Random Forest Classification Model Research Articles

A machine learning approach that measures pH using acidoCEST MRI of iopamidol.

Tumor acidosis is an important biomarker for aggressive tumors, and extracellular pH (pHe) of the tumor microenvironment can be used to predict and evaluate tumor responses to chemotherapy and immunotherapy. AcidoCEST MRI measures tumor pHe by exploiting the pH-dependent chemical exchange saturation transfer (CEST) effect of iopamidol, an exogenous CT agent repurposed for CEST MRI. However, all pH fitting methodologies for acidoCEST MRI data have limitations. Herein we present results of the application of machine learning for extracting pH values from CEST Z-spectra of iopamidol. We acquired 36,000 experimental CEST spectra from 200 phantoms of iopamidol prepared at five concentrations, five T1 values, and eight pH values at five temperatures, acquired at six saturation powers and six saturation times. We also acquired T1 , T2 , B1 RF power, and B0 magnetic field strength supplementary MR information. These MR images were used to train and validate machine learning models for the tasks of pH classification and pH regression. Specifically, we tested the L1-penalized logistic regression classification (LRC) model and the random forest classification (RFC) model for classifying the CEST Z-spectra for thresholds at pH 6.5 and 7.0. Our results showed that both RFC and LRC were effective for pH classification, although the RFC model achieved higher predictive value, and improved the accuracy of classification accuracy with CEST Z-spectra with a more limited set of saturation frequencies. Furthermore, we used LASSO and random forest regression (RFR) models to explore pH regression, which showed that the RFR model achieved higher accuracy and precision for estimating pH across the entire pH range of 6.2-7.3, especially when using a more limited set of features. Based on these results, machine learning for analysis of acidoCEST MRI is promising for eventual in vivo determination of tumor pHe.

A prediction model for thermal conductivity of metallic nuclear fuel based on multiple machine learning models

In this work, the presence of the BCC phase and its thermal conductivity in uranium-based metallic nuclear fuels at higher temperatures were predicted by coupling a random forest classification model with a multilayer perceptron network model. To analyze the effect of phase constitution on thermal conductivity in uranium-based alloys, the pearson correlation coefficient was introduced. The data samples in the current small-sample database of metallic nuclear fuels, created by collecting experimental and computational data from the literature, were first pre-processed with the guide of professional knowledge before machine learning, involving data screening and data completion. Based on temperature and composition characteristics, the existence of the BCC phases and the thermal conductivities of the candidates were predicted by our models. The optimized thermal conductivities of the candidate alloys coincide well with the experimental results, which shows the reliability of our model.

KEdge: Fuzzy-Based Multi-AI Model Coalescence Solution for Mobile Healthcare System

In this work, we propose a condition-aware analytical framework—KEdge—for health condition recommendation in Internet of Things (IoT) based mobile healthcare systems. Procuring data from multiple sensors and making a singular assessment from them is a challenging task. KEdge overcomes such an issue by determining the severity of the patient by determining a condition index (CI) using a two-step analytical framework and a multiple rule fuzzy inference system (FIS). In the first step, KEdge detects the heart severity condition using a convolutional neural network model and, in the second step, it detects the respiratory condition using a random forest classification model. KEdge also utilizes auscultation sounds from SkopEdge (a digital stethoscope) for assessing the heartbeats. Through extensive experiments, we observe that KEdge identifies the arrhythmia condition with an accuracy of 98.53% and respiratory condition by 98.68%. KEdge considers the analytical predictions and analysis from SkopEdge to evaluate the CI for recommending the overall health condition using Mamdani FIS. We observe that KEdge is suitable for resource-constrained IoT devices providing memory consumption of 6.6%. On offloading the same to the fog nodes, we observe improved CPU utilization with data upload rates in the order of 25 kb/s (KEdge) and 5 Mb/s (SkopEdge).

A powerful random forest-based pipeline for accurately identifying the structural variants on panel sequencing data.

e18898 Background: Numerous studies have reported that the identification of structural variants (SVs) can provide ideal targets for cancer targeting therapy. However, on the one hand, the current algorithms exhibited different performances in detecting different types and sizes of SVs; on the other hand, the whole-genome/exome sequencing for detecting SVs is costly, and only a portion of SVs can be used to guide the clinical treatment of cancer patients. Therefore, developing a powerful, highly accurate, and cost-effective pipeline to detect SVs for clinical application is urgently required. Methods: Based on simulated sequencing data, four common SV detection tools (Delly, Lumpy, SvABA, and Manta) were used to select the best performance one. This tool will be used for detecting SVs in panel sequencing data. Then, Integrative Genomics Viewer (IGV) was used to annotate these SVs as true positive (TP) and false positive (FP) SVs. These annotated TP and FP SVs were used as input in the random forest classifier (RFC) to train a model for predicting the true or false positive SVs. Two independent testing cohorts and standards were used to validate this constructed pipeline. In addition, IHC/FISH experiments were performed to further validate these predicted TP SVs. Results: In the simulation data, the detection tool with the highest SVs identification sensitivity was Delly. By Delly, a total of 1,303 SVs in 384 tumor samples were detected. Based on annotated TP and FP SVs, an RFC model was constructed in the training cohort, which predicted 334 and 560 TP and FP SVs, respectively. The accuracy rates (AR) were 99.85% and 100% of TP and FP SVs, respectively. The predicted TP and FP SVs by constructed pipeline also showed a high accuracy rate in cohort 1 (98.96% and 99.24% of TP and FP SVs, respectively), cohort 2 (98.48% and 100% of TP and FP SVs, respectively), and standards (the mean accuracy rate was 90%). Finally, based on these TP SVs predicted by the constructed pipeline, FISH/IHC experiment results further verified the robustness of this pipeline (the man accuracy rate was 95.83%). Conclusions: The constructed random forest-based pipeline can robustly and accurately identify SVs based on panel sequencing data, which will be helpful for clinical decision-making.

A quantitative analysis of the impact of COVID‐19 quarantining on in‐home eating and drinking habits in a cohort of people living with dementia

AbstractBackgroundSince the start of the coronavirus pandemic in 2019 (COVID‐19), people living with dementia (PLWD) have reported experiencing changes in their in‐home eating and drinking habits during quarantine. Previous research has been conducted solely through questionnaires, making it subject to researcher bias. However, remote monitoring technologies allow us the unprecedented ability to quantify behavioural patterns of PLWD in real‐world environments. We used data collected as part of an ongoing project at the UK Dementia Research Institute’s Care Research and Technology Centre. This work aims to understand better the effects of COVID‐19 quarantining on the in‐home eating and drinking habits of PLWD.MethodUsing COVID‐19 as a natural experiment, we compared data collated during the UK’s first COVID‐19 lockdown with the same period pre‐COVID. Household activity was observed using kitchen motion and appliance usage (refrigerator door, kettle, and oven) sensors, summed across four different daily periods (22:00, 04:00, 10:00, and 16:00), and then standardized across households by sensory device, resulting in 16 features. With this dataset, we trained a binary random forest classification model for each household independently, assessing performance using 10‐fold cross‐validation and also computing permutation feature importances (PFIs). Finally, using the PFIs of each household as a behavioural pattern reflecting in‐home eating and drinking habit changes as a function of quarantine, we used agglomerative hierarchical clustering (metric = correlation; method = ward) to identify similar behavioural changes across households.ResultOur cohort included 25 households of PLWD. All patients (11 female : 14 male; mean age at study start = 80.3 ± 6.78 years) had established diagnoses of dementia. Across households, binary classification accuracy was high (mean = 0.88 ± 0.08), with precision and recall balanced well (mean F1‐score = 0.83 ± 0.14). Hierarchical clusters of the household PFIs produced 4 main groupings that simple statistical observations showed each relied on distinct features for binary classification.ConclusionApplying combined supervised and unsupervised learning approaches to in‐home monitoring data has the potential to dramatically improve research for dementia care. Our findings show that in‐home eating and drinking habit changes among PLWD, in response to the COVID‐19 quarantine, are complex.

Early Warning Method of Transformer Core Clamp Disconnection Based on Random Forest Algorithm

Transformer core fault is one of the common faults, and core fault monitoring and early warning is an important research topic. An early warning method of transformer core clamp disconnection based on a random forest algorithm is proposed. Firstly, the values of transformer core clamp parameters under different working conditions are measured, and the parameter-state data set is established. Secondly, the features are processed and reduced in dimension, and the parameters with high correlation are selected. Finally, a random forest classification model training parameter-state training set is established to identify the fault of the iron core clamp.

Landscape context importance for predicting forest transition success in central Panama

ContextNaturally recovering secondary forests are frequently re-cleared before they can recover to pre-disturbance conditions. Identifying landscape factors associated with persistence success will help planning cost-efficient and effective forest restoration.ObjectivesThe ability of secondary forest to persist is an often undervalued requisite for long-term ecosystem restoration. Here we identify the landscape context for naturally regenerated forests to persist through time within central Panama.MethodsWe developed a random forest classification (RFC) calibration method to identify areas with high (≥ 90%) and low (< 90%) likelihood of forest persistence success based on their spatial relation with nine landscape explanatory variables.ResultsThe RFC model discriminated between secondary forests areas that persisted and did not persisted with an error rate of 2%. By tuning, we obtained a precision of 0.94 (94%) in the validation test. The two most important explanatory variables involved in the persistence dynamic were elevation and distance to the nearest rural area. Naturally regenerated forests lasted longer in patches that were closer to both Gatun and Alajuela Lakes as to protected areas, but further from rural communities, roads, urban areas and in patches with higher elevation and steeper slopes.ConclusionBy tracking remote sensed, landscape context metrics of easy collection, we developed a prediction map of central Panama areas with high (≥ 90%) and low (> 90%) probability of natural forest regeneration and persistence success within the next 30 years. This map represents a basis for management decisions and future investigations for effective, long-term forest-landscape restoration.

Metagenomic analysis reveals gut plasmids as diagnosis markers for colorectal cancer.

Colorectal cancer (CRC) is linked to distinct gut microbiome patterns. The efficacy of gut bacteria as diagnostic biomarkers for CRC has been confirmed. Despite the potential to influence microbiome physiology and evolution, the set of plasmids in the gut microbiome remains understudied. We investigated the essential features of gut plasmid using metagenomic data of 1,242 samples from eight distinct geographic cohorts. We identified 198 plasmid-related sequences that differed in abundance between CRC patients and controls and screened 21 markers for the CRC diagnosis model. We utilize these plasmid markers combined with bacteria to construct a random forest classifier model to diagnose CRC. The plasmid markers were able to distinguish between the CRC patients and controls [mean area under the receiver operating characteristic curve (AUC = 0.70)] and maintained accuracy in two independent cohorts. In comparison to the bacteria-only model, the performance of the composite panel created by combining plasmid and bacteria features was significantly improved in all training cohorts (mean AUCcomposite = 0.804 and mean AUCbacteria = 0.787) and maintained high accuracy in all independent cohorts (mean AUCcomposite = 0.839 and mean AUCbacteria = 0.821). In comparison to controls, we found that the bacteria-plasmid correlation strength was weaker in CRC patients. Additionally, the KEGG orthology (KO) genes in plasmids that are independent of bacteria or plasmids significantly correlated with CRC. We identified plasmid features associated with CRC and showed how plasmid and bacterial markers could be combined to further enhance CRC diagnosis accuracy.

Establishment of a Machine Learning Model for the Risk Assessment of Perineural Invasion in Head and Neck Squamous Cell Carcinoma.

Perineural invasion is a prevalent pathological finding in head and neck squamous cell carcinoma and a risk factor for unfavorable survival. An adequate diagnosis of perineural invasion by pathologic examination is limited due to the availability of tumor samples from surgical resection, which can arise in cases of definitive nonsurgical treatment. To address this medical need, we established a random forest prediction model for the risk assessment of perineural invasion, including occult perineural invasion, and characterized distinct cellular and molecular features based on our new and extended classification. RNA sequencing data of head and neck squamous cell carcinoma from The Cancer Genome Atlas were used as a training cohort to identify differentially expressed genes that are associated with perineural invasion. A random forest classification model was established based on these differentially expressed genes and was validated by inspection of H&E-stained whole image slides. Differences in epigenetic regulation and the mutational landscape were detected by an integrative analysis of multiomics data and single-cell RNA-sequencing data were analyzed. We identified a 44-gene expression signature related to perineural invasion and enriched for genes mainly expressed in cancer cells according to single-cell RNA-sequencing data. A machine learning model was trained based on the expression pattern of the 44-gene set with the unique feature to predict occult perineural invasion. This extended classification model enabled a more accurate analysis of alterations in the mutational landscape and epigenetic regulation by DNA methylation as well as quantitative and qualitative differences in the cellular composition in the tumor microenvironment between head and neck squamous cell carcinoma with or without perineural invasion. In conclusion, the newly established model could not only complement histopathologic examination as an additional diagnostic tool but also guide the identification of new drug targets for therapeutic intervention in future clinical trials with head and neck squamous cell carcinoma patients at a higher risk for treatment failure due to perineural invasion.

MWD Data-Based Marble Quality Class Prediction Models Using ML Algorithms

Brønnøy Kalk AS operates an open pit mine in Norway producing marble, mainly used by the paper industry. The final product is used as filler and pigment for paper production. Therefore, the quality of the product has utmost importance. In the mine, the primary quality indicator, called TAPPI, is quantified through a laborious sampling process and laboratory experiments. As a part of digital transformation, measurement while drilling (MWD) data have been collected in the mine. The purpose of this paper is to use the recorded MWD data for the prediction of marble quality to facilitate quality blending in the pit. For this purpose, two supervised classification modelling algorithms such as conventional logistic regression and random forest have been employed. The results show that the random forest classification model presents significantly higher statistical performance, and it can be used as a tool for fast and efficient marble quality assessment.

Accident severity prediction modeling for road safety using random forest algorithm: an analysis of Indian highways

Background: Road accidents claim around 1.35 million lives annually, with countries like India facing a significant impact. In 2019, India reported 449,002 road accidents, causing 151,113 deaths and 451,361 injuries. Accident severity modeling helps understand contributing factors and develop preventive strategies. AI models, such as random forest, offer adaptability and higher predictive accuracy compared to traditional statistical models. This study aims to develop a predictive model for traffic accident severity on Indian highways using the random forest algorithm. Methods: A multi-step methodology was employed, involving data collection and preparation, feature selection, training a random forest model, tuning parameters, and evaluating the model using accuracy and F1 score. Data sources included MoRTH and NHAI. Results: The classification model had hyperparameters ’max depth’: 10, ’max features’: ’sqrt’, and ’n estimators’: 100. The model achieved an overall accuracy of 67% and a weighted average F1-score of 0.64 on the training set, with a macro average F1-score of 0.53. Using grid search, a random forest Classifier was fitted with optimal parameters, resulting in 41.47% accuracy on test data. Conclusions: The random forest classifier model predicted traffic accident severity with 67% accuracy on the training set and 41.47% on the test set, suggesting possible bias or imbalance in the dataset. No clear patterns were found between the day of the week and accident occurrence or severity. Performance can be improved by addressing dataset imbalance and refining model hyperparameters. The model often underestimated accident severity, highlighting the influence of external factors. Adopting a sophisticated data recording system in line with MoRTH and IRC guidelines and integrating machine learning techniques can enhance road safety modeling, decision-making, and accident prevention efforts.

Public support for phasing out carbon-intensive technologies: the end of the road for conventional cars in Germany?

ABSTRACT Limiting the global mean temperature increase to 1.5°C requires phasing out fossil fuel combustion almost entirely within the next three decades and replacing carbon-intensive technologies with low-carbon alternatives. Such socio-technical transitions are politically feasible only if public acceptance is sufficiently high. Here we investigate German citizens’ views on the phase-out of internal combustion engine vehicles (ICEVs) using a random forest (decision trees) classification and logistic regression model. We surveyed a demographically representative sample (N = 1,663) in 2021, finding that the majority of respondents (67%) approve of an ICEV phase-out by 2040 or hold a neutral stance. Acceptability is best predicted by the degree to which environmental problems are attributed to ICEVs, followed by respondents’ willingness to abandon cars altogether or adopt electric vehicles (EVs). Our results further indicate that acceptability can be increased by providing people with information that present EVs in a more favourable, and ICEVs in a less favourable light. When the European Commission proposed to ban the sale of ICEVs by 2035, we conducted a follow-up survey to investigate whether this had influenced acceptability in Germany – with the result that it had not. In terms of concrete policies, pull measures such as public transport or electric vehicle purchase subsidies are preferred by the public over more restrictive policies such as taxes or bans. The findings of this study shed light on different dimensions of public opinion and their important implications for policymaking and the political feasibility of this socio-technical transition. Insights from this research can help policymakers in designing effective yet widely acceptable transport decarbonization policies.

Profiling low-proficiency science students in the Philippines using machine learning

Filipino students’ performance in global assessments of science literacy has always been low, and this was confirmed again in the PISA 2018, where Filipino learners’ average science literacy scores ranked second to last among 78 countries. In this study, machine learning approaches were used to analyze PISA data from the student questionnaire to test models that best identify the poorest-performing Filipino students. The goal was to explore factors that could help identify the students who are vulnerable to very low achievement in science and that could indicate possible targets for reform in science education in the Philippines. The random forest classifier model was found to be the most accurate and more precise, and Shapley Additive Explanations indicated 15 variables that were most important in identifying the low-proficiency science students. The variables related to metacognitive awareness of reading strategies, social experiences in school, aspirations and pride about achievements, and family/home factors, include parents’ characteristics and access to ICT with internet connections. The results of the factors highlight the importance of considering personal and contextual factors beyond the typical instructional and curricular factors that are the foci of science education reform in the Philippines, and some implications for programs and policies for science education reform are suggested.

Rainfall-driven machine learning models for accurate flood inundation mapping in Karachi, Pakistan

Urban pluvial flooding (UPF) has emerged as a serious natural hazard, especially in recent years. Previous research on UPF prediction has mainly focused on hydrological models, which required a large amount of data. However, a data-driven method can significantly reduce the computational cost by using rainfall amounts to predict pluvial flooding. Intensity-duration-frequency (IDF) curves using the Gumbel method can provide a better interpretation of the correlation between rainfall intensity, duration, and probability of occurrence of a given rainfall amount. In this study, machine learning models (ML) for rainfall amounts were used to identify flood points in a case study conducted in Karachi, Pakistan. Thirteen inundation factors were used for the ML models, including a new factor, curve number. Ten ML models were applied first on training and then on validation data, yielding the inundation points. The training and validation process of the model included 384 flood points. Several statistics were used to verify the performance and accuracy of the model. We found that the Light Gradient Boost Machine and Random Forest Classifier models were the most accurate in training and validating the model, while the Decision Tree and K-Nearest Neighbor models were the least accurate in training and validating the model. The study provides valuable information for decision makers to protect communities from flood hazards by incorporating the likely intensity and duration of rainfall events and carefully selecting influencing factors into flood event prediction models.

Artificial intelligence-based stratification of demographic, ocular surface high-risk factors in progression of keratoconus.

The purpose of this study was to identify and analyze the clinical and ocular surface risk factors influencing the progression of keratoconus (KC) using an artificial intelligence (AI) model. This was a prospective analysis in which 450 KC patients were included. We used the random forest (RF) classifier model from our previous study (which evaluated longitudinal changes in tomographic parameters to predict "progression" and "no progression") to classify these patients. Clinical and ocular surface risk factors were determined through a questionnaire, which included presence of eye rubbing, duration of indoor activity, usage of lubricants and immunomodulator topical medications, duration of computer use, hormonal disturbances, use of hand sanitizers, immunoglobulin E (IgE), and vitamins D and B12 from blood investigations. An AI model was then built to assess whether these risk factors were linked to the future progression versus no progression of KC. The area under the curve (AUC) and other metrics were evaluated. The tomographic AI model classified 322 eyes as progression and 128 eyes as no progression. Also, 76% of the cases that were classified as progression (from tomographic changes) were correctly predicted as progression and 67% of cases that were classified as no progression were predicted as no progression based on clinical risk factors at the first visit. IgE had the highest information gain, followed by presence of systemic allergies, vitamin D, and eye rubbing. The clinical risk factors AI model achieved an AUC of 0.812. This study demonstrated the importance of using AI for risk stratification and profiling of patients based on clinical risk factors, which could impact the progression in KC eyes and help manage them better.

Predicting patient-reported outcomes following lumbar spine surgery: development and external validation of multivariable prediction models

BackgroundThis study aimed to develop and externally validate prediction models of spinal surgery outcomes based on a retrospective review of a prospective clinical database, uniquely comparing multivariate regression and random forest (machine learning) approaches, and identifying the most important predictors.MethodsOutcomes were change in back and leg pain intensity and Core Outcome Measures Index (COMI) from baseline to the last available postoperative follow-up (3–24 months), defined as minimal clinically important change (MCID) and continuous change score. Eligible patients underwent lumbar spine surgery for degenerative pathology between 2011 and 2021. Data were split by surgery date into development (N = 2691) and validation (N = 1616) sets for temporal external validation. Multivariate logistic and linear regression, and random forest classification and regression models, were fit to the development data and validated on the external data.ResultsAll models demonstrated good calibration in the validation data. Discrimination ability (area under the curve) for MCID ranged from 0.63 (COMI) to 0.72 (back pain) in regression, and from 0.62 (COMI) to 0.68 (back pain) in random forests. The explained variation in continuous change scores spanned 16%-28% in linear, and 15%-25% in random forests regression. The most important predictors included age, baseline scores on the respective outcome measures, type of degenerative pathology, previous spinal surgeries, smoking status, morbidity, and duration of hospital stay.ConclusionsThe developed models appear robust and generalisable across different outcomes and modelling approaches but produced only borderline acceptable discrimination ability, suggesting the need to assess further prognostic factors. External validation showed no advantage of the random forest approach.

Loan Risk Prediction Model based on Random Forest

As people's consumption habits change, loan plays a crucial role in our modern society. It provides individuals who do not have sufficient money with funds to purchase residential property or start a business. However, for avoiding unpleasant loan defaults, all financial institutions will first assess the borrower's risk index. By predicting the default risk of the borrower to decide whether to lend money. Machine learning algorithms, including random forest, linear regression and so on, have been benefited most of the real-world applications. With the development of machine learning methods, this paper, based on the personal history loan data of an institution studies the loan default risk, and uses the random forest classification model to predict the possibility of loan default. The result showed that the accuracy of this method was 85.62%, which show its application ability of real-world loan prediction and benefits the manager to decide the degree of risk for loan grant.

Distinguishing Loss-of-Function and Gain-of-Function SCN8A Variants Using a Random Forest Classification Model Trained on Clinical Features.

Pathogenic variants at the voltage-gated sodium channel gene, SCN8A, are associated with a wide spectrum of clinical disease outcomes. A critical challenge for neurologists is to determine whether patients carry gain-of-function (GOF) or loss-of-function (LOF) variants to guide treatment decisions, yet in vitro studies to infer channel function are often not feasible in the clinic. In this study, we develop a predictive modeling approach to classify variants based on clinical features present at initial diagnosis. We performed an exhaustive search for individuals deemed to carry SCN8A GOF and LOF variants by means of in vitro studies in heterologous cell systems, or because the variant was classified as truncating, and recorded clinical features. This resulted in a total of 69 LOF variants: 34 missense and 35 truncating variants, including 9 nonsense, 13 frameshift, 6 splice site, 6 indels, and 1 large deletion. We then assembled a truth set of variants with known functional effects, excluding individuals carrying variants at other loci associated with epilepsy. We then trained a predictive model based on random forest using this truth set of 45 LOF variants and 45 GOF variants randomly selected from a set of variants tested by in vitro methods. Phenotypic categories assigned to individuals correlated strongly with GOF or LOF variants. All patients with GOF variants experienced early-onset seizures (mean age at onset = 4.5 ± 3.1 months) while only 64.4% patients with LOF variants had seizures, most of which were late-onset absence seizures (mean age at onset = 40.0 ± 38.1 months). With high accuracy (95.4%), our model including 5 key clinical features classified individuals with GOF and LOF variants into 2 distinct cohorts differing in age at seizure onset, development of seizures, seizure type, intellectual disability, and developmental and epileptic encephalopathy. The results support the hypothesis that patients with SCN8A GOF and LOF variants represent distinct clinical phenotypes. The clinical model developed in this study has great utility because it provides a rapid and highly accurate platform for predicting the functional class of patient variants during SCN8A diagnosis, which can aid in initial treatment decisions and improve prognosis.

Identification of diagnostic biomarkers in Alzheimer’s disease using metabolic-pathway related genes. (P10-6.011)

<h3>Objective:</h3> In this study, we aim to study the role of metabolic pathways in AD and to conduct a bioinformatics analysis and gene ontology to understand the candidate metabolic pathway genes in the pathogenesis of AD, and to construct a machine learning diagnostic model using those genes. <h3>Background:</h3> Alzheimer’s disease (AD) is the most common form of dementia. It is known that the metabolic pathways are altered in AD. So, we investigated their mechanisms to aid in finding a therapeutic targets for patients with AD. <h3>Design/Methods:</h3> We used (GSE159541, GSE104704, GSE15222, GSE5281) datasets from Gene Expression Omnibus database to find the differentially expressed genes (DEGs). Then we analyzed the metabolic pathway-related genes from the Molecular Signature database (MSigDB) and intersected them with the DEGs. We then constructed a machine learning diagnostic model using Support vector machine (SVM) and recursive feature elimination (RFE) to select the top 5 ranking genes. Then we fitted them in a random forest classification (RFC) to assess their performance. <h3>Results:</h3> We identified 2054 DEGs and intersected them with metabolic pathway-related genes resulting in 72 metabolic genes. Then we fitted them into a SVM and used RFE to choose the top 5 ranked genes which were (ENOPH1, PSMA2, PSMB2, TXN2, UROD). Among those genes, UROD was the highest contributing gene. Those genes were fitted into an RFC which was trained to predict the diagnosis of patients in a dataset of 343 total patients. The performance of the RFC model on the testing set was measured by a mean bootstrap estimate of 0.711 with 95% CI: [0.58–0.83], 10-fold cross-validation of 0.72, and AUC of 0.70. <h3>Conclusions:</h3> We identified 5 metabolic pathway-related genes and constructed a novel diagnostic model. Those results may help in understanding the pathogenesis in AD, which may aid in the creation of therapeutic targets and more widely available biomarkers. <b>Disclosure:</b> Miss Al-Bzour has nothing to disclose. Mr. Ahmed has nothing to disclose. Mr. Al-Khalili has nothing to disclose. Mr. Hamza has nothing to disclose. Ms. Ibrahim has nothing to disclose. Miss Al-Bzour has nothing to disclose.

A multifactorial fall risk assessment system for older people utilizing a low-cost, markerless Microsoft Kinect

Falls among older people are a major health concern. This study aims to develop a multifactorial fall risk assessment system for older people using a low-cost, markerless Microsoft Kinect. A Kinect-based test battery was designed to comprehensively assess major fall risk factors. A follow-up experiment was conducted with 102 older participants to assess their fall risks. Participants were divided into high and low fall risk groups based on their prospective falls over a 6-month period. Results showed that the high fall risk group performed significantly worse on the Kinect-based test battery. The developed random forest classification model achieved an average classification accuracy of 84.7%. In addition, the individual’s performance was computed as the percentile value of a normative database to visualise deficiencies and targets for intervention. These findings indicate that the developed system can not only screen out ‘at risk’ older individuals with good accuracy, but also identify potential fall risk factors for effective fall intervention. Practitioner summary: Falls are the leading cause of injuries in older people. We newly developed a multifactorial fall risk assessment system for older people utilising a low-cost, markerless Kinect. Results showed that the developed system can screen out ‘at risk’ individuals and identify potential risk factors for effective fall intervention.