Search NASA⌕ Search

SEARCH · Search NASA

Results for “Random forests”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 181 records · Page 10

GraphAlign: Graph-Enabled Machine Learning for Seismic Event Filtering

This report summarizes results from a 2 year effort to improve the current automated seismic event processing system by leveraging machine learning models that can operated over the inherent graph data structure of a seismic sensor network. Specifically, the GraphAlign project seeks to utilize prior information on which stations are more likely to detect signals originating from particular geographic regions to inform event filtering. To date, the GraphAlign team has developed a Graphical Neural Network (GNN) model to filter out false events generated by the Global Associator (GA) algorithm. The algorithm operates directly on waveform data that has been associated to an event by building a variable sized graph of station waveforms nodes with edge relations to an event location node. This builds off of previous work where random forest models were used to do the same task using hand crafted features. The GNN model performance was analyzed using an 8 week IMS/IDC dataset, and it was demonstrated that the GNN outperforms the random forest baseline. We provide additional error analysis of which events the GNN model performs well and poorly against concluded by future directions for improvements.

58 GEOSCIENCES↗

Power System Feature-Based Event Classification by Means of Multiple PMU Data

Abstract—Phasor Measurement Units (PMUs) provide time synchronized measurements across the power grid, enabling data driven event detection and classification for enhanced system monitoring and situational awareness. However, variations in event duration, spatial extent, and severity, along with coincident events, pose challenges for conventional classification models that require fixed-size inputs. This paper presents a feature-based framework that aggregates diverse attributes from all available PMUs for each event into a fixed-length vector, facilitating the application of standard machine learning classifiers, including Random Forest, XGBoost, and Multilayer Perceptron. A probabilistic post-processing scheme is further introduced to enable multi-label classification in the presence of overlapping events. Experiments using real-world PMU data demonstrate that the Random Forest model achieves 95% accuracy, while the proposed post-processing method yields an additional 3% improvement.

Nematirad, Reza↗

Spatial patterns of snow distribution in the sub-Arctic

Abstract. The spatial distribution of snow plays a vital role in sub-Arctic and Arctic climate, hydrology, and ecology due to its fundamental influence on the water balance, thermal regimes, vegetation, and carbon flux. However, the spatial distribution of snow is not well understood, and therefore, it is not well modeled, which can lead to substantial uncertainties in snow cover representations. To capture key hydro-ecological controls on snow spatial distribution, we carried out intensive field studies over multiple years for two small (2017–2019; ∼ 2.5 km2) sub-Arctic study sites located on the Seward Peninsula of Alaska. Using an intensive suite of field observations (> 22 000 data points), we developed simple models of the spatial distribution of snow water equivalent (SWE) using factors such as topographic characteristics, vegetation characteristics based on greenness (normalized different vegetation index, NDVI), and a simple metric for approximating winds. The most successful model was random forest, using both study sites and all years, which was able to accurately capture the complexity and variability of snow characteristics across the sites. Approximately 86 % of the SWE distribution could be accounted for, on average, by the random forest model at the study sites. Factors that impacted year-to-year snow distribution included NDVI, elevation, and a metric to represent coarse microtopography (topographic position index, TPI), while slope, wind, and fine microtopography factors were less important. The characterization of the SWE spatial distribution patterns will be used to validate and improve snow distribution modeling in the Department of Energy's Earth system model and for improved understanding of hydrology, topography, and vegetation dynamics in the sub-Arctic and Arctic regions of the globe.

54 ENVIRONMENTAL SCIENCES↗

Upscaling Soil Organic Carbon Measurements at the Continental Scale Using Multivariate Clustering Analysis and Machine Learning

Abstract Estimates of soil organic carbon (SOC) stocks are essential for many environmental applications. However, significant inconsistencies exist in SOC stock estimates for the U.S. across current SOC maps. We propose a framework that combines unsupervised multivariate geographic clustering (MGC) and supervised Random Forests regression, improving SOC maps by capturing heterogeneous relationships with SOC drivers. We first used MGC to divide the U.S. into 20 SOC regions based on the similarity of covariates (soil biogeochemical, bioclimatic, biological, and physiographic variables). Subsequently, separate Random Forests models were trained for each SOC region, utilizing environmental covariates and SOC observations. Our estimated SOC stocks for the U.S. (52.6 ± 3.2 Pg for 0–30 cm and 108.3 ± 8.2 Pg for 0–100 cm depth) were within the range estimated by existing products like Harmonized World Soil Database, HWSD (46.7 Pg for 0–30 cm and 90.7 Pg for 0–100 cm depth) and SoilGrids 2.0 (45.7 Pg for 0–30 cm and 133.0 Pg for 0–100 cm depth). However, independent validation with soil profile data from the National Ecological Observatory Network showed that our approach ( R 2 = 0.51) outperformed the estimates obtained from Harmonized World Soil Database ( R 2 = 0.23) and SoilGrids 2.0 ( R 2 = 0.39) for the topsoil (0–30 cm). Uncertainty analysis (e.g., low representativeness and high coefficients of variation) identified regions requiring more measurements, such as Alaska and the deserts of the U.S. Southwest. Our approach effectively captures the heterogeneous relationships between widely available predictors and the current SOC baseline across regions, offering reliable SOC estimates at 1 km resolution for benchmarking Earth system models.

58 GEOSCIENCES↗

PHACTS, a computational approach to classifying the lifestyle of phages

Motivation: Bacteriophages have two distinct lifestyles: virulent and temperate. The virulent lifestyle has many implications for phage therapy, genomics and microbiology. Determining which lifestyle a newly sequenced phage falls into is currently determined using standard culturing techniques. Such laboratory work is not only costly and time consuming, but also cannot be used on phage genomes constructed from environmental sequencing. Therefore, a computational method that utilizes the sequence data of phage genomes is needed. Results: Phage Classification Tool Set (PHACTS) utilizes a novel similarity algorithm and a supervised Random Forest classifier to make a prediction whether the lifestyle of a phage, described by its proteome, is virulent or temperate. The similarity algorithm creates a training set from phages with known lifestyles and along with the lifestyle annotation, trains a Random Forest to classify the lifestyle of a phage. PHACTS predictions are shown to have a 99% precision rate.

59 BASIC BIOLOGICAL SCIENCES↗

Long-Term Variability of Soil Moisture in the Southern Sierra: Measurement and Prediction

Using 6 yr (Water Year [WY] 2009–WY 2014) of hourly in situ measurements from a spatially distributed water-balance cluster, we quantified the long-term accuracy of an algorithm used to predict spatial patterns of depth-integrated soil-water storage within the rain–snow transition zone of the southern Sierra Nevada. The algorithm—the multivariate, non-parametric regression-tree estimator Random Forest—was used to predict soil-water storage based on a combination of attributes at each instrument cluster (soil texture, topographic wetness index, elevation, northness, and canopy cover). Out-of-bag R 2 (similar to cross-validation for Random Forest) was used to quantify the accuracy of the estimator for unobserved data. Accuracy was consistently high during the wet-up, snow-cover, and early recession periods of average and wet years. The accuracy declined at the end of a 3-yr dry period, and the relative rank of the independent variables in the model shifted. Soil texture was the highest-ranked independent variable across all years, followed by elevation and northness. Topographic wetness increased in importance during dry periods. Northness exhibited high importance during the wet-up and early recession periods of most water years. As a result, during dry years, the importance of elevation declined. In dry years, notable differences in soil-water storage at each depth include lower-than-average storage in the deeper regolith at the beginning of the water year and lower storage in near-surface layers during the winter resulting from transient snow cover.

54 ENVIRONMENTAL SCIENCES↗

Geographical Insights into Suicide Mortality Through Spatial Machine Learning

Suicide mortality is a leading cause of death in the United States, with an upward trend that emphasizes its significance as a public health issue. Previous research has employed global models like ordinary least squares (OLS) regression and local models such as geographically weighted regression (GWR). While local models are useful for analyzing spatial variations in suicide mortality, they share limitations with traditional global models, particularly about their inability to handle multi-collinearity and non-linear relationships. Machine learning approaches, like random forests (RF), can address some of these limitations but often fail to account for spatial variability. This gap highlights the need for spatial ML models specifically designed to tackle suicide mortality. This research seeks to fill this void by using a geographically weighted random forest model (GWRF) to examine the associations between county-level suicide mortality in the U.S. from 2010 to 2020 and various social and environmental determinants of health. A key aspect of our methodology is disciplined feature selection, which reduces the pool of explanatory variables by about 90%. This refinement enhances the explanatory power of both global (R2 improved from 0.59 to 0.67) and local (R2 improved from 0.64 to 0.67) RF models while reducing their run times. An analysis of the importance scores for these selected features reveals that the drivers of suicide mortality vary by context. Thus, to effectively address regional disparities and inform targeted public health interventions, a holistic approach that incorporates multiple county-level characteristics is essential.

Lebakula, Viswadeep [ORNL] (ORCID:0000000152935914↗

Evaluating county-level lung cancer incidence from environmental radiation exposure, PM 2.5 , and other exposures with regression and machine learning models

Characterizing the interplay between exposures shaping the human exposome is vital for uncovering the etiology of complex diseases. For example, cancer risk is modified by a range of multifactorial external environmental exposures. Environmental, socioeconomic, and lifestyle factors all shape lung cancer risk. However, epidemiological studies of radon aimed at identifying populations at high risk for lung cancer often fail to consider multiple exposures simultaneously. For example, moderating factors, such as PM 2.5 , may affect the transport of radon progeny to lung tissue. This ecological analysis leveraged a population-level dataset from the National Cancer Institute’s Surveillance, Epidemiology, and End-Results data (2013–17) to simultaneously investigate the effect of multiple sources of low-dose radiation (gross γ activity and indoor radon) and PM 2.5 on lung cancer incidence rates in the USA. County-level factors (environmental, sociodemographic, lifestyle) were controlled for, and Poisson regression and random forest models were used to assess the association between radon exposure and lung and bronchus cancer incidence rates. Tree-based machine learning (ML) method perform better than traditional regression: Poisson regression: 6.29/7.13 (mean absolute percentage error, MAPE), 12.70/12.77 (root mean square error, RMSE); Poisson random forest regression: 1.22/1.16 (MAPE), 8.01/8.15 (RMSE). The effect of PM 2.5 increased with the concentration of environmental radon, thereby confirming findings from previous studies that investigated the possible synergistic effect of radon and PM 2.5 on health outcomes. In summary, the results demonstrated (1) a need to consider multiple environmental exposures when assessing radon exposure’s association with lung cancer risk, thereby highlighting (1) the importance of an exposomics framework and (2) that employing ML models may capture the complex interplay between environmental exposures and health, as in the case of indoor radon exposure and lung cancer incidence.

63 RADIATION, THERMAL, AND OTHER ENVIRON. POLLUTAN↗

The Use of Machine Learning Models for Predicting the Dielectric Strength of Gases

Technological advancements in high voltage systems have pushed sulfur hexafluoride (SF6) to its operational limits. Furthermore, this gas has other drawbacks including a high liquefaction temperature and a high global warming potential. Therefore, there has been an urgent need to find alternative gases with high dielectric strength (DS). In this work, density functional theory (DFT) is used to calculate molecular descriptors that are fed into an artificial neural network (ANN) and a random forest (RF). These machine learning (ML) models are then used to predict the DS for hundreds of molecules. A finite element model (FEM) is also used to calculate the electric field profile of multiple simple electrode geometries as the applied voltage to the system is increased. Results indicate that the random forest model has better generalization to unseen data than the neural network. The highest DS value predicted by the RF was 2.16 relative to the experimental DS of SF6. The results also demonstrate how choosing a gas with a higher DS and a geometry with minimal edges and corners can significantly increase the operating voltage of an electrical system. Due to its superior generalization, the RF represents the most promising path toward an accurate DS predictor once sufficient experimental data are available.

Mileski, Matthew [AFIT]↗

Comparison of ANN (MLP), ANFIS, SVM, and RF models for the online classification of heating value of burning municipal solid waste in circulating fluidized bed incinerators

Highlights: • A rapid, cost-effective, and comparative methodology was proposed to evaluate the HVs of burning MSW online. • Models ranked in descending order of performance are ANFIS, RF, SVM, MLP. • A well trained model is capable of characterizing the variation trend of HVs accurately. - Abstract: The heating values, particularly lower heating values of burning municipal solid waste are critically important parameters in operating circulating fluidized bed incineration systems. However, the heating values change widely and frequently, while there is no reliable real-time instrument to measure heating values in the process of incinerating municipal solid waste. A rapid, cost-effective, and comparative methodology was proposed to evaluate the heating values of burning MSW online based on prior knowledge, expert experience, and data-mining techniques. First, selecting the input variables of the model by analyzing the operational mechanism of circulating fluidized bed incinerators, and the corresponding heating value was classified into one of nine fuzzy expressions according to expert advice. Development of prediction models by employing four different nonlinear models was undertaken, including a multilayer perceptron neural network, a support vector machine, an adaptive neuro-fuzzy inference system, and a random forest; a series of optimization schemes were implemented simultaneously in order to improve the performance of each model. Finally, a comprehensive comparison study was carried out to evaluate the performance of the models. Results indicate that the adaptive neuro-fuzzy inference system model outperforms the other three models, with the random forest model performing second-best, and the multilayer perceptron model performing at the worst level. A model with sufficient accuracy would contribute adequately to the control of circulating fluidized bed incinerator operation and provide reliable heating value signals for an automatic combustion control system.

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

The importance of round-robin validation when assessing machine-learning-based vertical extrapolation of wind speeds

The extrapolation of wind speeds measured at a meteorological mast to wind turbine rotor heights is a key component in a bankable wind farm energy assessment and a significant source of uncertainty. Industry-standard methods for extrapolation include the power-law and logarithmic profiles. The emergence of machine-learning applications in wind energy has led to several studies demonstrating substantial improvements in vertical extrapolation accuracy in machine-learning methods over these conventional power-law and logarithmic profile methods. In all cases, these studies assess relative model performance at a measurement site where, critically, the machine-learning algorithm requires knowledge of the rotor-height wind speeds in order to train the model. This prior knowledge provides fundamental advantages to the site-specific machine-learning model over the power-law and log profiles, which, by contrast, are not highly tuned to rotor-height measurements but rather can generalize to any site. Furthermore, there is no practical benefit in applying a machine-learning model at a site where winds at the heights relevant for wind energy production are known; rather, its performance at nearby locations (i.e., across a wind farm site) without rotor-height measurements is of most practical interest. To more fairly and practically compare machine-learning-based extrapolation to standard approaches, we implemented a round-robin extrapolation model comparison, in which a random-forest machine-learning model is trained and evaluated at different sites and then compared against the power-law and logarithmic profiles. We consider 20 months of lidar and sonic anemometer data collected at four sites between 50 and 100 km apart in the central United States. We find that the random forest outperforms the standard extrapolation approaches, especially when incorporating surface measurements as inputs to include the influence of atmospheric stability. When compared at a single site (the traditional comparison approach), the machine-learning improvement in mean absolute error was 28 % and 23 % over the power-law and logarithmic profiles, respectively. Using the round-robin approach proposed here, this improvement drops to 20 % and 14 %, respectively. These latter values better represent practical model performance, and we conclude that round-robin validation should be the standard for machine-learning-based wind speed extrapolation methods.

17 WIND ENERGY↗

Application of machine learning to discover new intermetallic catalysts for the hydrogen evolution and the oxygen reduction reactions

The adsorption energies for hydrogen, oxygen, and hydroxyl were calculated by means of density functional theory on the lowest energy surface of 24 pure metals and 332 binary intermetallic compounds with stoichiometries AB, A 2 B, and A 3 B taking into account the effect of biaxial elastic strains. This information was used to train two random forest regression models, one for the hydrogen adsorption and another for the oxygen and hydroxyl adsorption, based on 9 descriptors that characterized the geometrical and chemical features of the adsorption site as well as the applied strain. All the descriptors for each compound in the models could be obtained from physico-chemical databases. The random forest models were used to predict the adsorption energy for hydrogen, oxygen, and hydroxyl of ≈2700 binary intermetallic compounds with stoichiometries AB, A 2 B, and A 3 B made of metallic elements, excluding those that were environmentally hazardous, radioactive, or toxic. This information was used to search for potential good catalysts for the HER and ORR from the criteria that their adsorption energy for H and O/OH, respectively, should be close to that of Pt. Further, this investigation shows that the suitably trained machine learning models can predict adsorption energies with an accuracy not far away from density functional theory calculations with minimum computational cost from descriptors that are readily available in physico-chemical databases for any compound. Moreover, the strategy presented in this paper can be easily extended to other compounds and catalytic reactions, and is expected to foster the use of ML methods in catalysis.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Forest Soil Carbon Efflux Evaluation Across China: A New Estimate With Machine Learning

Forest soil respiration (Rs) plays an important role in the carbon balance of terrestrial ecosystems. China's forest occupies a large part of the world's forest. However, due to the lack of integrated observation data and appropriate upscaling methodologies, substantial uncertainties exist in the Rs evaluation, which limits our understanding of the carbon balance. Here, we re-evaluated the total soil carbon effluxes across China by combining field observations from 634 published annual Rs with a machine learning technique (i.e., Random Forest (RF)). Our results revealed that the combination of systematic measurements with the RF model allowed a definite estimate. The average annual Rs was 773.6 g C m -2 yr -1 , ranging from 404.6 to 1,772.3 g C m -2 yr -1 . Total forest soil carbon effluxes amounted to 1.17 Pg C yr -1 in China. Geographically, annual Rs showed a clear spatially increasing trend from northeast to southwest. Forest type is an important factor in determining the soil respiration rate. Bamboo and Evergreen broadleaf forests were higher than other types of forests. In conclusion, these results provide a unique insight into the magnitudes and mechanisms of soil CO 2 emissions in China's forest ecosystems.

58 GEOSCIENCES↗

SU-D-204-06: Integration of Machine Learning and Bioinformatics Methods to Analyze Genome-Wide Association Study Data for Rectal Bleeding and Erectile Dysfunction Following Radiotherapy in Prostate Cancer

Purpose: We investigated whether integration of machine learning and bioinformatics techniques on genome-wide association study (GWAS) data can improve the performance of predictive models in predicting the risk of developing radiation-induced late rectal bleeding and erectile dysfunction in prostate cancer patients. Methods: We analyzed a GWAS dataset generated from 385 prostate cancer patients treated with radiotherapy. Using genotype information from these patients, we designed a machine learning-based predictive model of late radiation-induced toxicities: rectal bleeding and erectile dysfunction. The model building process was performed using 2/3 of samples (training) and the predictive model was tested with 1/3 of samples (validation). To identify important single nucleotide polymorphisms (SNPs), we computed the SNP importance score, resulting from our random forest regression model. We performed gene ontology (GO) enrichment analysis for nearby genes of the important SNPs. Results: After univariate analysis on the training dataset, we filtered out many SNPs with p>0.001, resulting in 749 and 367 SNPs that were used in the model building process for rectal bleeding and erectile dysfunction, respectively. On the validation dataset, our random forest regression model achieved the area under the curve (AUC)=0.70 and 0.62 for rectal bleeding and erectile dysfunction, respectively. We performed GO enrichment analysis for the top 25%, 50%, 75%, and 100% SNPs out of the select SNPs in the univariate analysis. When we used the top 50% SNPs, more plausible biological processes were obtained for both toxicities. An additional test with the top 50% SNPs improved predictive power with AUC=0.71 and 0.65 for rectal bleeding and erectile dysfunction. A better performance was achieved with AUC=0.67 when age and androgen deprivation therapy were added to the model for erectile dysfunction. Conclusion: Our approach that combines machine learning and bioinformatics techniques enabled designing better models and identifying more plausible biological processes associated with the outcomes.

60 APPLIED LIFE SCIENCES↗

Feature-Based PMU Event Classification under Variable PMU Participation and Overlapping Events

Danovo Energy Solution's presented its paper named: Feature-Based PMU Event Classification under Variable PMU Participation and Overlapping Events at the 2026 Georgia Tech Fault & Disturbance Analysis Conference. The full paper can be found at OSTI ID# 3169150 Paper Abstract—Phasor Measurement Units (PMUs) stream time synchronized, high-resolution measurements from the grid, enabling data-driven techniques for event detection and classification. Accurate event classification improves grid reliability and stability. Events can be detected by varying numbers of PMUs and exhibit different durations depending on the event type. This variability challenges standard classifiers that require uniform input sizes. Moreover, multiple events may coincide, which increases classification complexity. Standard classifiers assign each instance to the class with the highest predicted probability, whereas overlapping events may exhibit comparable probabilities across multiple classes. In this study, to handle data size variability, we extract a wide range of time–frequency domain features from all available PMUs for each event into a fixed-length vector, facilitating the application of standard machine learning classifiers, including Random Forest, XGBoost, LightGBM, Support Vector Machine, and Multilayer Perceptron. To account for overlapping events, a probabilistic post-processing step is applied. For a given data instance, if multiple predicted class probabilities exceed 30% and the differences between them are less than 10%, the event is assigned to multiple classes. Experiments using real-world PMU data demonstrate that the Random Forest and XGBoost models achieve the highest accuracy, while the proposed post-processing method yields perfect classification performance on external unseen test sets.

Nematirad, Reza [Danova Energy Solutions]↗

Improving Medication Regimen Recommendation for Parkinson’s Disease Using Sensor Technology

Parkinson’s disease medication treatment planning is generally based on subjective data obtained through clinical, physician-patient interactions. The Personal KinetiGraph™ (PKG) and similar wearable sensors have shown promise in enabling objective, continuous remote health monitoring for Parkinson’s patients. In this proof-of-concept study, we propose to use objective sensor data from the PKG and apply machine learning to cluster patients based on levodopa regimens and response. The resulting clusters are then used to enhance treatment planning by providing improved initial treatment estimates to supplement a physician’s initial assessment. We apply k-means clustering to a dataset of within-subject Parkinson’s medication changes—clinically assessed by the MDS-Unified Parkinson’s Disease Rating Scale-III (MDS-UPDRS-III) and the PKG sensor for movement staging. A random forest classification model was then used to predict patients’ cluster allocation based on their respective demographic information, MDS-UPDRS-III scores, and PKG time-series data. Clinically relevant clusters were partitioned by levodopa dose, medication administration frequency, and total levodopa equivalent daily dose—with the PKG providing similar symptomatic assessments to physician MDS-UPDRS-III scores. A random forest classifier trained on demographic information, MDS-UPDRS-III scores, and PKG time-series data was able to accurately classify subjects of the two most demographically similar clusters with an accuracy of 86.9%, an F1 score of 90.7%, and an AUC of 0.871. A model that relied solely on demographic information and PKG time-series data provided the next best performance with an accuracy of 83.8%, an F1 score of 88.5%, and an AUC of 0.831, hence further enabling fully remote assessments. These computational methods demonstrate the feasibility of using sensor-based data to cluster patients based on their medication responses with further potential to assist with medication recommendations.

59 BASIC BIOLOGICAL SCIENCES↗

Feature-Based PMU Event Classification under Variable PMU Participation and Overlapping Events

This paper is the basis for a presentation help at the 2026 Georgia Tech Fault & Disturbance Analysis Conference, which can be found at OSTI # 3168287 Paper Abstract—Phasor Measurement Units (PMUs) stream time synchronized, high-resolution measurements from the grid, enabling data-driven techniques for event detection and classification. Accurate event classification improves grid reliability and stability. Events can be detected by varying numbers of PMUs and exhibit different durations depending on the event type. This variability challenges standard classifiers that require uniform input sizes. Moreover, multiple events may coincide, which increases classification complexity. Standard classifiers assign each instance to the class with the highest predicted probability, whereas overlapping events may exhibit comparable probabilities across multiple classes. In this study, to handle data size variability, we extract a wide range of time–frequency domain features from all available PMUs for each event into a fixed-length vector, facilitating the application of standard machine learning classifiers, including Random Forest, XGBoost, LightGBM, Support Vector Machine, and Multilayer Perceptron. To account for overlapping events, a probabilistic post-processing step is applied. For a given data instance, if multiple predicted class probabilities exceed 30% and the differences between them are less than 10%, the event is assigned to multiple classes. Experiments using real-world PMU data demonstrate that the Random Forest and XGBoost models achieve the highest accuracy, while the proposed post-processing method yields perfect classification performance on external unseen test sets.

Nematirad, Reza [Danovo Energy Solutions]↗

SU-D-207B-02: Early Grade Classification in Meningioma Patients Combining Radiomics and Semantics Data

Purpose: The clinical management of meningioma is guided by its grade and biologic behavior. Currently, diagnosis of tumor grade follows surgical resection and histopathologic review. Reliable techniques for pre-operative determination of tumor behavior are needed. We investigated the association between imaging features extracted from preoperative gadolinium-enhanced T1-weighted MRI and meningioma grade. Methods: We retrospectively examined the pre-operative MRI for 139 patients with de novo WHO grade I (63%) and grade II (37%) meningiomas. We investigated the predictive power of ten semantic radiologic features as determined by a neuroradiologist, fifteen radiomic features, and tumor location. Conventional (volume and diameter) imaging features were added for comparison. AUC was computed for continuous and χ{sup 2} for discrete variables. Classification was done using random forest. Performance was evaluated using cross validation (1000 iterations, 75% training and 25% validation). All p-values were adjusted for multiple testing. Results: Significant association was observed between meningioma grade and tumor location (p<0.001) and two semantic features including intra-tumoral heterogeneity (p<0.001) and overt hemorrhage (p=0.01). Conventional (AUC 0.61–0.67) and eleven radiomic (AUC 0.60–0.70) features were significant from random (p<0.05, Noether test). Median AUC values for classification of tumor grade were 0.57, 0.71, 0.72 and 0.77 respectively for conventional, radiomic, location, and semantic features after using random forest. By combining all imaging data (semantic, radiomic, and location), the median AUC was 0.81, which offers superior predicting power to that of conventional imaging descriptors for meningioma as well as radiomic features alone (p<0.05, permutation test). Conclusion: We demonstrate a strong association between radiologic features and meningioma grade. Pre-operative prediction of tumor behavior based on imaging features offers promise for guiding personalized medicine and improving patient management.

60 APPLIED LIFE SCIENCES↗