Chronic Kidney Disease Detection from Clinical Data using CNN

Chronic Kidney Disease (CKD) is a concerning health issue worldwide as it affects a huge population with a high mortality rate. CKD patients are at increased risk of developing adverse effects such as anemia, bone diseases, cardiac disorders and hormonal problems. Since the loss of renal function occurs gradually and its symptoms are devoid, advanced technologies are needed to find the patterns and relationships in medical data for early diagnosis. This work aims to focus on detecting CKD from clinical data using Convolutional Neural Network (CNN) and comparing their findings with various machine learning algorithms. As the data available has some missing values, numerical data are imputed using k-nearest neighbor and categorical data are imputed with the most frequently occurring category. Hence, this article exposes the best method to automatically diagnose CKD from clinical data. The empirical results indicated that CNN outperforms other classifiers, with a promising accuracy of 99.12%.