Real-world Data Research Articles

Economic and Environmental Assessment of Ammonia as a Sustainable Fuel for Tugboat Operations at the Port of Texas

This study assesses the feasibility of ammonia as a sustainable alternative fuel for tugboat operations in the Port of Texas, a major U.S. maritime, oil and gas hub. Tugboats, while critical to port logistics, contribute significantly to localized air pollution and greenhouse gas emissions, impacting both environmental quality and public health in surrounding communities. Given the International Maritime Organization’s (IMO) goals to reduce carbon emissions by 40% by 2030 and 70% by 2050, exploring alternative fuels has become imperative. Ammonia, a carbon-free fuel, presents both opportunities and challenges. While ammonia combustion produces no direct CO₂ emissions, making it attractive from a decarbonization standpoint, its lower energy density results in higher fuel volumes and increased operational costs compared to marine fuel oil (MFO). This paper presents a comparative analysis based on real-world data, mathematical modeling, and emission trade-offs, examining the economic implications, environmental impact, and safety concerns of adopting ammonia as a tugboat fuel. Results indicate that while ammonia could enable significant CO₂ reductions, challenges related to fuel storage, cost, and infrastructure must be addressed. The findings suggest that with appropriate policy support and investment in handling infrastructure, ammonia can become a viable fuel option for sustainable maritime operations.

Correction: Role of neoadjuvant chemotherapy in patients with locally advanced and clinically positive nodes Upper Tract Urothelial Carcinoma treated with Nephroureterectomy: real-world data from the ROBUUST 2.0 Registry

Correction: Role of neoadjuvant chemotherapy in patients with locally advanced and clinically positive nodes Upper Tract Urothelial Carcinoma treated with Nephroureterectomy: real-world data from the ROBUUST 2.0 Registry

Retrospective Analysis of the Impact of Screening for Psoriatic Arthritis (PsA) with the Psoriasis Epidemiology Screening Tool (PEST) in Clinical Dermatology Practice

Introduction: Patients with plaque psoriasis (PSO) have an increased risk of developing PsA as the disease severity progresses. While it is estimated that 30% of patients with PSO have PsA, nearly 16% of patients with PSO may have undiagnosed PsA.1,2 Implementing the PEST questionnaire for PsA screening in dermatologic practice may increase early intervention by improving rates of timely diagnosis.3 This observational study aimed to assess the proportion of patients with PSO and no previous PsA diagnosis who screened positive on the PEST and were then diagnosed with PsA by a healthcare provider in the PEST deployment period within the Advanced Dermatology and Cosmetic Surgery (ADCS) specialty dermatology network. Procedure/Study: De-identified, real-world patient data [2018–2019 and 2021–2022] were sourced from electronic health records within the ADCS network of outpatient facilities across 14 states in the United States. Patients were included in the analysis if they received care at an ADCS facility where the PEST was deployed, had ≥1 encounter with a diagnosis code for PSO, and never had a diagnosis code for PsA before their first diagnosis code for PSO. For outcomes directly related to PEST evaluations, data from 1/1/2021–12/31/2022 (PEST deployment period) were used. For outcomes not directly related to PEST results, data from 1/1/2018–12/31/2019 (period before PEST deployment) were evaluated. Descriptive statistics, including frequencies, percentages, and 95% confidence intervals (CIs) for categorical variables, were used to characterize the patient cohorts by study period. Results: Among ADCS facilities using the PEST during the deployment period (44% of practices in the network), the final study population included 14,308 and 21,142 patients with PSO in the period before PEST deployment and during the PEST deployment period, respectively. In the PEST deployment cohort, 1,100 (5.2%; 95% CI: 4.9–5.5) patients were screened using the PEST. In the period before PEST deployment, 327 (2.3%; 95% CI: 2.0–2.5) patients with PSO and no previous PsA diagnosis were subsequently diagnosed with PsA. During the PEST deployment period, 449 (2.1%; 95% CI: 1.9–2.3) patients with PSO and no previous PsA diagnosis received a subsequent PsA diagnosis without PEST screening. Of all patients with PSO who screened positive for PsA on the PEST (score ≥3; n=296), 10.8% (n=32; 95% CI: 7.6–15.1) subsequently received a formal PsA diagnosis during the PEST deployment period. Conclusion: These results suggest that a meaningful proportion of patients with PsA may be underdiagnosed in the dermatology setting. Deployment of the PEST in the clinical dermatologic workflow may provide an opportunity to diagnose PsA earlier and improve patient outcomes.

An Intelligent Kick Detection Model for Large-Hole Ultra-Deep Wells in the Sichuan Basin

The Sichuan Basin has abundant deep and ultra-deep natural gas resources, making it a primary target for exploration and the development of China’s oil and gas industry. However, during the drilling of ultra-deep wells in the Sichuan Basin, complex geological conditions frequently lead to gas kicks, posing significant challenges to well control and safety. Compared to traditional kick detection methods, artificial intelligence technology can improve the accuracy and timeliness of kick detection. However, there are limited real-world kick data available from drilling operations, and the datasets are extremely imbalanced, making it difficult to train intelligent models with sufficient accuracy and generalization capabilities. To address this issue, this paper proposes a kick data augmentation method based on a time-series generative adversarial network (TimeGAN). This method generates synthetic kick samples from real datasets and then employs a long short-term memory (LSTM) neural network to extract multivariate time-series features of surface drilling parameters. A multilayer perceptron (MLP) network is used for data classification tasks, constructing an intelligent kick detection model. Using real drilling data from ultra-deep wells in the SY block of the Sichuan Basin, the effects of k-fold cross-validation, data dimensionality, various imbalanced data handling techniques, and the sample imbalance ratio on the model’s kick detection performance are analyzed. Ablation experiments are also conducted to assess the contribution of each module in identifying kick. The results show that TimeGAN outperforms other imbalanced data handling techniques. The accuracy, recall, precision, and F1-score of the kick identification model are highest when the sample imbalance ratio is at 1 but decrease as the imbalance ratio increases. This indicates that maintaining a balance between positive and negative samples is essential for training a reliable intelligent kick detection model. The trained model is applied during the drilling of seven ultra-deep wells in Sichuan, demonstrating its effectiveness and accuracy in real-world kick detection.

Provider Burden Associated With Apremilast Adverse Events

Introduction: In December 2021, the US Food and Drug Administration (FDA) expanded its initial approval for apremilast to include moderate and severe cases of psoriasis. However, real-world epidemiology data regarding the use of apremilast, the influence of adverse events (AEs), and the burden of AEs on managing physicians, has been lacking. This retrospective, observational, multisite study aimed to describe real-world patient characteristics, AEs (including diarrhea, nausea, and headache), and healthcare provider-level burden associated with the psoriasis treatment apremilast in the US. Methods: Dermatologists abstracted medical record data for adults treated with apremilast on or after January 1, 2018, who had ≥3 months of data available. Dermatologist and patient characteristics, including demographics and clinical characteristics, AEs, and provider burden in managing AEs were summarized using descriptive statistics. Results: Forty-seven dermatologists nationwide responded to a survey and conducted a chart review of 308 patients. Dermatologists practiced in various US geographic regions. The most common practice setting was in a private community with 2 to 5 physicians (34.0%). Most patients were White (74.4%), non-Hispanic (86.0%), and male (55.2%). The mean age at diagnosis was 43.6 years (SD, 16.8 years). AEs occurred in 29.2% of patients, with the most common being diarrhea (16.2%), followed by nausea (13.6%) and headache (4.9%). Diarrhea mostly occurred ≤30 days into treatment (87.2%) and resulted in 1 additional office visit (39.5%) and 3 phone calls per patient to their providers (34.9%). Nausea mostly occurred within 30 days (70.2%) and resulted in 1 additional visit (29.7%) and 1 phone call per patient (35.1%). The burdens of managing both diarrhea and nausea were similar, although diarrhea resulted in more phone calls to the healthcare provider. Conclusion: Diarrhea and nausea were common and resulted in an additional time and resource burden on dermatologists and their staff. Future research on the AE burden on providers and patients related to apremilast is needed.

Ixekizumab Improves Nail, Scalp, and Skin Response and Quality of Life Independent of Baseline Psoriasis Severity: Results From the Psoriasis in Special Areas (PSoSA) Study

Background: Ixekizumab has shown impressive efficacy among patients with nail, skin, and scalp psoriasis. Here, we present real-world data from the PSoSA study evaluating effectiveness of ixekizumab by cohorts of patients with mild and moderate/severe nail psoriasis. Methods: PSoSA is an ongoing, prospective, multicenter, single-arm, observational study enrolling adult patients with confirmed diagnosis of moderate-to-severe psoriasis and nail involvement, with or without scalp involvement, initiating ixekizumab in a US dermatology clinic. Assessment of improvements in nail (by modified Nail Psoriasis Severity Index [mNAPSI], with scoring range 0–130), skin (by Psoriasis Area and Severity Index [PASI], with scoring range 0–72), and scalp (by Psoriasis Scalp Severity Index [PSSI], with scoring range 0–72) response and in quality of life (by Dermatology Life Quality Index [DLQI], with scoring range 0–30) were collected at weeks (W) 4, 12, 24, and 52. Percent change was calculated for patients with non-missing baseline values. Data were stratified by nail psoriasis severity at baseline; patients with mNAPSI scores <20 were categorized as mild while those with scores ≥20 were categorized as moderate/severe. In our second interim analysis, we will be reporting data to W24. Results: Of 182 patients included in the analysis, 101 patients had baseline mNAPSI scores <20 and 81 had scores ≥20. Median percent changes from baseline in mNAPSI response at W4, W12, and W24 were −12.2, −33.3, and −71.9, respectively, for patients with baseline scores <20 and −6.1, −38.7, and −68.6, respectively, for patients with scores ≥20. Median percent changes from baseline in PASI response at W4, W12, and W24 were −60.0, −91.2, and −100.0, respectively, for patients with baseline mNAPSI scores <20 and −62.7, −91.4, and −96.7, respectively, for patients with scores ≥20. Median percent changes from baseline in PSSI response at W4, W12, and W24 were −79.3, −100.0, and −100.0, respectively, for patients with baseline mNAPSI scores <20 and −87.5, −100.0, and −100.0, respectively, for patients with scores ≥20. Median percent changes from baseline in DLQI response at W4, W12, and W24 were −66.7, −77.8, and −89.7, respectively for patients with baseline mNAPSI scores <20, and −57.1, −70.0, and −88.7, respectively, for patients with scores ≥20. Conclusions: In a real-world setting, patients with moderate-to-severe psoriasis initiating ixekizumab saw major improvements in nail, skin, and scalp response as well as quality of life as early as 4 weeks and sustained through 24 weeks, independent of nail psoriasis severity at baseline.

Machine learning approach for detection of MACE events within clinical trial data.

Randomized controlled trials (RCTs) are the gold standard for clinical research but may not accurately reflect the impact of medicines in real-world settings. Supplementing RCTs with insights from real-world data (RWD) can address known limitations by including more diverse patient populations, additional types of sites-of-care, and practices more representative of the care most people receive. One current challenge in using RWD is the lack of an algorithmic approach to identifying outcomes. To address this, machine learning models for identifying a frequently used outcome, Major Adverse Cardiovascular Events (MACE), were developed in Clinical Trial Data (CTD). Anonymized CTD sourced from the Medidata Enterprise Data Store were used to develop model features on the condition that they would be useful for labelling MACE events and that they could also be found in RWD. These features were used to train three random forest models to identify each component of 3-point MACE in a patient's clinical trial journey. Performance metrics for the models are presented (recall = 0.72 [0.07], precision = 0.68 [0.12] - mean, [SD]) along with the top contributing features. We show that the models can be tuned specifically to replicate the adjudication panels' results and present a cost-benefit analysis for deploying such models in clinical trial settings. We demonstrate the viability of using advanced algorithms for identifying clinical outcomes in prospective clinical trials. Deployment of such models could reduce the resources required to conduct RCTs. Extending such models to RWD would facilitate approval of pragmatic clinical trials for regulatory submissions.

What is the Relationship Between Time in Range, Time in Tight Range, and HbA1c in Youth and Young Adults With Type 1 Diabetes? Results From the German/Austrian/Luxembourgian/Swiss Diabetes Prospective Follow-Up Registry.

Time in range (TIR, 70-180 mg/dL) is an established marker of glycemic control. More recently, time in tight range (TTR, 70-140 mg/dL) has been proposed as well. The aim of this study was to examine the relationship between TIR, TTR, and HbA1c in youth and young adults with type 1 diabetes (T1D) in the German/Austrian/Luxembourgian/Swiss Diabetes Prospective Follow-up (DPV) registry. Data of youth and young adults aged ≤25 years with T1D for >3 months, documented in the DPV registry between 2019 and 2022 were analyzed. The most recent available HbA1c and corresponding continuous glucose monitoring (CGM) profiles in the 12 preceding weeks with at least 80% completeness were included. Associations were investigated using correlation and adjusted regression models. 1901 individuals (median age 14.0 years [IQR 10.4-16.9]) were included in the analysis. TIR and TTR correlated strongly, r = 0.965 (95% CI [0.962, 0.968]), P < .001. TTR estimates predicted from TIR were significantly higher in the group with high coefficient of variation (CV group ≥ 36%), P < .001. Correlations between TIR or TTR and HbA1c were both strong, r = -0.764 (95% CI [-0.782, -0.745]) and r = -0.777 (95% CI [-0.795, -0.759]), both P < .001, with no significant difference (P = .312) However, adjusted regression models indicated a slightly better fit for the prediction of HbA1c from TIR compared with TTR. Based on large, real-world data from a multinational registry, TIR and TTR correlated strongly, and both showed a good prediction of HbA1c. TTR estimates predicted from TIR were significantly higher in people with high glucose variability (CV).

Analysis of a fractional order epidemiological model for tuberculosis transmission with vaccination and reinfection.

This study has been carried out using a novel mathematical model on the dynamics of tuberculosis (TB) transmission considering vaccination, endogenous re-activation of the dormant infection, and exogenous re-infection. We can comprehend the behavior of TB under the influence of vaccination from this article. We compute the basic reproduction number ( ) as well as the vaccination reproduction number ( ) using the next-generation matrix (NGM) approach. The theoretical analysis demonstrates that the disease-free equilibrium point is locally asymptotically stable, and the fractional order system is Ulam-Hyers type stable. We perform numerical simulation of our model using the Adams-Bashforth 3-step method to verify the theoretical results and to show the model outputs graphically. By performing data fitting, we observe that our formulated model produces results that closely match real-world data. Our findings indicate that vaccinating a limited segment of the population can effectively eradicate the disease. The numerical simulations also show that vaccination can reduce the number of susceptible and infectious individuals in the population. Moreover, the graphical representations illustrate that the number of infected individuals rises due to both exogenous reinfection and endogenous reactivation.

Research on Well Selection Method for Intermittent Pumping in Oil Wells Based on the Analytic Network Process and Fuzzy Logic

In the later stages of oilfield development, the decline in oil well production and the increase in development costs, attributed to issues such as insufficient liquid supply, necessitate the implementation of intermittent pumping measures. However, current methods for selecting these wells lack comprehensiveness in the decision-making process. This article proposes a novel method for selecting intermittent pumping wells utilizing the analytic network process (ANP) and fuzzy logic. Initial surveys identified the main factors influencing intermittent pumping effectiveness. The ANP was employed to screen and integrate six core factors, including submergence depth and water cut. Subsequently, a fuzzy-logic-based model was developed, incorporating trapezoidal and rectangular membership functions to establish detailed correlations among the factors. The model’s efficacy was validated and tested using real-world data from the oilfield. Results indicate that the model’s assessments of intermittent pumping wells align closely with professional engineering judgments. This approach not only provides clear guidance for well selection but also demonstrates high scalability and adaptability across different oilfields by adjusting membership functions, thereby showcasing significant practical value.

Clinical Outcomes According to Aortic Stenosis Management: Insights From Real-World Practice.

Real-world data regarding clinical outcomes according to aortic stenosis (AS) management are scarce. Therefore, we aimed to investigate long-term management across the spectrum of outpatients with AS. Between May 2016 and December 2017, consecutive outpatients with mild (peak aortic velocity, 2.5-2.9 m/s), moderate (3-3.9 m/s), and severe AS (≥4 m/s) were included by 117 cardiologists in the VALVENOR (Follow-Up of a Cohort of Patients With Valvular Aortic Stenosis in the Nord-pas-de-Calais Region) study and followed-up for aortic valve replacement (AVR) and modes of death. Among 2704 patients included, 1156 (42.7%) had mild, 1121 (41.5%) moderate, and 427 (15.8%) severe AS. After a median follow-up of 5 years, 993 AVRs (488 surgical and 505 transcatheter) and 1098 deaths occurred. The 5-year cumulative incidence of AVR or of the composite of death or AVR was 13.3% and 45.2% in mild AS, 45.5% and 75.3% in moderate AS, and 62.8% and 90.6% in severe AS, respectively. Of the 292 patients who met the criteria for AVR but were not treated, AVR was considered futile in 137 patients and 155 patients refused AVR. Mortality rates after 3 years were high: 86% for anticipated futility and 72.3% for refusal. While patients at anticipated futility showed a well-balanced proportion of cardiovascular and noncardiovascular deaths, cardiovascular deaths predominated among those who refused AVR. At 5-year follow-up, only two thirds of patients with severe AS underwent AVR. Patients with untreated severe AS experienced high mortality rates, mostly cardiovascular for patients who declined AVR. This advocates for better patient education based on shared decision making and for optimizing AS quality of care, from diagnosis to treatment.

LIMUNet: A Lightweight Neural Network for Human Activity Recognition Using Smartwatches

The rise of mobile communication, low-power chips, and the Internet of Things has made smartwatches increasingly popular. Equipped with inertial measurement units (IMUs), these devices can recognize user activities through artificial intelligence (AI) analysis of sensor data. However, most existing AI-based activity recognition algorithms require significant computational power and storage, making them unsuitable for low-power devices like smartwatches. Additionally, discrepancies between training data and real-world data often hinder model generalization and performance. To address these challenges, we propose LIMUNet and its smaller variant LIMUNet-Tiny—lightweight neural networks designed for human activity recognition on smartwatches. LIMUNet utilizes depthwise separable convolutions and residual blocks to reduce computational complexity and parameter count. It also incorporates a dual attention mechanism specifically tailored to smartwatch sensor data, improving feature extraction without sacrificing efficiency. Experiments on the PAMAP2 and LIMU datasets show that the LIMUNet improves recognition accuracy by 2.9% over leading lightweight models while reducing parameters by 88.3% and computational load by 58.4%. Compared to other state-of-the-art models, LIMUNet achieves a 9.6% increase in accuracy, with a 60% reduction in parameters and a 57.8% reduction in computational cost. LIMUNet-Tiny further reduces parameters by 75% and computational load by 80%, making it even more suitable for resource-constrained devices.

Drug Survival, Retention, and Persistence of Dupilumab in Adults and Adolescents with Atopic Dermatitis: A Narrative Literature Review.

Atopic dermatitis (AD) is a chronic, relapsing inflammatory skin condition that can have a negative impact on a patient's quality of life. Long-term effectiveness is required to manage the symptoms of AD (skin inflammation, eczematous lesions, and itching). Because some of the systemic immunosuppressants used to treat AD have been associated with serious adverse events (AEs), other safer, more effective options, including dupilumab, have been proven effective long-term for treatment of adult and adolescent patients with moderate-to-severe AD. The long-term safety and effectiveness of a drug are usually confirmed in real-world studies by evaluating its performance over time. Measures such as drug survival, drug retention, drug persistence, or retention rates reflect whether treatment may be considered as satisfactory by both patients and physicians, meeting key clinical needs. This review aimed to describe the survival, retention, or persistence of dupilumab therapy in adults and adolescents with moderate-to-severe AD by conducting a PubMed search in March 2023 and screening for relevant publications. Globally, real-world studies with dupilumab have regularly reported high drug survival rates after 1, 2, and 3years of observation, being consistently at 80-90%, with low rates of treatment discontinuation due to lack of efficacy or AEs. These findings are notably higher than 1- and 2-year drug survival rates of systemic immunosuppressants (including cyclosporine [37% and 20%, respectively] and methotrexate [41% and 33%, respectively]). Overall, real-world data on drug survival have confirmed that dupilumab provides long-term sustained efficacy and acceptable safety in patients with moderate-to-severe AD.

A New Tool Supporting the Selection of the Best Hematopoietic Stem Cell Donor by Modelling Local Own Real-World Data

Background: The selection of the best donor for each specific patient is crucial for the success of allogeneic hematopoietic stem cell transplantation (HSCT). However, there is debate on the choice of the best donor when multiple suitable donors exist. Methods: By using own data from two transplant centers, we have developed a calculator able to provide the patients’ 2-year overall survival (OS) associated with each of the potential donor options during the selection process, in order to support the transplant physician during the choice. Data on 737 HSCTs with HLA-identical siblings, and unrelated or related haploidentical donors from January 2010 to July 2022 have been retrospectively obtained. Results: Patients’ age, disease, comorbidity index, and donor type were found to be significant variables able to predict the outcome with robustness (concordance index: 0.677). Estimates are provided within an example in the text showing outcomes with four donor options for a specific patient. Conclusions: We present the prototype of a tool supporting the selection of the best donor, guiding transplant physicians during the delicate process of donor selection before HSCT. This approach relies on real data from the centers, reflecting their local clinical experience. Improvements are underway with a larger, ongoing multicenter study.

The Harmony Index: Evaluating, Predicting, and Visualizing Effectiveness in Multi-Agent Team Dynamics

Team-based games are a keystone pillar of the gaming industry. Sadly, the understanding of team dynamics—and the recommendations for both human and AI-based teammates—are based on a rudimentary understanding of human-AI teaming. We propose a superior metric, which provides information about team effectiveness in an efficient and easily-replicable manner. Without an accurate and effective metric for team evaluation, it is nigh-impossible to provide feedback to players and game designers to improve team balance. We provide such a metric. The Harmony Index, a novel algorithm using real-world data, provides simpler and more accurate actionable directives to improve game design across MOBAs and other game genres. We prove its predictive power in a separate analysis and make recommendations for its use in assessing team effectiveness as well as its future use in additional domains.

Critical Data for Critical Care: A Primer on Leveraging Electronic Health Record Data for Research From Society of Critical Care Medicine’s Panel on Data Sharing and Harmonization

A growing body of critical care research draws on real-world data from electronic health records (EHRs). The bedside clinician has myriad data sources to aid in clinical decision-making, but the lack of data sharing and harmonization standards leaves much of this data out of reach for multi-institution critical care research. The Society of Critical Care Medicine (SCCM) Discovery Data Science Campaign convened a panel of critical care and data science experts to explore and document unique advantages and opportunities for leveraging EHR data in critical care research. This article reviews and illustrates six organizing topics (data domains and common data elements; data harmonization; data quality; data interoperability and digital infrastructure; data access, sharing, and governance; and ethics and equity) as a data science primer for critical care researchers, laying a foundation for future publications from the SCCM Discovery Data Harmonization and Sharing Guiding Principles Panel.

Refraction-Aware Structure from Motion for Airborne Bathymetry

In this work, we introduce the first pipeline that combines a refraction-aware structure from motion (SfM) method with a deep learning model specifically designed for airborne bathymetry. We accurately estimate the 3D positions of the submerged points by integrating refraction geometry within the SfM optimization problem. This way, no refraction correction as post-processing is required. Experiments with simulated data that approach real-world capturing conditions demonstrate that SfM with refraction correction is extremely accurate, with submillimeter errors. We integrate our refraction-aware SfM within a deep learning framework that also takes into account radiometrical information, developing a combined spectral and geometry-based approach, with further improvements in accuracy and robustness to different seafloor types, both textured and textureless. We conducted experiments with real-world data at two locations in the southern Mediterranean Sea, with varying seafloor types, which demonstrate the benefits of refraction correction for the deep learning framework. We made our refraction-aware SfM open source, providing researchers in airborne bathymetry with a practical tool to apply SfM in shallow water areas.

Word-Level Political Sentiments Inferred From Social Media and Application in Recommendation Diversification

Political polarization is commonly observed in democratic countries. While it allows individual citizens to freely choose sides, it also causes the problem of separation and isolation. Especially in information-seeking behaviors, echo chambers and filter bubbles are observed. In this paper, we present a political sentiment dictionary for analyzing political polarization and increasing information heterogeneity. It takes advantage of large-scale social media data and is thus superior in accuracy and coverage compared to manually crafted dictionaries. Generated from Japanese tweets, more than 50k words in this dictionary cover aspects ranging from political parties and public entities to foods and personal hobbies. We describe in detail the method to construct this dictionary, which can be replicated for other languages and countries. We demonstrate the use of this dictionary in the application of recommendation diversification. We show with real-world e-commerce data that the use of the dictionary can generally increase the diversity in product recommendations, effectively mitigating the filter bubbles.

Filtered partial differential equations: a robust surrogate constraint in physics-informed deep learning framework

Embedding physical knowledge into neural network (NN) training has been a hot topic. However, when facing the complex real world, most of the existing methods still strongly rely on the quantity and quality of observation data. Furthermore, the NNs often struggle to converge when the solution to the real equation is very complex. Inspired by large eddy simulation in computational fluid dynamics, we propose an improved method based on filtering. We analysed the causes of the difficulties in physics-informed machine learning, and proposed a surrogate constraint (filtered partial differential equation, FPDE) of the original physical equations to reduce the influence of noisy and sparse observation data. In the noise and sparsity experiment, the proposed FPDE models (which are optimized by FPDE constraints) have better robustness than the conventional PDE models. Experiments demonstrate that the FPDE model can obtain the same quality solution with 100 % higher noise and 12 % quantity of observation data of the baseline. Besides, two groups of real measurement data are used to show the FPDE improvements in real cases. The final results show that the FPDE still gives more physically reasonable solutions when facing the incomplete equation problem and the extremely sparse and high-noise conditions. The proposed FPDE constraint is helpful for merging real-world experimental data into physics-informed training, and it works effectively in two real-world experiments: simulating cell movement in scratches and blood velocity in vessels.

Immunotherapy benefits for large brain metastases in non-small cell lung cancer.

Non-small cell lung cancer (NSCLC) patients with large brain metastases (BrM) defined as >2cm in diameter historically face grim prognoses. With immunotherapy emerging as a promising avenue for BrM management and being commonly used in NSCLC, its application in addressing large BrM remains underexplored. This retrospective study conducted across the MedStar Georgetown Cancer Network aimed to assess the efficacy of immunotherapy in non-biomarker driven NSCLC patients with large BrM following initial treatment. Thirty-six patients were included, all of whom underwent neurosurgery and/or radiation before commencing immunotherapy. The median intracranial progression-free survival (PFS) was 9.2 months and the median overall survival (OS) reached 31 months. Utilizing multivariable Cox penalized regression, the intracranial PFS hazard ratio (HR) was 0.07 (95% confidence interval (CI), 0.02-0.26) for patients who received at least 90 days of immunotherapy compared to those who did not. Each additional 30 days of immunotherapy was associated with an OS HR 0.77 (95% CI, 0.67-0.90). This real-world data highlights the potential of immunotherapy in large BrM NSCLC patients, a population often excluded from clinical trials. This study contributes insights that can inform future treatment approaches, emphasizing the need for further exploration of immunotherapy's role in enhancing outcomes for this challenging patient population.