scholarly journals Multiobjective Calibration of Disease Simulation Models Using Gaussian Processes

2019 ◽  
Vol 39 (5) ◽  
pp. 540-552
Author(s):  
Aditya Sai ◽  
Carolina Vivas-Valencia ◽  
Thomas F. Imperiale ◽  
Nan Kong
Keyword(s):  
Natural History ◽  
Goodness Of Fit ◽  
Transition Probabilities ◽  
Pareto Frontier ◽  
Treatment Strategies ◽  
Simulation Models ◽  
Optimal Parameters ◽  
Dimensional Parameter ◽  
Best Fitting ◽  
Multiobjective Calibration

Background. Developing efficient procedures of model calibration, which entails matching model predictions to observed outcomes, has gained increasing attention. With faithful but complex simulation models established for cancer diseases, key parameters of cancer natural history can be investigated for possible fits, which can subsequently inform optimal prevention and treatment strategies. When multiple calibration targets exist, one approach to identifying optimal parameters relies on the Pareto frontier. However, computational burdens associated with higher-dimensional parameter spaces require a metamodeling approach. The goal of this work is to explore multiobjective calibration using Gaussian process regression (GPR) with an eye toward how multiple goodness-of-fit (GOF) criteria identify Pareto-optimal parameters. Methods. We applied GPR, a metamodeling technique, to estimate colorectal cancer (CRC)–related prevalence rates simulated from a microsimulation model of CRC natural history, known as the Colon Modeling Open Source Tool (CMOST). We embedded GPR metamodels within a Pareto optimization framework to identify best-fitting parameters for age-, adenoma-, and adenoma staging–dependent transition probabilities and risk factors. The Pareto frontier approach is demonstrated using genetic algorithms with both sum-of-squared errors (SSEs) and Poisson deviance GOF criteria. Results. The GPR metamodel is able to approximate CMOST outputs accurately on 2 separate parameter sets. Both GOF criteria are able to identify different best-fitting parameter sets on the Pareto frontier. The SSE criterion emphasizes the importance of age-specific adenoma progression parameters, while the Poisson criterion prioritizes adenoma-specific progression parameters. Conclusion. Different GOF criteria assert different components of the CRC natural history. The combination of multiobjective optimization and nonparametric regression, along with diverse GOF criteria, can advance the calibration process by identifying optimal regions of the underlying parameter landscape.

scholarly journals Comparison of growth models to describe growth from birth to 6 years in a Beninese cohort of children with repeated measurements

BMJ Open ◽  
2020 ◽  
Vol 10 (9) ◽  
pp. e035785
Author(s):  
Shukrullah Ahmadi ◽  
Florence Bodeau-Livinec ◽  
Roméo Zoumenou ◽  
André Garcia ◽  
David Courtin ◽  
...  
Keyword(s):  
Growth Model ◽  
Goodness Of Fit ◽  
Growth Models ◽  
Information Criterion ◽  
Height Growth ◽  
Sub Saharan Africa ◽  
Individual Growth ◽  
Secondary Outcome ◽  
Growth Trajectories ◽  
Best Fitting

ObjectiveTo select a growth model that best describes individual growth trajectories of children and to present some growth characteristics of this population.SettingsParticipants were selected from a prospective cohort conducted in three health centres (Allada, Sekou and Attogon) in a semirural region of Benin, sub-Saharan Africa.ParticipantsChildren aged 0 to 6 years were recruited in a cohort study with at least two valid height and weight measurements included (n=961).Primary and secondary outcome measuresThis study compared the goodness-of-fit of three structural growth models (Jenss-Bayley, Reed and a newly adapted version of the Gompertz growth model) on longitudinal weight and height growth data of boys and girls. The goodness-of-fit of the models was assessed using residual distribution over age and compared with the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC). The best-fitting model allowed estimating mean weight and height growth trajectories, individual growth and growth velocities. Underweight, stunting and wasting were also estimated at age 6 years.ResultsThe three models were able to fit well both weight and height data. The Jenss-Bayley model presented the best fit for weight and height, both in boys and girls. Mean height growth trajectories were identical in shape and direction for boys and girls while the mean weight growth curve of girls fell slightly below the curve of boys after neonatal life. Finally, 35%, 27.7% and 8% of boys; and 34%, 38.4% and 4% of girls were estimated to be underweight, wasted and stunted at age 6 years, respectively.ConclusionThe growth parameters of the best-fitting Jenss-Bayley model can be used to describe growth trajectories and study their determinants.

scholarly journals Machine learning approaches to calibrate individual-based infectious disease models

2021 ◽  
Author(s):  
Theresa Reiker ◽  
Monica Golumbeanu ◽  
Andrew Shattock ◽  
Lydia Burgert ◽  
Thomas A. Smith ◽  
...  
Keyword(s):  
Machine Learning ◽  
Disease Transmission ◽  
Goodness Of Fit ◽  
Epidemiological Data ◽  
Bayesian Optimization ◽  
Model Complexity ◽  
Learning Approaches ◽  
Dimensional Parameter ◽  
Novel Approach ◽  
Transmission Models

AbstractIndividual-based models have become important tools in the global battle against infectious diseases, yet model complexity can make calibration to biological and epidemiological data challenging. We propose a novel approach to calibrate disease transmission models via a Bayesian optimization framework employing machine learning emulator functions to guide a global search over a multi-objective landscape. We demonstrate our approach by application to an established individual-based model of malaria, optimizing over a high-dimensional parameter space with respect to a portfolio of multiple fitting objectives built from datasets capturing the natural history of malaria transmission and disease progression. Outperforming other calibration methodologies, the new approach quickly reaches an improved final goodness of fit. Per-objective parameter importance and sensitivity diagnostics provided by our approach offer epidemiological insights and enhance trust in predictions through greater interpretability.One Sentence SummaryWe propose a novel, fast, machine learning-based approach to calibrate disease transmission models that outperforms other methodologies

Brain arteriovenous malformations

Neurology ◽  
2020 ◽  
Vol 95 (20) ◽  
pp. 917-927 ◽  
Author(s):  
Ching-Jen Chen ◽  
Dale Ding ◽  
Colin P. Derdeyn ◽  
Giuseppe Lanzino ◽  
Robert M. Friedlander ◽  
...  
Keyword(s):  
Natural History ◽  
Minimally Invasive ◽  
Arteriovenous Malformations ◽  
Cerebral Arteries ◽  
Treatment Strategies ◽  
Latency Period ◽  
Management Options ◽  
Brain Arteriovenous Malformations ◽  
Risk Of Hemorrhage ◽  
Brain Avms

Brain arteriovenous malformations (AVMs) are anomalous direct shunts between cerebral arteries and veins that convalesce into a vascular nidus. The treatment strategies for AVMs are challenging and variable. Intracranial hemorrhage and seizures comprise the most common presentations of AVMs. However, incidental AVMs are being diagnosed with increasing frequency due to widespread use of noninvasive neuroimaging. The balance between the estimated cumulative lifetime hemorrhage risk vs the risk of intervention is often the major determinant for treatment. Current management options include surgical resection, embolization, stereotactic radiosurgery (SRS), and observation. Complete nidal obliteration is the goal of AVM intervention. The risks and benefits of interventions vary and can be used in a combinatorial fashion. Resection of the AVM nidus affords high rates of immediate obliteration, but it is invasive and carries a moderate risk of neurologic morbidity. AVM embolization is minimally invasive, but cure can only be achieved in a minority of lesions. SRS is also minimally invasive and has little immediate morbidity, but AVM obliteration occurs in a delayed fashion, so the patient remains at risk of hemorrhage during the latency period. Whether obliteration can be achieved in unruptured AVMs with a lower risk of stroke or death compared with the natural history of AVMs remains controversial. Over the past 5 years, multicenter prospective and retrospective studies describing AVM natural history and treatment outcomes have been published. This review provides a contemporary and comprehensive discussion of the natural history, pathobiology, and interventions for brain AVMs.

scholarly journals “Malignant” Perivascular Epithelioid Cell Neoplasm: Risk Stratification and Treatment Strategies

Sarcoma ◽  
2012 ◽  
Vol 2012 ◽  
pp. 1-12 ◽  
Author(s):  
Jonathan S. Bleeker ◽  
J. Fernando Quevedo ◽  
Andrew L. Folpe
Keyword(s):  
Natural History ◽  
Risk Stratification ◽  
English Literature ◽  
Treatment Strategies ◽  
Mitotic Rate ◽  
Epithelioid Cell ◽  
Diagnostic Features ◽  
Mtor Inhibition ◽  
Perivascular Epithelioid Cell ◽  
Not Otherwise Specified

Purpose. Perivascular epithelioid cell tumors (PEComas) are a rare collection of tumors characterized by a myomelanocytic phenotype, and PEComas occurring in “nonclassic” anatomic distributions are known as perivascular epithelioid cell tumor not otherwise specified (PEComa-NOS). This review aims to compile and analyze cases of PEComa-NOS in an effort to better define their natural history.Design. We evaluated all 234 cases of PEComa-NOS reported in the English literature, extracting information regarding diagnostic features, treatment approaches, and outcomes. Multivariate analysis of a number of variables evaluable on pathologic review was performed to refine preexisting risk stratification criteria. Outcomes for patients receiving nonsurgical treatment are also reported.Results. Primary tumor size ≥5 cm (P=0.02) and a high (1/50 HPF) mitotic rate (P<0.0001) were the only factors significantly associated with recurrence following surgical resection. Cytotoxic chemotherapy and radiation therapy have shown little benefit in treating PEComa-NOS; mTOR inhibition is emerging as a treatment option.Conclusion. Progress has been made in understanding the natural history and molecular biology of PEComa-NOS. This review further clarifies risk of recurrence in this disease, allowing clinicians to better risk stratify patients. Further work should focus on applying this knowledge to making treatment decisions for patients with this disease.

scholarly journals Emulator-based Bayesian optimization for efficient multi-objective calibration of an individual-based model of malaria

2021 ◽  
Vol 12 (1) ◽  
Author(s):  
Theresa Reiker ◽  
Monica Golumbeanu ◽  
Andrew Shattock ◽  
Lydia Burgert ◽  
Thomas A. Smith ◽  
...  
Keyword(s):  
Malaria Transmission ◽  
Goodness Of Fit ◽  
Epidemiological Data ◽  
Bayesian Optimization ◽  
Model Complexity ◽  
Dimensional Parameter ◽  
Optimization Framework ◽  
Individual Based Models ◽  
Objective Parameter ◽  
History Of

AbstractIndividual-based models have become important tools in the global battle against infectious diseases, yet model complexity can make calibration to biological and epidemiological data challenging. We propose using a Bayesian optimization framework employing Gaussian process or machine learning emulator functions to calibrate a complex malaria transmission simulator. We demonstrate our approach by optimizing over a high-dimensional parameter space with respect to a portfolio of multiple fitting objectives built from datasets capturing the natural history of malaria transmission and disease progression. Our approach quickly outperforms previous calibrations, yielding an improved final goodness of fit. Per-objective parameter importance and sensitivity diagnostics provided by our approach offer epidemiological insights and enhance trust in predictions through greater interpretability.

scholarly journals Discrete k-nearest neighbor resampling for simulating multisite precipitation occurrence and adaption to climate change

2018 ◽  
Author(s):  
Taesam Lee ◽  
Vijay P. Singh
Keyword(s):  
Climate Change ◽  
Nearest Neighbor ◽  
Transition Probabilities ◽  
Precipitation Amount ◽  
Simulation Models ◽  
Simulation Method ◽  
Good Alternative ◽  
Discrete Version ◽  
K Nearest Neighbor ◽  
Precipitation Occurrence

Abstract. Stochastic weather simulation models are commonly employed in water resources management and agricultural applications. The data simulated by these models, such as precipitation, temperature, and wind, are used as input for hydrological and agricultural models. Stochastic simulation of multisite precipitation occurrence is a challenge because of its intermittent characteristics as well as spatial and temporal cross-correlation. Employing a nonparametric technique, k-nearest neighbor resampling (KNNR), and coupling it with Genetic Algorithm (GA), this study proposes a novel simulation method for multisite precipitation occurrence. The proposed discrete version of KNNR (DKNNR) model is compared with an existing parametric model, called multisite occurrence model with standard normal variate (MONR). The datasets simulated from both the DKNNR model and the MONR model are tested using a number of statistics, such as occurrence and transition probabilities as well as temporal and spatial cross-correlations. Results show that the proposed DKNNR model can be a good alternative for simulating multisite precipitation occurrence. We also tested the model capability to adapt climate change. It is shown that the model is capable but further improvement is required to have specific variations of the occurrence probability due to climate change. Combining with the generated occurrence, the multisite precipitation amount can then be simulated by any multisite amount model.

Multibrand Transition Probabilities as a Function of Explanatory Variables: Estimation by a Least-Squares-Based Approach

1986 ◽  
Vol 23 (2) ◽  
pp. 177-183 ◽  
Author(s):  
Fred S. Zufryden
Keyword(s):  
Least Squares ◽  
Goodness Of Fit ◽  
Transition Probabilities ◽  
Weighted Least Squares ◽  
Estimation Procedure ◽  
Explanatory Variables ◽  
First Order ◽  
Proposed Model ◽  
Markov Transition ◽  
The Relationship

A model is formulated to express the relationship between first-order Markov transition probabilities for a multibrand market and explanatory variables. The author shows that the parameters of the model can be estimated through a proposed restricted weighted least squares procedure. An empirical implementation of the estimation procedure illustrates the structure, goodness of fit, and predictive validity of the proposed model.

Incorporating Biological Natural History in Simulation Models: Empirical Estimates of the Progression of End-Stage Liver Disease

2005 ◽  
Vol 25 (6) ◽  
pp. 620-632 ◽  
Author(s):  
Oguzhan Alagoz ◽  
Cindy L. Bryce ◽  
Steven Shechter ◽  
Andrew Schaefer ◽  
Chung-Chou H. Chang ◽  
...  
Keyword(s):  
Liver Disease ◽  
Natural History ◽  
Simulation Models ◽  
End Stage Liver Disease ◽  
Empirical Estimates ◽  
End Stage

Cost-effectiveness of nodal observation versus completion lymphadenectomy in patients with melanoma and sentinel lymph node metastases.

2020 ◽  
Vol 38 (15_suppl) ◽  
pp. e22076-e22076
Author(s):  
Elena Parvez ◽  
Teodora Dumitra ◽  
Dimitra Panagiotoglou ◽  
Sarkis H. Meterissian ◽  
Sinziana Dumitra
Keyword(s):  
Cost Effectiveness ◽  
Transition Probabilities ◽  
Treatment Strategies ◽  
Cost Savings ◽  
Cost Effective ◽  
Health State ◽  
Complication Rates ◽  
Disease States ◽  
Payer Perspective ◽  
Life Years

e22076 Background: The MSLT-II trial demonstrated no survival benefit of completion lymphadenectomy (CLND) compared to nodal observation (NO) and subsequent therapeutic lymphadenectomy (TLND) in the case of macroscopic nodal relapse in patients with melanoma and SLN metastases. NO avoids the upfront cost and morbidity of CLND. However, patients followed with NO must undergo intensive surveillance and if TLND is required, it is associated with a higher complication rate than CLND. The cost-effectiveness of NO versus CLND in light of data from MSLT-II has not been previously studied. Methods: A Markov model with a 10-year time horizon was constructed to simulate two hypothetical cohorts of patients with SLN metastases undergoing NO and subsequent TLND for nodal recurrence or upfront CLND. Transition probabilities between disease states were derived from the MSLT-II trial. Remaining parameters including complication rates and health state utilities were obtained from an extensive review of the literature. Direct health care system costs were obtained from published US Medicare cost data and the literature. Primary outcomes were cost and quality-adjusted life years (QALYs) saved. Incremental cost-effectiveness ratio (ICER) was used to compare treatment strategies. Sensitivity analysis was performed in order to evaluate model uncertainty. A threshold of acceptance of $100,000/QALY was used. Results: Total projected cost over the study period for CLND was $28,609.87, while that of NO was lower at $20,865.27, resulting in $7,744.60 saved for the NO treatment strategy. Ten-year utility was 4.840 for CLND compared to 5.379 for NO, resulting in a gain of 0.539 QALYs in the NO arm. The NO strategy is dominant in the model as it results in both cost-savings and a gain in health effects, with an average ICER of -$14,368.46/QALY gained. Conclusions: From the payer perspective, the strategy of NO compared to CLND in patients with melanoma and SLN metastases is associated with an improvement in health outcomes and reduction in cost. Taking into account MSLT-II trial data, this study demonstrates NO is more cost-effective than CLND.

Calibrating Wiedemann-99 Model Parameters to Trajectory Data of Mixed Vehicular Traffic

2021 ◽  
pp. 036119812110375
Author(s):  
Ankit Anil Chaudhari ◽  
Karthik K. Srinivasan ◽  
Bhargava Rama Chilukuri ◽  
Martin Treiber ◽  
Ostap Okhrin
Keyword(s):  
Goodness Of Fit ◽  
Simulation Models ◽  
Microscopic Level ◽  
Model Parameters ◽  
Mixed Traffic ◽  
Relative Speed ◽  
Trajectory Data ◽  
Conventional Vehicle ◽  
Car Following ◽  
Emergency Braking

We propose a new methodology for calibrating Wiedemann-99 vehicle-following parameters for mixed traffic (different conventional vehicle classes) based on trajectory data. The existing acceleration equations of the Wiedemann model are modified to represent more realistic driving behavior. Exploratory analysis of simulation data revealed that different Wiedemann-99 model parameters could lead to similar macroscopic behavior, highlighting the importance of calibration at the microscopic level. Therefore, the proposed methodology is based on optimizing performance measures at the microscopic level (acceleration, speed, and trajectory profiles) to estimate suitable calibration parameters. Further, the goodness of fit for the observed data is sensitive to the numerical integration method used to compute vehicles’ velocity and position. We found that the calibrated parameters using the proposed methodology perform better than other approaches for calibrating mixed traffic. The results reveal that the calibrated parameter values and, consequently, the thresholds that delineate closing, following, emergency braking, and opening regimes, vary between two-wheelers and cars. The window (in the relative speed versus gap plot) for the unconscious following is larger for cars while the free-flow regime is more extensive for two-wheelers. Moreover, under the same relative speed and gap stimulus, two-wheelers and cars may be in different regimes and display different acceleration responses. Thus, accurate calibration of each vehicle’s parameters is essential for developing micro-simulation models for mixed traffic. The calibration analysis results of strict and overlapping staggered car following signify an impact of staggered car following compared with strict car following which demands separate calibration for strict and staggered following.

Sign in / Sign up

Export Citation Format

Share Document