanying high-pressure phase transition in AlOOH

Abstract An attempt to compare and describe the differences in the electron density distribution between two phase structures of AlOOH has been made. High-resolution, high-pressure experiments with α-AlOOH diaspore were conducted using single-crystal synchrotron X-ray diffraction data. A multipole model of experimental electron density in the α-AlOOH single crystal was refined. Simultaneously, similar multipole refinement was conducted for another phase of diaspore (δ-AlOOH), this time based on a previously published data set. Both results were compared and supported by density functional theory (DFT) calculations. Although the results are affected by the limited quality of the data, it is clear that the phase transition caused significant changes in the shape and arrangement of the atomic basins. Atomic basins are a much better tool to present subtle electron density distribution changes than traditional polyhedra. Straightforward comparison of datasets available in older scientific papers and current datasets is challenging because of differences in data quality and collection parameters. However, augmenting experimental data with computational results can help reveal important information in even incomplete datasets.

Is a Rapid Recovery Protocol for THA and TKA Associated With Decreased 90-day Complications, Opioid Use, and Readmissions in a Health Safety-net Hospital?

Patients treated at a health safety-net hospital have increased medical complexity and social determinants of health that are associated with an increasing risk of complications after TKA and THA. Fast-track rapid recovery protocols (RRPs) are associated with reduced complications and length of stay in the general population; however, whether that is the case among patients who are socioeconomically disadvantaged in health safety-net hospitals remains poorly defined. When an RRP protocol is implemented in a health safety-net hospital after TKA and THA: (1) Was there an associated change in complications, specifically infection, symptomatic deep venous thromboembolism (DVT), symptomatic pulmonary embolism (PE), myocardial infarction (MI), and mortality? (2) Was there an associated difference in inpatient opioid consumption? (3) Was there an associated difference in length of stay and 90-day readmission rate? (4) Was there an associated difference in discharge disposition? An observational study with a historical control group was conducted in an urban, academic, tertiary-care health safety-net hospital. Between May 2022 and April 2023, an RRP consistent with current guidelines was implemented for patients undergoing TKA or THA for arthritis. We considered all patients aged 18 to 90 years presenting for primary TKA and THA as eligible. Based on these criteria, 562 patients with TKAs or THAs were eligible. Of these 33% (183) were excluded because they were lost before 90 days of follow-up and had incomplete datasets, leaving 67% (379) for evaluation. Patients in the historical control group (September 2014 to May 2022) met the same criteria, and 2897 were eligible. Of these, 31% (904) were excluded because they were lost before 90 days of follow-up and had incomplete datasets, leaving 69% (1993) for evaluation. The mean age in the historical control group was 61 ± 10 years and 63 ± 10 years in the RRP group. Both groups were 36% (725 of 1993 and 137 of 379) men. In the historical control group, 39% (770 of 1993) of patients were Black and 33% (658 of 1993) were White, compared with 38% (142 of 379) and 32% (121 of 379) in the RRP group, respectively. English was the most-spoken primary language, by 69% (1370 of 1993) and 68% (256 of 379) of the historical and RRP groups, respectively. A total of 65% (245 of 379) of patients in the RRP group had a peripheral nerve block compared with 54% (1070 of 1993) in the historical control group, and 39% (147 of 379) of them received spinal anesthesia, compared with 31% (615 of 1993) in the historical control group. The main elements of the RRP were standardization of preoperative visits, nutritional management, neuraxial anesthesia, accelerated physical therapy, and pain management. The primary outcomes were the proportions of patients with 90-day complications and opioid consumption. The secondary outcomes were length of stay, 90-day readmission, and discharge disposition. A multivariate analysis adjusting for age, BMI, gender, race, American Society of Anaesthesiologists class, and anesthesia type was performed by a staff biostatistician using R statistical programming. After controlling for the confounding variables as noted, patients in the RRP group had fewer complications after TKA than those in the historical control group (odds ratio 2.0 [95% confidence interval 1.3 to 3.3]; p = 0.005), and there was a trend toward fewer complications in THA (OR 1.8 [95% CI 1.0 to 3.5]; p = 0.06), decreased opioid consumption during admission (517 versus 676 morphine milligram equivalents; p = 0.004), decreased 90-day readmission (TKA: OR 1.9 [95% CI 1.3 to 2.9]; p = 0.002; THA: OR 2.0 [95% CI 1.6 to 3.8]; p = 0.03), and increased proportions of discharge to home (TKA: OR 2.4 [95% CI 1.6 to 3.6]; p = 0.01; THA: OR 2.5 [95% CI 1.5 to 4.6]; p = 0.002). Patients in the RRP group had no difference in the mean length of stay (TKA: 3.2 ± 2.6 days versus 3.1 ± 2.0 days; p = 0.64; THA: 3.2 ± 2.6 days versus 2.8 ± 1.9 days; p = 0.33). Surgeons should consider developing an RRP in health safety-net hospitals. Such protocols emphasize preparing patients for surgery and supporting them through the acute recovery phase. There are possible benefits of neuraxial and nonopioid perioperative anesthesia, with emphasis on early mobility, which should be further characterized in comparative studies. Continued analysis of opioid use trends after discharge would be a future area of interest. Analysis of RRPs with expanded inclusion criteria should be undertaken to better understand the role of these protocols in patients who undergo revision TKA and THA. Level III, therapeutic study.

Comparison between intrauterine application of an antibiotic and an herbal product to treat clinical endometritis in dairy cattle – A randomized multicentre field study

Uterine diseases are main indications for antibiotic use in dairy cows. To test a non-antibiotic treatment option, we compared the effect of an intrauterine cephapirin (Metricure®; cefapirin benzathin 500 mg per dosis; CEPH) with an intrauterine applied herbal product (25 ml of EucaComp® PlantaVet containing alcoholic extracts of Calendula officinalis L., Mellissa officinalis L., Origanum majorana L. and Eucalyptus globulus Labill. (EUC)) on the clinical cure of endometritis.Examinations of 816 cows between 21 and 35 days after calving were performed and cases of clinical endometritis (n = 169) were included. Diagnosis based on a scoring system for vaginal discharge. Study animals were randomly assigned to one of two treatment groups and treated immediately. After excluding animals with incomplete datasets, 136 cows (EUC: n = 61; CEPH: n = 75) remained for the final analysis.In total, 64% (EUC: 61%, CEPH: 67%) of analysed endometritis cases were considered as clinically cured 14 ± 2 days after the first treatment, 15% stayed uncured after the application of a maximum of two consecutive treatments, leading to an overall clinical cure rate of 85% (EUC: 82%, CEPH: 88%). No statistically difference in clinical cure rates could be observed between both treatment groups nor 14 ± 2 days after the first treatment (p = 0.956) neither regarding the overall cure rate (p = 0.923).In conclusion, the clinical cure of dairy cows' endometritis after the intrauterine application of the herbal product was non-inferior to the intrauterine application of the antibiotic cephapirin. These results could contribute to reduce the antimicrobial use in the daily veterinary routine treatment of endometritis.

Robust deep learning from incomplete annotation for accurate lung nodule detection

Deep learning plays a significant role in the detection of pulmonary nodules in low-dose computed tomography (LDCT) scans, contributing to the diagnosis and treatment of lung cancer. Nevertheless, its effectiveness often relies on the availability of extensive, meticulously annotated dataset. In this paper, we explore the utilization of an incompletely annotated dataset for pulmonary nodules detection and introduce the FULFIL (Forecasting Uncompleted Labels For Inexpensive Lung nodule detection) algorithm as an innovative approach. By instructing annotators to label only the nodules they are most confident about, without requiring complete coverage, we can substantially reduce annotation costs. Nevertheless, this approach results in an incompletely annotated dataset, which presents challenges when training deep learning models.Within the FULFIL algorithm, we employ Graph Convolution Network (GCN) to discover the relationships between annotated and unannotated nodules for self-adaptively completing the annotation. Meanwhile, a teacher–student framework is employed for self-adaptive learning using the completed annotation dataset. Furthermore, we have designed a Dual-Views loss to leverage different data perspectives, aiding the model in acquiring robust features and enhancing generalization. We carried out experiments using the LUng Nodule Analysis (LUNA) dataset, achieving a sensitivity of 0.574 at a False positives per scan (FPs/scan) of 0.125 with only 10% instance-level annotations for nodules. This performance outperformed comparative methods by 7.00%. Experimental comparisons were conducted to evaluate the performance of our model and human experts on test dataset. The results demonstrate that our model can achieve a comparable level of performance to that of human experts. The comprehensive experimental results demonstrate that FULFIL can effectively leverage an incomplete pulmonary nodule dataset to develop a robust deep learning model, making it a promising tool for assisting in lung nodule detection.

Optimizing Pneumonia Diagnosis Using RCGAN-CTL: A Strategy for Small or Limited Imaging Datasets

In response to the urgent need for efficient pneumonia diagnosis—a significant health challenge that has been intensified during the COVID-19 era—this study introduces the RCGAN-CTL model. This innovative approach combines a coupled generative adversarial network (GAN) with relativistic and conditional discriminators to optimize performance in contexts with limited data resources. It significantly enhances the efficacy of small or incomplete datasets through the integration of synthetic images generated by an advanced RCGAN. Rigorous evaluations using a wide range of lung X-ray images validate the model’s effectiveness. In binary classification tasks that differentiate between normal and pneumonia cases, RCGAN-CTL demonstrates exceptional accuracy, exceeding 99%, with an area under the curve (AUC) of around 95%. Its capabilities extend to a complex triple classification task, accurately distinguishing between normal, viral pneumonia, and bacterial pneumonia, with precision scores of 89.9%, 95.5%, and 90.5%, respectively. A notable improvement in sensitivity further evidences the model’s robustness. Comprehensive validation underscores RCGAN-CTL’s superior accuracy and reliability in both binary and triple classification scenarios. This advancement is pivotal for enhancing deep learning applications in medical diagnostics, presenting a significant tool in addressing the challenges of pneumonia diagnosis, a key concern in contemporary healthcare.

Incomplete Multiview Clustering via Semidiscrete Optimal Transport for Multimedia Data Mining in IoT

With the wide deployment of the Internet of Things (IoT), large volumes of incomplete multiview data that violates data integrity is generated by various applications, which inevitably produces negative impacts on the quality of service of IoT systems. Incomplete multiview clustering (IMC), as an essential technique of data processing, has the potential for mining patterns of incomplete IoT data. However, previous methods utilize notion-strong distances that can only measure differences between distributions at the overlap of data manifolds in fusing complementary information of data for pattern mining. They may suffer from biased estimation and information loss in capturing intrinsic structures of incomplete multiview data. To address these challenges, a semidiscrete multiview optimal transport (SD-MOT) is defined for IMC, which utilizes distances with weak notions to capture intrinsic structures of incomplete multiview data. Specifically, IMC is recast as an equivalent optimal transport between continuous incomplete multiview data and discrete clustering centroids, to avoid the strict assumption on overlap between manifolds in pattern mining. Then, SD-MOT is instantiated as a deep incomplete contrastive clustering network to remedy biased estimation and information loss on intrinsic structures of incomplete multiview data. Afterwards, a variational solution to SD-MOT is derived to effectively train the network parameters for pattern mining. Finally, extensive experiments on four representative incomplete multiview datasets verify the superiority of SD-MOT in comparison with nine baseline methods.

Rock mass quality prediction on tunnel faces with incomplete multi-source dataset via tree-augmented naive Bayesian network

Rock mass quality serves as a vital index for predicting the stability and safety status of rock tunnel faces. In tunneling practice, the rock mass quality is often assessed via a combination of qualitative and quantitative parameters. However, due to the harsh on-site construction conditions, it is rather difficult to obtain some of the evaluation parameters which are essential for the rock mass quality prediction. In this study, a novel improved Swin Transformer is proposed to detect, segment, and quantify rock mass characteristic parameters such as water leakage, fractures, weak interlayers. The site experiment results demonstrate that the improved Swin Transformer achieves optimal segmentation results and achieving accuracies of 92%, 81%, and 86% for water leakage, fractures, and weak interlayers, respectively. A multi-source rock tunnel face characteristic (RTFC) dataset includes 11 parameters for predicting rock mass quality is established. Considering the limitations in predictive performance of incomplete evaluation parameters exist in this dataset, a novel tree-augmented naive Bayesian network (BN) is proposed to address the challenge of the incomplete dataset and achieved a prediction accuracy of 88%. In comparison with other commonly used Machine Learning models the proposed BN-based approach proved an improved performance on predicting the rock mass quality with the incomplete dataset. By utilizing the established BN, a further sensitivity analysis is conducted to quantitatively evaluate the importance of the various parameters, results indicate that the rock strength and fractures parameter exert the most significant influence on rock mass quality.

Semi-supervised community detection method based on generative adversarial networks

Community detection in complex networks often suffers from insufficient data and limited utilization of prior knowledge. In this paper we propose “Semi-supervised Generative Adversarial Network” (GANSE), a novel algorithm that integrates Generative Adversarial Networks (GANs) and semi-supervised learning to address these challenges. This method addresses the issues above through a multi-step process. Initially, the network is rewired using vertex similarity metrics, thereby enhancing its structural integrity. Subsequently, a novel generative adversarial network model is designed, and our model facilitates the reconstruction of the network, thereby yielding partitions. Which form the basis for identifying core communities. Additionally, the local clustering coefficient is incorporated as a reward signal and injected into the node selection process. Moreover, isolated nodes are reallocated, ultimately culminating in the derivation of the final community structure. Experimental results on four large real-life datasets demonstrate the clear superiority of the proposed algorithm in terms of F1 and Jaccard metrics when compared to existing algorithms. Notably, our GANSE method outperforms the traditional algorithms in networks with “missing data”. Thus showing its robustness and effectiveness in real-world incomplete datasets. Our findings highlight the potential of GANs and semi-supervised learning for enhancing community detection accuracy in complex networks.

Data Imputation of Soil Pressure on Shield Tunnel Lining Based on Random Forest Model.

With the advancement of engineering techniques, underground shield tunneling projects have also started incorporating emerging technologies to monitor the forces and displacements during the construction and operation phases of shield tunnels. Monitoring devices installed on the tunnel segment components generate a large amount of data. However, due to various factors, data may be missing. Hence, the completion of the incomplete data is imperative to ensure the utmost safety of the engineering project. In this research, a missing data imputation technique utilizing Random Forest (RF) is introduced. The optimal combination of the number of decision trees, maximum depth, and number of features in the RF is determined by minimizing the Mean Squared Error (MSE). Subsequently, complete soil pressure data are artificially manipulated to create incomplete datasets with missing rates of 20%, 40%, and 60%. A comparative analysis of the imputation results using three methods-median, mean, and RF-reveals that this proposed method has the smallest imputation error. As the missing rate increases, the mean squared error of the Random Forest method and the other two methods also increases, with a maximum difference of about 70%. This indicates that the random forest method is suitable for imputing monitoring data.

Curation of myeloma observational study MALIMAR using XNAT: solving the challenges posed by real-world data

ObjectivesMAchine Learning In MyelomA Response (MALIMAR) is an observational clinical study combining “real-world” and clinical trial data, both retrospective and prospective. Images were acquired on three MRI scanners over a 10-year window at two institutions, leading to a need for extensive curation.MethodsCuration involved image aggregation, pseudonymisation, allocation between project phases, data cleaning, upload to an XNAT repository visible from multiple sites, annotation, incorporation of machine learning research outputs and quality assurance using programmatic methods.ResultsA total of 796 whole-body MR imaging sessions from 462 subjects were curated. A major change in scan protocol part way through the retrospective window meant that approximately 30% of available imaging sessions had properties that differed significantly from the remainder of the data. Issues were found with a vendor-supplied clinical algorithm for “composing” whole-body images from multiple imaging stations. Historic weaknesses in a digital video disk (DVD) research archive (already addressed by the mid-2010s) were highlighted by incomplete datasets, some of which could not be completely recovered. The final dataset contained 736 imaging sessions for 432 subjects. Software was written to clean and harmonise data. Implications for the subsequent machine learning activity are considered.ConclusionsMALIMAR exemplifies the vital role that curation plays in machine learning studies that use real-world data. A research repository such as XNAT facilitates day-to-day management, ensures robustness and consistency and enhances the value of the final dataset. The types of process described here will be vital for future large-scale multi-institutional and multi-national imaging projects.Critical relevance statementThis article showcases innovative data curation methods using a state-of-the-art image repository platform; such tools will be vital for managing the large multi-institutional datasets required to train and validate generalisable ML algorithms and future foundation models in medical imaging.Key points• Heterogeneous data in the MALIMAR study required the development of novel curation strategies.• Correction of multiple problems affecting the real-world data was successful, but implications for machine learning are still being evaluated.• Modern image repositories have rich application programming interfaces enabling data enrichment and programmatic QA, making them much more than simple “image marts”.Graphical

Dorsal-to-palmar branch neuroanastomosis in horses undergoing palmar digital neurectomy does not reduce neuroma formation or improve outcome.

Chronic foot pain, a common cause of forelimb lameness, can be treated by palmar digital neurectomy (PDN). Complications include neuroma formation and lameness recurrence. In humans, neuroanastomoses are performed to prevent neuroma formation. The aim of the study was to evaluate the outcome of horses undergoing dorsal-to-palmar branch neuroanastomosis following PDN. Retrospective case series. Eighty-five horses with PDN and dorsal-to-palmar branch neuroanastomosis. Medical records for horses undergoing this procedure at two hospitals between 2015 and 2020 were reviewed. Palmar and dorsal nerve branches of the PDN were transected and end-to-end neuroanastomosis was performed by apposition of the perineurium. Follow-up was obtained from medical records and telephone interviews. Success was defined as resolution of lameness for at least one year. Lameness resolved following surgery in 81/85 (95%) horses with 57/84 (68%) sound at one year. Postoperative complications occurred in 19/85 (22%) cases. The main limitations of the study were an incomplete data set, inaccurate owner recall, and variations in procedure. Compared to previous studies, this technique resulted in similar numbers of horses sound immediately after surgery, a comparable rate of postoperative neuroma formation but a higher recurrence of lameness rate at 1 year postoperatively. End-to-end neuroanastomosis of the dorsal and palmar branches of the PDN does not reduce the rate of neuroma formation in horses. Long-term outcome was less favorable compared to previously reported PDN techniques.