Joint Prediction Model Research Articles

Development of a joint prediction model based on both the radiomics and clinical factors for preoperative prediction of circumferential resection margin in middle-low rectal cancer using T2WI images.

A circumferential resection margin (CRM) is an independent risk factor for local recurrence, distant metastasis, and poor overall survival of rectal cancer. In this study, we developed and validated a radiomics prediction model to predict perioperative surgical margins in patients with middle and low rectal cancer following neoadjuvant treatment and for decisions about treatment plans for patients. This study retrospectively analyzed 275 patients from center 1(training cohort) and 120 patients from center 2(verification cohort) with rectal cancer diagnosed at two centers from July 2020 to July 2022 who underwent neoadjuvant therapy and had their CRM status confirmed by preoperative high-resolution magnetic resonance imaging (MRI) scans. Radiomics signatures were extracted and screened from MRI images and a radiomics signature was built by the least absolute shrinkage and selection operator (LASSO) logistic regression model, which was combined with clinical signatures to construct a nomogram. The receiver operating characteristic (ROC) curve and area under the curve (AUC) value, sensitivity, specificity, positive predictive value, negative predictive value, and calibration curve were used to evaluate the predictive performance of the model. In our research, the combined model has the best performance. In the training group, the radiomics model based on high-spatial-resolution T2-weighted imaging (HR-T2WI), clinical model and combined model demonstrated an AUC of 0.819 (0.802-0.833), 0.843 (0.822-0.861), and 0.910 (0.880-0.940), respectively. In the validation group, they demonstrated an AUC of 0.745 (0.715-0.788), 0.827 (0.798-0.850), and 0.848 (0.779-0.917), respectively. The calibration curve confirmed the clinical applicability of the model. The individualized prediction model established by combining radiomics signatures and clinical signatures can efficiently and objectively predict perioperative margin invasion in patients with middle and low rectal cancer.

Grey correlation analysis of macro- and micro-scale properties of aeolian sand concrete under the salt freezing effect

To study the effect of the pore structure of aeolian sand concrete (ASC) with different compositions on its compressive strength in a freeze-thaw environment, ASC with 0% (S0), 25% (S25), 50% (S50), 75% (S75), and 100% (S100) of aeolian sand (AS) content was prepared in our experiment. The damage and degradation laws of the compressive strength of ASC cube in the presence of chloride freeze-thaw were studied, and the compressive strengths of S0, S50 and S100 were predicted from the pore structure parameters. Compressive strength, nuclear magnetic resonance (NMR) and scanning electron microscopy (SEM) tests of the ASC samples were carried out. Grey correlation analysis was used to analyze the correlation between the pore structure parameters and the compressive strength, and a model was developed to predict the compressive strength of S0, S50 and S100 ASC subjected to freeze-thaw cycles (FTCs). It was shown that ASC with 25% AS content showed a large improvement in compressive strength, which was 6.86% higher than that of the S0 group. With the progression of FTCs, the porosity of ASC increases, and the bound fluid saturation and the proportion of 0–0.1 µm pores decrease. According to the grey correlation analysis, the saturation of bound fluid and the proportion of 0–0.1 µm pores have the strongest correlation with the compressive strength of ASC. The joint prediction model for the compression strength has a correlation coefficient of 0.937, and is able to better predict the compressive strength of ASC at different AS replacement rates during the FTCs.

Exploiting local label correlation from sample perspective for multi-label classification via three-way decision theory

In multi-label classification, the expansion of output dimension seriously interferes learning performance, and even fails to build a joint prediction model. In order to restrain the proliferation of multi-label classifier’s hypothesis space, the current works focus on the application of global positive label correlation. However, the “black or white” mechanism ignore other possible forms of label correlation, such as negative or neutral correlation. By introducing the doctrine of the mean, three-way decision (3WD) theory provides a solution for in-depth research on local label correlation, and aims to handle the uncertainty of multi-label learning tasks. In this paper, a novel learning algorithm for multi-label joint classification, namely ML-3WD, is proposed by considering the 3WD label correlation from the perspective of samples. According to the weights of different features on any label, the comprehensive loss of each sample to three action strategies can be measured. Obviously, the 3WD rules for any label variable in multi-label output space is obtained. By aggregating the cutting thresholds between different labels, the division principles of 3WD label correlation are further established. Given any multi-label sample, the local fuzzy membership to co-occurrence or mutual state for label pair is examined based on kernelized fuzzy rough sets. The 3WD local label relevance of each sample is confirmed, that is, positive, negative or neutral. The global application strategy for multi-label classification is utilized to avoid over-fitting induced by local mining strategy. Based on the integral mean of the distribution of 3WD local label relevance in multi-label sample space, two different versions of empirical label relevance are constructed. By constraining the relative position between sub-separation hyperplanes, the 3WD label correlation distribution-based model for multi-label joint classification is designed. The experiment results on fifteen real world multi-label datasets reflect that our algorithm achieves good classification ability and versatility. The impact of core parameters on learning performance is also dissected.

Federated Semi-Supervised Learning for Medical Image Segmentation via Pseudo-Label Denoising.

Distributed big data and digital healthcare technologies have great potential to promote medical services, but challenges arise when it comes to learning predictive model from diverse and complex e-health datasets. Federated Learning (FL), as a collaborative machine learning technique, aims to address the challenges by learning a joint predictive model across multi-site clients, especially for distributed medical institutions or hospitals. However, most existing FL methods assume that clients possess fully labeled data for training, which is often not the case in e-health datasets due to high labeling costs or expertise requirement. Therefore, this work proposes a novel and feasible approach to learn a Federated Semi-Supervised Learning (FSSL) model from distributed medical image domains, where a federated pseudo-labeling strategy for unlabeled clients is developed based on the embedded knowledge learned from labeled clients. This greatly mitigates the annotation deficiency at unlabeled clients and leads to a cost-effective and efficient medical image analysis tool. We demonstrated the effectiveness of our method by achieving significant improvements compared to the state-of-the-art in both fundus image and prostate MRI segmentation tasks, resulting in the highest Dice scores of 89.23% and 91.95% respectively even with only a few labeled clients participating in model training. This reveals the superiority of our method for practical deployment, ultimately facilitating the wider use of FL in healthcare and leading to better patient outcomes.

A combined wind-PV power prediction method based on clustering prescreening

Due to the volatility caused by the high uncertainty of renewable energy, the large-scale integration of renewable energy into the electric power grid brings huge challenges to the scheduling and operation of the electric power system. Effective power prediction of power source can not only relieve the pressure of power system peaking and frequency regulation, but also improve the accuracy of decision-making. So as to make the most of the information between wind and photovoltaic power source, this paper puts forward a wind-PV joint power forecasting model. First, the raw power data is pre-screened by clustering to select the closer connection between wind and PV power; next, the screened data are passed through the upper CNN structure respectively after pre-processed. Then, the outputs from CNN modules are fed together into the prediction layer, which include LSTM layers and fully connected layers. Finally, the proposed wind-PV joint prediction model is verified by simulation experiments.

Development of risk prediction models for depression combining genetic and early life risk factors.

Both genetic and early life risk factors play important roles in the pathogenesis and progression of adult depression. However, the interplay between these risk factors and their added value to risk prediction models have not been fully elucidated. Leveraging a meta-analysis of major depressive disorder genome-wide association studies (N = 45,591 cases and 97,674 controls), we developed and optimized a polygenic risk score for depression using LDpred in a model selection dataset from the UK Biobank (N = 130,092 European ancestry individuals). In a UK Biobank test dataset (N = 278,730 European ancestry individuals), we tested whether the polygenic risk score and early life risk factors were associated with each other and compared their associations with depression phenotypes. Finally, we conducted joint predictive modeling to combine this polygenic risk score with early life risk factors by stepwise regression, and assessed the model performance in identifying individuals at high risk of depression. In the UK Biobank test dataset, the polygenic risk score for depression was moderately associated with multiple early life risk factors. For instance, a one standard deviation increase in the polygenic risk score was associated with 1.16-fold increased odds of frequent domestic violence (95% CI: 1.14-1.19) and 1.09-fold increased odds of not having access to medical care as a child (95% CI: 1.05-1.14). However, the polygenic risk score was more strongly associated with depression phenotypes than most early life risk factors. A joint predictive model integrating the polygenic risk score, early life risk factors, age and sex achieved an AUROC of 0.6766 for predicting strictly defined major depressive disorder, while a model without the polygenic risk score and a model without any early life risk factors had an AUROC of 0.6593 and 0.6318, respectively. We have developed a polygenic risk score to partly capture the genetic liability to depression. Although genetic and early life risk factors can be correlated, joint predictive models improved risk stratification despite limited improvement in magnitude, and may be explored as tools to better identify individuals at high risk of depression.

Predictive value of preoperative inflammatory indexes for postoperative early recurrence of hepatitis B-related hepatocellular carcinoma.

To investigate the predictive value of preoperative neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), systemic inflammation response index (SIRI), and systemic immune inflammation index (SII) for early recurrence after liver resection in patients with hepatitis B-related hepatocellular carcinoma. A retrospective study was conducted on 162 patients who underwent hepatitis B-related hepatocellular carcinoma (HCC) resection between January 2013 and April 2016. The Youden index was utilized to calculate the optimal cut-off value. The Pearson Chi-square test was applied to analyze the relationship between inflammatory indexes and common clinical and pathological features. The Kaplan-Meier method and Log-Rank test were implemented to compare the recurrence-free survival rate within 2 years of the population. The Cox regression analysis was used to identify the risk factors for early postoperative recurrence. The best cut-off values of SIRI, PLR, NLR and SII were 0.785, 86.421, 2.231 and 353.64, respectively. Tumor diameter, degree of tumor differentiation, vascular invasion, SIRI>0.785, PLR>86.421, NLR>2.231 and SII>353.64 were risk factors for early recurrence. Combining the above seven risk factors to construct a joint index, the AUC of the joint prediction model was 0.804. The areas under the ROC curves of SIRI, PLR, NLR, and SII were 0.659, 0.725, 0.680, and 0.723, respectively. There was no significant difference in the predictive ability between the single inflammatory index models, but the predictive performance of the joint prediction model was significantly higher than that of the single inflammatory index models. The patients with lower SIRI, PLR, NLR, SII and joint index value had longer recurrence-free survival within 2 years. The joint index CIP, constructed by combining preoperative SIRI, PLR, NLP and SII with pathological features, can better predict the early recurrence of HBV-related HCC patients after surgery, which is beneficial in identifying high-risk patients and assisting clinicians to make better clinical choices.

Construction and validation of a survival prognostic model for stage III hepatocellular carcinoma: a real-world, multicenter clinical study

ObjectiveTo construct a survival prediction model for patients with TNM stage III hepatocellular carcinoma (HCC) to guide the clinical diagnosis and treatment of HCC patients and improve prognosis.MethodsBased on data from patients with stage III (AJCC 7th TNM stage) recorded by the American Institute of Cancer Research from 2010 to 2013, risk factors affecting the prognosis were screened by Cox univariate and multivariate regression, line plots was constructed, and the credibility of the model was verified by Boostrap method. ROC operating curves, calibration curves and DCA clinical decision curves were used to evaluate the model, and Kaplan–Meier was used for survival analysis was used to evaluate the efficacy of the model. External survival data from patients newly diagnosed with stage III hepatocellular carcinoma during 2014–2015 were used to validate and fit the model and to optimize the model.ResultsAge > 75 years vs.18-53 years [HR = 1.502; 95%CI(1.134–1.990)], stage IIIC vs. Stage IIIA [HR = 1.930; 95%CI(1.509–2.470)], lobotomy vs. non-surgery [HR = 0.295; 95%CI(0.228–0.383)], radiotherapy vs. non-radiotherapy [HR = 0.481; 95%CI(0.373–0.619)], chemotherapy vs. Non-chemotherapy [HR = 0.443; 95%CI(0.381–0.515)], positive serum AFP before treatment vs. negative [HR = 1.667; 95%CI(1.356–2.049)], the above indicators are independent prognostic factors for patients with stage III hepatocellular carcinoma, and the P values for the above results were less than 0.05. A joint prediction model was constructed based on age, TNM stage, whether and how to operate, whether to receive radiotherapy, whether to receive chemotherapy, pre-treatment serum AFP status and liver fibrosis score. The consistency index of the improved prognosis model was 0.725.ConclusionsThe traditional TNM staging has limitations for clinical diagnosis and treatment, while the Nomogram model modified by TNM staging has good predictive efficacy and clinical significance.

Using Quantum hybrid join prediction model to predict future trajectory motion during transit train movement

Predicting future train motion from the yard to the main track via an autonomous driving train is an essential task in a terminal railway station. Existing models fail to address the multiple agents that cross the track as the train transits. Still, there is available research in this area, such as how to predict scene-compliant trajectories across multiple agents jointly. The number of agents significantly increases prediction space as the number of challenges grows exponentially. This study focuses on joint motion prediction between an interacting agent and a transit train. The complex joint prediction problem is divided into marginal sub-prediction issues. In the joint prediction space where the marginal prediction model and conditional prediction model are processed, our Quantum Hybrid Joint Prediction (QHJP) model effectively classifies both influencers and reactors. The proposed model combined the interacting agent features to improve prediction likelihood at the joint relative motion. We test and analyse our model's effectiveness as subjected to the acquired database from publicly real-world surveillance camera records from terminal railway junctions. We recreate a simulation view of an authentic, interactive agent scenario and compare the evaluation result with the well-known prediction benchmark Waymo Open Motion (WOM) dataset for experimental purposes.

Contrast-enhanced ultrasonography-based renal blood perfusion in brain-dead donors predicts early graft function.

The aim of this study was to quantify renal microcirculatory perfusion in braindead donors using contrast-enhanced ultrasonography (CEUS), and to establish an accurate, noninvasive, and convenient index for predicting delayed graft function (DGF) post-transplantation. In total, 90 brain-dead donor kidneys (training group, n=60; validation group, n=30) examined between August 2020 and November 2022 were recruited in this prospective study. CEUS was performed on the kidneys of brain-dead donors 24 hours before organ procurement and time-intensity curves were constructed. The main measures were arrival time, time to peak, and peak intensity of the kidney segmental arteries, cortex, and medulla. Recipients were divided into DGF and non-DGF groups according to early post-transplant graft function. The area under the receiver operating characteristic curve (AUC) was used to assess diagnostic performance. The arrival time of the kidney segmental artery and cortex and the time interval between the time to peak of the segmental artery and cortex were identified as independent factors associated with DGF by multivariate stepwise regression analysis. A new index for the joint prediction model of three variables, the contrast-enhanced ultrasonography/Kidney Donor Profile index (CEUS-KDPI), was developed. CEUS-KDPI showed high accuracy for predicting DGF (training group: AUC, 0.91; sensitivity, 90.5%; specificity, 92.3%; validation group: AUC, 0.84; sensitivity, 75.0%; specificity, 92.3%). CEUS-KDPI accurately predicted DGF after kidney transplantation. CEUS may be a potential noninvasive tool for bedside examinations before organ procurement and may be used to predict early renal function after kidney transplants kidneys from donors after brain death.

Value of clinical features combined with multimodal ultrasound in predicting lymph node metastasis in cervical central area of papillary thyroid carcinoma.

To explore the clinical features, multimodal ultrasound features and multimodal ultrasound imaging features in predicting lymph node metastasis in the central cervical region of papillary thyroid carcinoma. A total of 129 patients with papillary thyroid carcinoma (PTC) confirmed by pathology were selected from our hospital from September 2020 to December 2022. According to the pathological results of cervical central lymph nodes, these patients were divided into metastatic group and non-metastatic group. Patients were randomly sampled and divided into training group (n = 90) and verification group (n = 39) according to the ratio of 7:3. The independent risk factors for central lymph node metastasis (CLNM) were determined by least absolute shrinkage and selection operator and multivariate logistic regression. Based on independent risk factors to build a prediction model, select the best diagnostic effectiveness of the prediction model sketch line chart, and finally, the line chart calibration and clinical benefits were evaluated. A total of 8, 11 and 17 features were selected from conventional ultrasound images, shear wave elastography (SWE) images and contrast-enhanced ultrasound (CEUS) images to construct the Radscore of conventional ultrasound, SWE and CEUS, respectively. After univariate and multivariate logistic regression analysis, male, multifocal, encapsulation, iso-high enhancement and multimodal ultrasound imaging score were independent risk factors for cervical CLNM in PTC patients (p < 0.05). Based on independent risk factors, a clinical combined with multimodal ultrasound feature model was constructed, and multimodal ultrasound Radscore were added to the clinical combined with multimodal ultrasound feature model to form a joint prediction model. In the training group, the diagnostic efficacy of combined model (AUC = 0.934) was better than that of clinical combined with multimodal ultrasound feature model (AUC = 0.841) and multimodal ultrasound radiomics model (AUC = 0.829). In training group and validation group, calibration curves show that the joint model has good predictive ability for cervical CLNM of PTC patients; The decision curve shows that most of the net benefits of the nematic chart are higher than those of clinical + multimodal ultrasound feature model and multimodal ultrasound radiomics model within a reasonable risk threshold range. Male, multifocal, capsular invasion and iso-high enhancement are independent risk factors of CLNM in PTC patients, and the clinical plus multimodal ultrasound model based on these four factors has good diagnostic efficiency. The joint prediction model after adding multimodal ultrasound Radscore to clinical and multimodal ultrasound features has the best diagnostic efficiency, high sensitivity and specificity, which is expected to provide objective basis for accurately formulating individualized treatment plans and evaluating prognosis.

Improving the spaceborne GNSS-R altimetric precision based on the novel multilayer feedforward neural network weighted joint prediction model

Global navigation satellite system-reflection (GNSS-R) sea surface altimetry based on satellite constellation platforms has become a new research direction and inevitable trend, which can meet the altimetric precision at the global scale required for underwater navigation. At present, there are still research gaps for GNSS-R altimetry under this mode, and its altimetric capability cannot be specifically assessed. Therefore, GNSS-R satellite constellations that meet the global altimetry needs to be designed. Meanwhile, the matching precision prediction model needs to be established to quantitatively predict the GNSS-R constellation altimetric capability. Firstly, the GNSS-R constellations altimetric precision under different configuration parameters is calculated, and the mechanism of the influence of orbital altitude, orbital inclination, number of satellites and simulation period on the precision is analyzed, and a new multilayer feedforward neural network weighted joint prediction model is established. Secondly, the fit of the prediction model is verified and the performance capability of the model is tested by calculating the R2 value of the model as 0.9972 and the root mean square error (RMSE) as 0.0022, which indicates that the prediction capability of the model is excellent. Finally, using the novel multilayer feedforward neural network weighted joint prediction model, and considering the research results and realistic costs, it is proposed that when the constellation is set to an orbital altitude of 500 km, orbital inclination of 75° and the number of satellites is 6, the altimetry precision can reach 0.0732 m within one year simulation period, which can meet the requirements of underwater navigation precision, and thus can provide a reference basis for subsequent research on spaceborne GNSS-R sea surface altimetry.

Location-and-Preference Joint Prediction for Task Assignment in Spatial Crowdsourcing

With the rapid development of mobile networks and the ubiquity of mobile devices, spatial crowdsourcing (SC), which refers to assigning spatial–temporal tasks to moving workers, has drawn increasing attention. Thus, many researchers aim at various task assignment methods in SC. However, existing works generally consider workers’ location and preference categories separately. Ignorance of the correlation between them can often lead to poor assignment results. In this article, we propose a location-and-preference joint prediction model (JPM) to predict workers’ locations and preference categories jointly at each sample timestamp. Based on the predictive location probability distribution and preference probability distribution, we elaborately design a greedy multiattribute joint task assignment algorithm (MAJA) to maximize the average number of completed tasks under constraints. Then, an overall procedure incorporating the JPM and MAJA, called the location-and-preference joint prediction-based task assignment (LPJTA), is implemented to focus on assigning tasks to workers who are near the task location and willing to perform the task based on predicting locations and preference categories. We theoretically analyze the time complexity and approximation ratio of the proposed methods and construct extensive experiments on three real datasets to empirically verify their effectiveness, comparing with the state-of-the-art baselines.

Assessing groundwater storage anomalies in Beijing based on the new multifactor-quantitative joint prediction model

Water shortages and groundwater depletion are critical issues in the world, leading to unsustainable agricultural production and adverse ecological impacts. Here, the new Multifactor-Quantitative joint Prediction Model (MQPM) is developed to quantitatively predict the Groundwater Storage Anomalies (GWSA), which includes an annual multifactor module and a monthly quantitative module. The correlative coefficients from two modules between simulated GWSA and observed GWSA reach up to 0.98 and 0.87, respectively. Taking Beijing as an example, results show that GWSA trends before South-to-North Water Diversion (SNWD) (2005–2014) and after SNWD (2015–2018) are at a rate of −3.00 × 108 m3/yr and 1.95 × 108 m3/yr, respectively, which reflects the effectiveness of water diversion. Additionally, the predicted results show that GWSA from the multifactor module will increase to 45.97 × 108 m3 by 2028. The quantitative module designs four scenarios under different climate changes and policies, from which the predicted GWSA with values ranging from 28.49 × 108 m3 to 63.06 × 108 m3. For the latter module, the groundwater level will recover to ∼8.9 m up to 2028, combining multiple favorable conditions. Finally, factors consisting of water diversion, climate change, and water-saving policies have a vital influence on groundwater variations, and contributions of these factors to the GWSA account for 46%, 27%, and 27%, respectively.

The value of CT radiomics features to predict visceral pleural invasion in ≤3 cm peripheral type early non-small cell lung cancer.

To investigate predictive value of CT-based radiomics features on visceral pleural invasion (VPI) in ≤3.0 cm peripheral type early non-small cell lung cancer (NSCLC). A total of 221 NSCLC cases were collected. Among them, 115 are VPI-positive and 106 are VPI-negative. Using a stratified random sampling method, 70% cases were assigned to training dataset (n = 155) and 30% cases (n = 66) were assigned to validation dataset. First, CT findings, imaging features, clinical data and pathological findings were retrospectively analyzed, the size, location and density characteristics of nodules and lymph node status, the relationship between lesions and pleura (RAP) were assessed, and their mean CT value and the shortest distance between lesions and pleura (DLP) were measured. Next, the minimum redundancy-maximum relevance (mRMR) and least absolute shrinkage and selection operator (LASSO) features were extracted from the imaging features. Then, CT imaging prediction model, texture feature prediction model and joint prediction model were built using multifactorial logistic regression analysis method, and the area under the ROC curve (AUC) was applied to evaluate model performance in predicting VPI. Mean diameter, density, fractal relationship with pleura, and presence of lymph node metastasis were all independent predictors of VPI. When applying to the validation dataset, the CT imaging model, texture feature model, and joint prediction model yielded AUC = 0.882, 0.824 and 0.894, respectively, indicating that AUC of the joint prediction model was the highest (p < 0.05). The study demonstrates that the joint prediction model containing CT morphological features and texture features enables to predict the presence of VPI in early NSCLC preoperatively at the highest level.

GCMT: a graph-contextualized multitask spatio-temporal joint prediction model for cellular trajectories

GCMT: a graph-contextualized multitask spatio-temporal joint prediction model for cellular trajectories

Construction and assessment of a joint prediction model and nomogram for colorectal cancer

Colorectal cancer (CRC) is one of the most common tumors in the digestive system, and all its risk factors are not yet known. It is important to identify valuable clinical indicators to predict the risk of CRC. A total of 227 participants, comprising 162 healthy adults and 65 patients diagnosed with CRC at Tianjin Hospital from January 2017 to March 2022, were included in this study. Electrochemiluminescence was adopted to test the expression levels of carcinoembryonic antigen (CEA), carbohydrate antigen 19-9 (CA199). Univariate and multivariate logistic regression analyses were performed to identify independent risk factors for CRC, and a joint prediction model was then constructed. A nomogram was prepared, and the model was later assessed using the receiver operating characteristic curve and calibration curve. The univariate analysis showed that there were statistically significant differences between the two groups in terms of smoking (χ2=8.67), fecal occult blood (χ2=119.41), Helicobacter pylori (H. pylori) infection (χ2=30.87), a history of appendectomy (χ2=5.47), serum total bile acid levels (t=19.80), serum CEA levels (t=37.82), serum CA199 levels (t=6.82), and serum ferritin levels (t=54.31) (all P<0.05). The multiple logistic regression analysis showed that smoking, fecal occult blood, H. pylori infection, a history of appendectomy, serum CEA levels, and serum CA199 levels were independent risk factors for CRC (all P<0.05). Based on the above findings, a joint prediction model was constructed, and the area under the receiver operator characteristic (ROC) curve of the model was 0.842. A nomogram and calibration curve was drawn, and the internal validation results indicated that the model had good diagnostic value. Smoking, fecal occult blood, H. pylori infection, a history of appendectomy, serum CEA levels, and serum CA199 levels are independent risk factors for CRC, and the prediction model based on these factors had good predictive ability.

Robust Depth Completion with Uncertainty-Driven Loss Functions

Recovering a dense depth image from sparse LiDAR scans is a challenging task. Despite the popularity of color-guided methods for sparse-to-dense depth completion, they treated pixels equally during optimization, ignoring the uneven distribution characteristics in the sparse depth map and the accumulated outliers in the synthesized ground truth. In this work, we introduce uncertainty-driven loss functions to improve the robustness of depth completion and handle the uncertainty in depth completion. Specifically, we propose an explicit uncertainty formulation for robust depth completion with Jeffrey's prior. A parametric uncertain-driven loss is introduced and translated to new loss functions that are robust to noisy or missing data. Meanwhile, we propose a multiscale joint prediction model that can simultaneously predict depth and uncertainty maps. The estimated uncertainty map is also used to perform adaptive prediction on the pixels with high uncertainty, leading to a residual map for refining the completion results. Our method has been tested on KITTI Depth Completion Benchmark and achieved the state-of-the-art robustness performance in terms of MAE, IMAE, and IRMSE metrics.

Anomaly detection model of mooring system based on LSTM PCA method

Aiming at the difficulty of damage detection of the mooring system, this paper proposes an anomaly detection framework combined with statistical analysis and neural network methods. Specifically, based on the real design parameters of a semi-submersible platform, a hydrodynamic model was established. Taking the damage degree of 5%, 10%, 15%, and 20% as examples, the analysis of the responses was carried out when the anchor chain was damaged in different positions. And then, based on the Long Short-Term Memory (LSTM) network and normal state data, a joint prediction model of the platform's six-degree-of-freedom (6DoF) motion was established considering the marine environment loads. The motion prediction error was only 0.0078. The Principle Component Analysis (PCA) together with the prediction residual sequences was tailored and employed to build the anomaly detection model. The Upper Confident Level (UCL) of the model was also given. The recognition accuracy reached 100% which can guide the safe service of the platform.

Application of a Deep Learning Network for Joint Prediction of Associated Fluid Production in Unconventional Hydrocarbon Development

Machine learning (ML) approaches have risen in popularity for use in many oil and gas (O&amp;G) applications. Time series-based predictive forecasting of hydrocarbon production using deep learning ML strategies that can generalize temporal or sequence-based information within data is fast gaining traction. The recent emphasis on hydrocarbon production provides opportunities to explore the use of deep learning ML to other facets of O&amp;G development where dynamic, temporal dependencies exist and that also hold implications to production forecasting. This study proposes a combination of supervised and unsupervised ML approaches as part of a framework for the joint prediction of produced water and natural gas volumes associated with oil production from unconventional reservoirs in a time series fashion. The study focuses on the pay zones within the Spraberry and Wolfcamp Formations of the Midland Basin in the U.S. The joint prediction model is based on a deep neural network architecture leveraging long short-term memory (LSTM) layers. Our model has the capability to both reproduce and forecast produced water and natural gas volumes for wells at monthly resolution and has demonstrated 91 percent joint prediction accuracy to held out testing data with little disparity noted in prediction performance between the training and test datasets. Additionally, model predictions replicate water and gas production profiles to wells in the test dataset, even for circumstances that include irregularities in production trends. We apply the model in tandem with an Arps decline model to generate cumulative first and five-year estimates for oil, gas, and water production outlooks at the well and basin-levels. Production outlook totals are influenced by well completion, decline curve, and spatial and reservoir attributes. These types of model-derived outlooks can aid operators in formulating management or remedial solutions for the volumes of fluids expected from unconventional O&amp;G development.