A study comparing 4 different artificial intelligence (AI) models found that some models were better than others at detecting early-stage chronic kidney disease (CKD), when patients do not experience visible symptoms.
When implementing a machine learning model for the prevention of chronic kidney disease (CKD), decision forest artificial intelligence (AI) outperformed other models at detecting early-stage CKD, according to a recent study.
The study, published in Computational Intelligence and Neuroscience, explored the detection capabilities of 4 AI models to determine whether implementing a machine learning model can allow for diagnosis of CKD in its initial stages and which models are the most accurate.
Patients with early-stage CKD often do not experience visible symptoms, which can make the condition difficult to diagnose in these patients. CKD is considered a high-cost pathology, because it can contribute to a high economic impact on finances of the health care system and can have a dramatic effect on patients' quality of life, especially in regard employment of those who have CKD as well as their families.
Reducing the high mortality rates associated of CKD requires research focused on the early-stages of the disease. Automatic learning can help with early classification, which reduces the time required for diagnosis and allows patients to receive treatment before their CKD progresses to a later, harder-to-treat stage.
“It is necessary that the technical help tools that are based on data can support the decision-making process in the initial diagnoses quickly, with high precision, and at low cost,” wrote the investigators.
The investigators collected anonymized data (373,770 samples) from the Baghdad Renal Clinic in Iraq and processed it in the cloud on the Azure platform, where the sample data was unbalanced. The cross industry standard process for data mining (CRISP-DM) was used as a reference. The data was balanced using the SMOTE technique. After data balancing, the investigators used 4 AI matching algorithms to determine the most effective CKD detector: logistic regression, decision-forest, artificial neural network, and jungle of decisions.
The data set was divided randomly into training sets and test sets. When carrying out the experiment, 70.0% of the data was used as training data and the other 30.0% was reserved to evaluate the model’s efficiency.
Decision forests was the highest performing model, with a 92.2% accuracy and 92.1% completeness value, followed by neural networks (80.6% accuracy; 80.0% completeness) and decision jungle (75.4%% accuracy; 75.0%). The lowest performer was the logistic regression model, with 68.9% accuracy and a 68.9% completeness value.
“Thanks to the models, we can see how changing the characteristics affects the search for the target value with a simple change of column selection or improvements in the data…. Furthermore, this methodology could apply to clinical data of other diseases and pathologies inaccurate medical diagnoses,” the investigators noted.
The study had some limitations, including that there was not a significant data sample available because of the restrictions of medical data and its legal effects that exist in Iraq. The investigators suggested that expansion of the database could reduce the limited generalization error for the model and allow the severity of the disease to be detected.
“Although the validity of this research is internal, since the document corpus is private and cannot be published for other works, it will help interested professionals with machine learning to carry out their studies in the classification area,” the investigators wrote.
Alhamazani KT, Alshudukhi J, Aljaloud S, Abebaw S. Implementation of machine learning models for the prevention of kidney diseases (CKD) or their derivatives. Comput Intell Neurosci. December 30, 2021;2021: 3941978. doi: 10.1155/2021/3941978