#6830 APPLICATION OF THE NEURAL NETWORK ALGORITHM IN THE ASSESSMENT OF THE STAGE OF ACUTE KIDNEY DAMAGE AND THE SELECTION OF TREATMENT

Abstract

Abstract Background and Aims Acute kidney injury is a syndrome that occurs in different clinical situations, under the influence of different risk factors, in patients of all ages, especially those in the intensive care unit. The problem of acute kidney damage refers to numerous clinical situations, different disease mechanisms and poor outcomes. In addition, with the current definition of acute kidney injury, only one conventional marker is available for diagnosis in daily clinical practice. To recognize the disease promptly and start treatment, and thereby improving the outcome, the paper presents a model for recognizing acute kidney damage in the stage that requires hemodialysis treatment. Method The study included 86 hospitalized patients with acute renal impairment who were divided according to the stage of renal impairment at admission to hospital treatment into three groups. The assessment of acute renal impairment and the classification of disease stages were based on the diagnostic stages by the K / DIGO group. All examined patients were over 18 years of age. In the first stage of the disease it was 12.79%, in the second stage it was 15.12%, and in the third 72.09% of patients. Test methods included clinical processing, laboratory evaluation, functional tests, and echosonographic examination. Clinical data included demographic, comorbidity and vital parameters, baseline full blood count, biochemistry and renal function tests. Further laboratory investigations were performed as indicated clinically thereafter. Spot blood and urine samples were collected from all patients in the morning after overnight fasting. The neural network was created in feed forward back propagation of the connection between the data of 69 patients with 31 parameters. In pattern recognition, the neural network learns to recognize the conditions in which it was decided to send the patient to dialysis 10 or to send to classical treatment 01. The training function analyzed the patients separated in three special categories 80% training, 10% test and 10% validation. The algorithm of the learning function is a scaling gradient that adapts the input data to the output data. Results The neural network achieved a performance of 0.14322 in 6 epochs. In a test with the data of 17 patients, it was shown that the neural network indicates for which patients dialysis treatment is a better option that fits with AKI stage 3. For patients with the first and second stages of the disease, conservative treatment is a better therapeutic choice. Conclusion This paper deals with a current topic by analyzing data using a neural network algorithm to help assess the stage of the disease in high-risk hospitalized patients with acute kidney injury and select treatment modalities. The conclusions derived from this analysis should help with stage recognition and the selection of treatment modalities for patients with acute kidney injury.