Volume 5, Issue 1, Pp19-28, 2022

APPLICATION OF ARTIFICIAL INTELLIGENCE MULTIMODAL ULTRASONOGRAPHY FOR PREDICTING THE RISK OF ACUTE KIDNEY INJURY IN SEPSIS

Piotr Materzok

University of Wrocław,Poland

Abstract

The occurrence of acute kidney injury in sepsis represents a common complication in hospitalized and critically injured patients that is usually associated with negative prognosis, which presents additional consequences in terms of the risk of developing chronic kidney disease coupled with significantly higher mortality. To intervene prematurely in high-risk patients, improve poor prognosis and further enhance the success rate of resuscitation in severe cases, a diagnostic grading standard of acute kidney injury is employed as the basis in this paper. Firstly, an artificial intelligence multimodal ultrasound imaging technique is conceived by incorporating conventional ultrasound, ultrasonography and shear wave elastography examination approaches. Secondly, the acquired focal lesion images in the kidney lumen are mapped into a knowledge map and then injected into feature mining of a multicenter clinical dataset to accomplish risk prediction for the occurrence of acute kidney injury. Eventually, the clinical decision curve demonstrated that the application of the prediction model can provide net benefits to patients in the range of 0.017 to 0.89 for threshold probability fluctuations. Additionally, the values of model sensitivity, specificity, accuracy, and AUC are 67.9%, 82.48%, 76.86%, and 0.692, respectively, which confirms out that multimodal ultrasonography not only improves the diagnostic sensitivity of the model, but also dramatically raises the risk prediction ability, illustrating that the predictive model possesses promising validity and accuracy.

Keywords: Artificial intelligence; multimodal ultrasonography; sepsis; acute kidney injury; risk forecast