Submission
| Title: | INTELLIGENT HEALTHCARE DATA ANALYTICS FOR HEPATIC STEATOSIS PREDICTION |
| Presenter: | Heiner Castro |
| Institution: | Purdue University |
| Authors: | Heiner Castro, School of Engineering Technology, Purdue University; Suranjan Panigrahi, School of Engineering Technology, Purdue University; |
Abstract
| Background/Significance/Rationale: | The increasing prevalence of Hepatic Steatosis (HS) necessitates the development of effective screening tools to aid clinical decision-making. This study builds upon the work of Deo (2022), which proposed a set of machine learning (ML) algorithms for HS detection using the NHANES III survey dataset. While Deo’s algorithms demonstrated promising performance, they did not incorporate the sample weights recommended for analysis. Our research aims to evaluate the impact of these sample weights on the performance of an HS screening model, utilizing physiological and liver biochemistry parameters. |
| Methods: | We employed a methodology that retained key predictors, including age, BMI, HDL, plasma glucose, AST, ALT, and ASP, while excluding samples from individuals outside specified demographic criteria. To address the imbalanced nature of the dataset, we implemented under-sampling and SMOTE techniques. We trained Support Vector Machine (SVM) models with linear, quadratic, and Gaussian kernels, assessing their performance metrics, including accuracy, sensitivity, and specificity. |
| Results/Findings: | Our findings indicate that while the quadratic kernel SVM with weights achieved comparable performance to models without weights, the introduction of sample weights improved sensitivity but reduced specificity. |
| Conclusions/Discussion: | Overall, the sample weights did not enhance the performance of the SVM models significantly. Future work will explore additional algorithms tested by Deo (2022) to further improve HS prediction outcomes |
| Translational/Human Health Impact: | This research contributes to the ongoing efforts to refine clinical decision support tools for the early detection of Hepatic Steatosis, ultimately aiming to improve patient outcomes in healthcare settings. |
