Chronic Kidney Disease Detection using Imputation-Aware Deep Neural Network

Abstract

Chronic renal disease damages kidney function. Hypertension, diabetes, and cardiovascular disease are associated with chronic kidney disease. Risk factors for chronic kidney disease encompass age, genetic predisposition, hypertension, diabetes, obesity, proteinuria, and dyslipidemia. Tests, blood pressure measurements, and medical imaging can assist deep neural networks in diagnosing chronic renal disease. These models can identify minuscule patterns that are imperceptible to individuals with an accuracy of 97–100%. Artificial intelligence has progressed through deep neural networks, which are capable of processing intricate data and exhibit enhanced accuracy. Deep neural networks for tabular data reconstruct multi-layered connections among data points. Data, statistics, and machine learning experts employ this strategy to evaluate datasets and analyses containing missing data. Imputation substitutes missing data with alternative values. It infers each absent item, analyzes each completed dataset individually, and subsequently amalgamates the results after inputting various values. Imputation-aware deep neural networks effectively manage absent values. Managing real-world datasets with absent data is challenging. This is relevant to numerous enterprises, including the healthcare sector. To safeguard valuable data, these networks employ fundamental imputation to prevent users from removing missing rows or columns. This strategy preserves the sample size of the dataset to ensure the validity of statistical tests. It enables models to assimilate comprehensive data, reducing bias and improving precision.